Technology Topic Identification and Trend Prediction of New Energy Vehicle Using LDA Modeling

Hu, Renjie; Ma, Wencong; Lin, Weiqiang; Chen, Xiude; Zhong, Zuchang; Zeng, Chuhong

doi:https://doi.org/10.1155/2022/9373911

Complexity

On this page

Abstract Introduction Literature Review Conclusions Data Availability Conflicts of Interest Acknowledgments References Copyright Related Articles

Special Issue

Flow Space and the Complexity of Urban Spatial Networks

View this Special Issue

Research Article | Open Access

Volume 2022 | Article ID 9373911 | https://doi.org/10.1155/2022/9373911

Technology Topic Identification and Trend Prediction of New Energy Vehicle Using LDA Modeling

Renjie Hu,¹Wencong Ma,²Weiqiang Lin,²Xiude Chen,³Zuchang Zhong,¹and Chuhong Zeng⁴

Academic Editor: Zhipeng Tang

Received24 Nov 2021

Revised06 Feb 2022

Accepted21 Feb 2022

Published23 Mar 2022

Abstract

As new energy vehicle (NEV) is the future of automobile development, it is of great significance to dig deeper into the technical topics and development trends of new energy vehicles for accurately understanding the technical trends of the new energy vehicle industry, grasping development opportunities, and scientifically formulating strategic plans. This paper takes the patent texts in the field of new energy vehicles from 2000 to 2020 in the patent database of CNKI as the data source, identifies 25 technical topics implied in the patent texts by using the LDA (Latent Dirichlet Allocation) topic model, analyzes the evolution trend of the 25 technical topics in terms of importance and popularity, and predicts the popularity and development trend of each technical topic in new energy vehicles from 2021 to 2025 by constructing the ARIMA model. The popularity and development trend of each technology topic of new energy vehicles in China from 2021 to 2025 are predicted by constructing ARIMA model. Drawing on quantitative evidence, the study found that there are top five technical topics in terms of importance in this field, namely, Topic 8 (Installation and Fixation), Topic 5 (Heat Dissipation), Topic 14 (Vehicle Data Monitoring), Topic 9 (Charging Pile), and Topic 15 (Damping). From 2014 to 2020, the importances of Topic 5 (Heat Dissipation), Topic 8 (Installation and Fixation), Topic 6 (Electric Drive System), Topic 9 (Charging Pile), and Topic 15 (Damping) are gradually rising. In terms of popularity of technical topics, from 2014 to 2020, the first to fifth topics are Topic 20 (Safety), Topic 8 (Installation and Fixation), Topic 3 (Cable Insulation Materials), Topic 15 (Damping), and Topic 10 (Pump Cooling). Based on the prediction of ARIMA model, it is found that the popularity of these five technical topics is steadily increasing from 2021 to 2025, among which the popularity of Topic 20 (Safety) will increase from 63.58 to 113.07, the largest increase in popularity among all technical topics. The paper provides implications for countries dedicated to developing the new energy vehicle industry.

1. Introduction

Countries around the world have different expressions about new energy vehicles, which are called “alternative fuel vehicles” in the United States, “low pollution vehicles” in Japan, and “new energy vehicles” in China. China has used the terms “electric vehicles,” “low energy consumption and new energy vehicles,” and “new energy vehicles” in its terminology for new energy vehicles. According to the “Entry Management Rules for New Energy Vehicle Manufacturing Enterprises and Products” implemented in China on July 1, 2009, a new energy vehicle refers to a vehicle that uses unconventional automotive fuels as its power source (or uses conventional automotive fuels but with a new vehicle power unit) and integrates advanced technologies in power control and drive of the vehicle to form an advanced technical principle with new technologies and structures [1]. There are five main types of new energy vehicles, namely, fuel cell new energy vehicles, pure electric vehicles, hybrid vehicles, hydrogen engine vehicles, and other new energy vehicles. Since then, the expression “new energy vehicles” has been used uniformly in government documents, enterprises, and academia.

With the air pollution caused by the exhaust emission of fuel vehicles and continuous consumption of nonrenewable resources, it is irreversible to change the dependence on oil and develop clean energy such as water, wind, and light as the driving force of vehicles. In recent years, the world’s major automotive countries have strengthened strategic planning and policy support, multinational automotive enterprises to increase investment in research and development and improve the industrial layout; new energy vehicles have become the direction of transformation and development of the global automotive industry and promote the world’s sustained economic growth of an important engine. Since 2018, the global output of new energy vehicles has been growing rapidly. In 2019, the global output of new energy vehicles was 2.17 million, and, in 2020, the output was 2.55 million. In 2020, the European new energy vehicles market accounted for 43.06% of the global market, followed by 41.27% in China, 10.12% in the United States, and 0.96% in Japan. In order to implement the State Council’s decision to develop strategic emerging industries and strengthen energy conservation and emission reduction and to accelerate the cultivation and development of energy conservation and new energy vehicle industry, the State Council issued the “Energy Conservation and New Energy Vehicle Industry Development Plan (2012–2020)” in 2012. In November 2020, the General Office of the State Council issued the Development Plan for New Energy Vehicle Industry (2021–2035), which calls for the in-depth implementation of the national strategy for developing new energy vehicles, promoting the high-quality and sustainable development of China’s new energy vehicle industry, and accelerating the construction of an automobile power.

Since 2012, China has adhered to the strategic orientation of pure electric drive and made great achievements in the development of new energy automobile industry, becoming one of the important forces in the development and transformation of the world automobile industry. According to the China Association of Automobile Manufacturers (CAAM), the production and sales of new energy vehicles in China were 1.366 million and 1.367 million, respectively, in 2020, up 7.5% and 10.9% year-on-year, with production and sales reaching a record high. Meantime, the development of new energy vehicles in China also faces problems such as weak innovation capability of core technologies [2, 3], infrastructure construction still lags [4], the industrial ecology is not yet sound [5, 6], the quality assurance system needs to be improved [7, 8], and the market competition is increasing [9–11]. Therefore, based on patent literature data, identifying the technical topics of new energy vehicles and predicting the development trend of new energy vehicles are important for tracking the technological frontier of new energy vehicles, grasping technological development opportunities, improving R&D (research and development) efficiency, perfecting industrial ecology, and achieving high-quality development.

The rest of this paper is organized as follows. Section 2 sorts out the literature review. Section 3 is devoted to the description of the relevant methodology. The empirical results are presented in Section 4. Section 5 proposes policy implications and concludes the paper.

2. Literature Review

2.1. Technical Topic Identification and Forecast

As the most important output result in science and technology innovation, patent literature carries 90%–95% of technical information worldwide [12]. Data mining and topic identification of patent literature can not only shorten R&D time and save research expenditure but also help to grasp the frontier of technological innovation and predict technological development trend. There are mainly patent classification method, patent citation method, and text mining method for technology topic identification based on patent literature.

2.1.1. Analysis Method Based on Patent Classification

Patent classification is a simple and generic technology classification system provided based on the technical content revealed by the patent [13]. The main technical subject analyses based on patent classification numbers are statistical analysis and coclassification analysis. Jun used international patent classification numbers as the technical topic of literature in a key technology prediction study [14]. Kostoff et al. used word frequency analysis to identify the technical topic of disruptive technologies [15]. Zhang et al. predicted the future development trend of driverless cars based on the statistics and analysis of key technology patents such as automatic braking, cruise control, and lane keeping [16]. Suzuki et al. used the International Patent Classification (IPC) number cooccurrence method to study technical topic identification and trend prediction [17]. Jeong et al. studied the strength of IPC cooccurrence relationship by Jaccard coefficients and analyzed the main types of technology topics [18]. Lee et al. performed link prediction analysis on the cooccurrence network of IPC to predict possible future emerging technology topics and used topic analysis to extract keywords to identify possible future emerging areas [19]. Huang et al. used association rule analysis method to analyze the IPC cooccurrence in two technology areas, namely, information technology and biotechnology, and analyzed the characteristics of technology topics in terms of support, confidence, and lift [20]. By statistically analyzing the number of applications and the number of IPCs for driverless car patents, they concluded that the technological research in this field has entered a stable state to a certain extent [21].

2.1.2. Analysis Method Based on Patent Citation

The patent citation method analyzes the citation relationships among patent literature and between patent literature and scientific literature and analyzes the track of technical topic evolution by constructing a citation relationship network [22]. Kwon et al. comprehensively analyzed technical topics by constructing a patent citation coupling network and a cocitation network to synthesize the distribution of patents [23]. Choi and Park constructed a patent citation network and used the main path analysis algorithm to identify technical topics [24]. Hsueh and Wang combined patent citation time with patent citation relationship to predict the frontier technology in the field of LCD (Liquid Crystal Display) technology [25]. Geum et al. [26] and Zhai et al. [27] used the knowledge flow of citation networks between technology categories to study technology convergence. Kim and Seol fused the similarity of patented technologies and patent cocitation features to achieve a frontier technology identification method based on patent network analysis by constructing a patent cooperation network [28]. Small et al. combined patent direct citation network and cocitation network to identify novel technology topics by community clustering algorithm [29]. Li and Chen proposed an emerging technology identification model based on patent citation coupled clustering and empirically analyzed the field of nanotechnology [30]. Lee et al. collected Google’s patents on self-driving cars and used social network analysis methods to identify the company’s core patents in this field based on the citation relationships between patents and predicted that Google’s R&D capabilities would be focused on hardware control [31].

2.1.3. Analysis Methods Based on the Content of Patent Texts

Patent text mining can discover potential data patterns and internal relations from a large amount of unstructured textual information and is an important method for technical topic evolution analysis. In 2003, Beli [32] first proposed the LDA (Latent Dirichlet Allocation) topic model, which introduces the Dirichlet prior distribution, based on statistical probability level to express the semantic relationship between words and mine document topics. Qin and Le used LDA model to conduct a study on the variation of oncology domain in terms of topic content and intensity by formulating topic association filtering rules [33]. Yan used LDA model to mine the document dataset in library intelligence domain and identified the technical topics in the domain [34]. Yang and Yang used LDA topic model for topic mining of policy text data in the international climate domain [35]. Liao and Le introduced IPC classification numbers to measure technical topic strength based on LDA modeling and realized a study on three aspects of topic strength, topic content, and technical topic strength [36]. Wang et al. identified technical topics in the field of ocean acidification using the LDA model [37]. Fan et al. proposed four characteristic indicators of novelty, innovation, interdisciplinarity, and high interest based on the LDA topic identification results and based on the characteristics of literature in the medical field to identify frontier topics in the medical field [38]. Yi et al. identified technology topic identification and its structural features in the field of graphene based on LDA model and strategic coordinate map [39]. Shen et al. constructed an IPC cooccurrence network by Apriori algorithm, applied Louvain clustering algorithm to divide the network into different technical communities, and discovered the technical topics of each community based on LDA topic model [40]. Ma et al. identified common technologies in the field of new materials based on patent data, using LDA topic model to extract the implied technology topics in the text [41]. Mi et al. dynamically identified technology topics in 3D printing field based on the use of LDA topic identification model combined with time series analysis [42]. Li and Xuan combined LDA topic model with patent value evaluation index and proposed a quantitative method for mining technological innovation topics [43]. Tang and Qiu used LDA model to obtain topic words from multisource heterogeneous texts and carried out fusion analysis to extract emerging technology topics in the field of intelligent network vehicles [44].

2.2. New Energy Vehicle

In recent years, the number of new energy vehicle patents and research literature has grown rapidly [45, 46]. In the patent database of CNKI (China National Knowledge Infrastructure), the patent literature of new energy vehicles can be traced back to as early as 2002. The number of new energy vehicle patents exceeded 100 for the first time in 2013, and the number of new energy vehicle patents has been growing rapidly since then.

Song and Zhu identified the technical topics in the field of artificial intelligence. The research shows that new energy vehicles are the frontier technical topic in the field of artificial intelligence, with a significant upward trend in the next three years [47]. Yang et al. extracted data from the EPO global patent statistics database and classified the technical topics of new energy vehicles by IPC classification number [48]. Xie et al. obtained 1393 pieces of patent data from 2003 to 2012 from the China Intellectual Property Rights Network and analyzed the development trend of new energy vehicles from five aspects: the number of patent applications, the distribution of patent owners, the technology life cycle, the IPC analysis, and the geographical distribution of patent R&D subjects [49]. Liang et al. analyzed the development trend of new energy vehicles by using technology activity analysis and patent technology cooccurrence to analyze the technology research frontier and development trend of new energy vehicles. The new materials, safe electricity, structural parts, and related electric traction are the popularities of R&D in the future period [50]. Wang et al. retrieved 19,610 pieces of patent data from the database of the State Intellectual Property Office and conducted statistics on the annual publication trend of patents, the portfolio structure of patent application subjects, and the distribution of IPC classification numbers, and the statistical results showed that the top three technical fields in patent number are battery, electrical machine, and electric control system [51]. Shen et al. constructed the IPC cooccurrence network of new energy vehicles by Apriori algorithm [40]. Li and Fan conducted a comparative analysis of key technology areas of international and Shanghai Automotive Group Corporation from three dimensions: patent distribution map of major technology areas, topography map of key technology R&D areas, and technology hotspot change map, and extracted the technology areas in which Shanghai Automotive Group Corporation has comparative advantages [52]. Shi et al. analyzed the overall development of Chinese new energy vehicle technology in the past 30 years from three aspects: number of patents, innovation subjects, and technology categories [53]. Sun et al. took the top 38 automotive manufacturers with the most NEV-related patents published in China as the target of their study and generated a patent cooperation map through a patent citation matrix to further analyze the cooperation behavior among different patent owners [54]. Feng et al. found that the topics of the four most promising ones are mainly about battery arrangement and protection, control systems, framework design, and charging connectors [55]. Miao et al. predicted the sales volume of new energy vehicles in China based on ARIMA model [56]. Wu et al. conducted an exploratory study on the ecosystem construction process of new energy vehicles, taking the NIO new energy vehicle as an example [57]. Fang and Zhang used the global patents of charging pile technology from 1990 to 2018 in the Smart Bud patent database to predict the new energy vehicle charging pile technology based on the life cycle and international patent classification numbers. The results of the study indicate that the charging pile technology in China will enter a mature period in 2020 and a recession period in 2025 [58]. Based on machine learning and empirical mode decomposition method, a technical topic was identified for Chinese new energy vehicles, and the results showed that China has a relatively high technological advantage in the field of battery cooling or maintaining low temperature technology and vehicle battery application technology [59].

From the above discussion, it can be found that the existing literature is mainly based on the analysis method of patent classification and patent citations to study the technological development of new energy vehicles, and although the analysis method of technical topic evolution based on patent classification and patent citations can discover the technological development trend from a macro perspective, it cannot show the specific evolution details of technical topics, and the technical topic identification based on patent texts can make up for this deficiency. Therefore, this paper downloads the patent texts of new energy vehicles from the patent database of CNKI and introduces the LDA topic model into the identification of technical topics in the field of new energy vehicles, which can solve the problems of “time lag” and “multiple meanings of words” of traditional research methods and accurately accurately discover the topics hidden in massive texts. On this basis, this paper predicts the popularity of China’s new energy vehicle technology topics from 2021 to 2025 by constructing ARIMA model.

3. Methodology

3.1. LDA Model-Based Technical Topic Identification

The LDA (Latent Dirichlet Allocation) topic model is a typical probability generation model [32] (see Figure 1), a text mining technology that can train large-scale document sets to identify potential topic information. It can realize unsupervised, without setting any topic vocabulary in advance and professional domain knowledge can complete topic crawling. M is the number of documents in the corpus, N is the number of words in the documents, K is the number of topics implied in all documents, denotes an M K document topic distribution matrix, denotes the topic distribution of the m-th document, denotes a K N topic word distribution matrix, denotes the distribution of topic words numbered k, is the prior distribution of topics for each document, is the prior distribution of words for each topic, represents the observable words, and is the potential topic distribution of each observed word.

In this paper, the technical topic matter implied in the patent text is set to obey the distribution as follows:

denotes the distribution of patent text d in technical topic k. For each technical topic k, the term distribution is generated. For each patent text d, the subject term distribution is generated, and, for the n-th term in each patent text, the subject term and the term are generated.

From the perspective of LDA, the generation process of a document can be decomposed into the following steps: First, the topic distribution of document d is generated by sampling from the Dirichlet distribution α, where . Second, the polynomial distribution of topics is sampled to generate topic for the n-th word in document d. Third, the word distribution of topic is generated by sampling from the Dirichlet distribution . Fourth, word is generated by sampling from the polynomial distribution of words .

The above generation process can be represented by the joint distribution of all visible and hidden variables.

The LDA topic model uses the Dirichlet distribution as the prior distribution for topic information mining. Compared with other generative probability models, LDA topic model has obvious advantages, which can describe document production mode more accurately, and the analysis results are better than other mixed topic models. The LDA topic model has the feature of discovering potential topics and has obvious advantages in topic research in emerging technology fields.

The main steps of technology topic mining using LDA model are as follows.

3.1.1. Text Segmentation

Patent text is very different from unstructured information such as news articles, novel biographies, and letters and emails. It records innovative contributions in a specific technical field and involves many specialized vocabulary and technical terms. Therefore, the Viterbi algorithm can be used to split the text of patent data, generate a directed acyclic graph based on the lexicon, and find the shortest path to accurately intercept the patent text according to the exact pattern of jieba splitting. In addition, although the patent abstract has a very standardized structure, including two main parts, introduction and subject introduction, there are still pronouns, articles, conjunctions, and punctuation in the text, which are not very useful in the text subject recognition and must be eliminated.

3.1.2. TF-IDF Weight

Since the new energy vehicle patent text contains specialized words from different disciplines and different words have different effects among themselves in expressing the text topic, this paper invokes the TF-IDF (Term Frequency-Inverse Document Frequency) method to assign weights to words. The main idea of TF-IDF is as follows: two indicators are needed for a word to maximize the representation of the text topic, one is whether the word appears many times in some texts and the other is that the word rarely appears in all texts [60]. Therefore, the value of TF-IDF is equal to the “Term Frequency” (TF) multiplied by the “Inverse Document Frequency” (IDF), which is calculated as follows:

denotes the number of occurrences of word in text , and denotes the total number of occurrences of all words in text .

denotes the total number of documents in the corpus, and denotes the number of documents containing word . When the TF and IDF values of a word are calculated, they are multiplied together to obtain the TF-IDF value of the word. The higher the TF-IDF value of a word is, the more important the word is in the text and the more likely it is a keyword of the text.

3.1.3. Model Parameter Setting

The main influence on the results of the LDA topic model is the preprocessing step, which requires more careful processing. For parameter setting, since the LDA topic model is an unsupervised machine learning model, for hyperparameter of the document topic prior Dirichlet distribution and hyperparameter of the topic word prior Dirichlet distribution , we generally use the scikit-learn library's default value 1/K, which is because, during the continuous iteration of Gibbs Sampling, these two parameters will keep approximating the true value of the text training set under the training of the LDA topic model. Therefore, the parameter setting of the LDA topic model is mainly focused on the discussion of the number of topics k and the maximum number of iterations.

The selection of the number of topics k is directly related to the training effect of the LDA topic model. There are three common methods for setting the number of topics k: ① Blei et al. used the size of confusion to evaluate the goodness of the model. A smaller value of perplexity means better training result of the training set. ② The Hierarchical Dirichlet Processes (HDP) method based on the Dirichlet process proposed by Teh et al. assumes that the document sets share the same topics before training and the number of topics is unlimited, but the exact number of topics will be determined during the derivation of the Dirichlet parameters [61]. ③ Griffiths et al. proposed the application of a Bayesian model approach to determine the optimal number of topics. Through literature review, it is found that the existing literature mainly uses method ① to determine the number of topics k by the method of perplexity [62], and both method ② and method ③ have the disadvantage of high computational complexity, and this paper uses method ① to calculate the number of topics k. Method ① of determining the number of iterations of the LDA topic model is mainly based on whether the iteration results converge to determine the optimal number of iterations.

3.1.4. Technical Topic Identification and Trend Analysis

Through the above steps, the LDA topic model can be trained, and the corresponding training results can be derived to identify each technical topic, respectively, and the development trend of each technical topic can be calculated by using the document topic distribution matrix.

3.2. Trend Forecasting of Technical Topics Based on ARIMA Model

ARIMA models can be divided into seasonal ARIMA models and nonseasonal ARIMA models. Because the technical topic of this paper is not affected by seasonality and produces obvious periodic changes, the nonseasonal ARIMA model is used in the modeling process. The specific process of model construction is shown in Figure 2 [63].(1)Sequence stabilization. Usually, the data predicted by ARIMA model is an unsteady temporal dataset, so the first process of construction is to preprocess by observing the changing trend of the original data. Generally, there are three ways to deal with it: difference, seasonal difference, and natural logarithm transformation. In this paper, differential preprocessing is mainly used. Generally, only one or two difference calculations are needed to get the average value of the data close to zero. In addition, in order to get a smooth series, it is also necessary to test the value of the autocorrelation function and the value of the partial correlation function of the data to see whether the value of the function is not significantly different from zero.(2)Model identification. The parameters of the ARIMA correlation model are determined as shown in Table 1. If the partial autocorrelation function (PACF) of the smooth series is truncated and the autocorrelation function ACF is trailing, it is determined that the series is suitable for constructing the AR() model, and the value of is the order of the truncated partial autocorrelation function; if the autocorrelation function (ACF) of the smooth series is truncated and the partial autocorrelation function (PACF) is trailing, it can be determined that the series is suitable for the MA(q) model, and the value of q is the order of the truncated autocorrelation function; if the autocorrelation function and the partial autocorrelation function of the smooth series are both trailing, it can be decided that the series is suitable for the construction of the ARMA(, q) model, and the values of and q need to be calculated by the AIC criterion, SC criterion, and BIC criterion for all possible values, from which the combination of the lowest values of AIC, SC, and BIC is selected as the optimal model parameters.(3)Model testing. The main test is whether the model is constructed reasonably enough, and this step requires an autocorrelation test on the residual series of the model with the optimal values determined in the previous step to determine whether it is a white noise series [64].(4)Model prediction. The fitting effect of the ARIMA model is further analyzed by calculating the relative error between the fitted and actual values; in the prediction stage, the first case after the evaluation is set to the specified date, and the future trend is forecasted.

The relative error is calculated by the following formula:

Thereinto, denotes the relative error between the fitted value and the true value, is the absolute error, which takes a value equal to the difference of the predicted value minus the true value, and L is the true value.

4. Empirical Research

4.1. Data Acquisition

In this paper, the patent database of CNKI was used to search the patent data related to new energy vehicles, the search time was February 1, 2020, the search subject was new energy vehicle, and a total of 18,583 patents related to new energy vehicles were obtained. The latest patent is “a remote monitoring recorder for new energy vehicles,” and the patent publication date is November 20, 2020. The latest patent is “a new energy vehicle remote monitoring recorder,” with a patent publication date of November 20, 2020, and the application time in the patent document collection spans from October 2000 to December 2020. The results of the patent search mainly focused on topics such as new energy vehicle battery, charging pile, new energy electric vehicle, charging device, power battery, battery box, and preparation method. The information obtained includes Patent Author, Applicant, Title, Country Name, Publication Number (PubNo), Publication Date (PubTime), Summary, and Claims. After removing the data of patents with incomplete information, the total number of remaining patent texts is 17, 816, and the specific number of publications is shown in Figures 3 and 4.

As can be seen from Figure 3, the development of Chinese new energy vehicle patents can be divided into three stages: 2002–2010, the number of new energy vehicle patents issued was small; 2011–2015, the number of new energy vehicle patents issued increased from 44 to 237, and the annual growth rate was 52.3%; 2016–2020, under the combined effect of policy support and market promotion, the number of new energy vehicle patents issued showed a rapid growth trend from 780 to 5888, and the annual growth rate was 65.8%.

4.2. New Energy Vehicle Technology Topic Identification and Evolution Analysis Based on LDA Topic Model

4.2.1. Text Preprocessing

The process of text preprocessing has three main parts: reading data, segmenting sentences, and saving segmentation results. Before the code is run, a proprietary data file named 1.csv, a deactivated words table named stopwords.txt, and a custom dictionary named userdict.txt have been prepared. The first step is to read the data from a file named 1.csv with the read csv( ) method of pandas. The second step is to load the prerecorded custom dictionary, segment the sentences using the jieba( ) method, save them to the sentence seged as an array data type, and delete the words contained in the deactivated word list by traversing each segmented word in the sentence seged. Execute the third step, save the result of the previous steps into a file named fenci.csv, and the effect of document word segmentation is shown in Figure 5.

4.2.2. Calculation of TF-IDF Assignment Weights and Optimal Number of Topics

Firstly, the text feature extraction function CountVectorizer( ) is applied for TF-IDF weight assignment. Secondly, in the LDA topic model parameter setting module, the number of iterations is set to 100 times, and the corresponding perplexity values from 2 to 50 topic counts are calculated sequentially. As shown in Figure 6, the curve changes sharply and then slowly. When the document set is divided into 28 topics, the minimum confusion value is 319.18.

4.2.3. Unsupervised Training of LDA Topic Models

The LDA topic model was trained, and the number of iterations was continuously increased until the topics of each word converged to a stable state, and the top 10 technical terms in terms of probability under each topic were derived, and the three topics with little association with the main technology of new energy vehicles were excluded, and the final 25 technical topic contents were determined as shown in Table 2.

Through the division of new energy vehicle technology topics, it can be divided into 25 technology areas, and the topics are named as shown in Table 3.

4.2.4. Technical Topic Evolution Analysis

A 17817 25 document topic distribution matrix can be obtained after training by the LDA topic model. The sum of the values of each row of this matrix is 1, and the values in each column are different. The greater the sum of the values in each column, the greater the importance of the topic to the corpus. The top 10 technical topics in the corpus of new energy vehicles in terms of importance are Topic 8: Installation and Fixation, Topic 5: Heat Dissipation and Cooling, Topic 14: Vehicle Data Monitoring, Topic 9: Charging Pile, Topic 15: Vibration Damping, Topic 6: Electric Drive System, Topic 3: Cable Insulation Materials, Topic 13: Production and Processing, Topic 23: Circuit Control, and Topic 21: Wireless Charging.

The time dimension data of patents is introduced to portray the changes of each technology topic with time by taking the year as the unit; see Figure 7. Since 2014, China’s new energy vehicles have achieved fruitful results in patent technology and the continuous improvement of the overall technology level. The achievements of Topic 5: Heat Dissipation and Cooling, Topic 8: Mounting Fixation, Topic 6: Electric Drive System, Topic 9: Charging Pile, and Topic 15: Vibration Damping stand out. These technologies are closely related to new energy vehicles and directly determine the performance level of new energy vehicles. This is because the first element that car owners pay attention to when purchasing a new energy vehicle is the driving efficiency of the vehicle, such as maximum mileage, 100 km consumption, power, and other factors. The development of this type of technology can be a great attraction to consumers and can greatly accelerate the speed of the market promotion of new energy vehicles. Topic 7: Chassis, Topic 10: Water Pump, Topic 18: Electric Motor, Topic 24: Module, and other technologies are developing relatively slowly, and new innovations and breakthroughs are urgently needed. Topic 12: Vacuum Pump and Topic 21: Wireless Charging are relatively less important in 2019–2020.

Figure 8 is plotted based on Figure 7, and the value of each box in the figure is equal to the importance of the corresponding topic in that year divided by the overall importance of the technical topic. Through observation, it can be found that the importance of Topic 3: Cable Insulation, Topic 14: Vehicle Data Monitoring, and Topic 23: Circuit Control declined from 2014 to 2020, while Topic 5: Heat Dissipation and Cooling and Topic 8: Mounting and Fixing continued to increase in importance. With the continuous improvement of the drive system of new energy vehicles, this part must require the electric motor to provide higher power with more stable output. Although the electric motor is a low heating component, the heat generated during normal operation is low, but the motor will lead to high coolant temperature after long time and high-power operation. This reflects that, in recent years, with the maturity of new energy vehicle technology, the technical level has been able to meet the needs of consumers, and the most important concern of automobile manufacturers is to develop an efficient process to realize the mass production of new energy vehicles, so the importance of installation and fixing technology is increasing.

4.3. New Energy Vehicle Technology Topic Popularity Screening Based on LDA Model

In the training results of the LDA topic model, the distribution of topics corresponding to each document can be obtained, and the changing trend of technical topic popularity over time is calculated according to the date of publication corresponding to each document. Using each month’s record as a cycle, the popularity of each period is connected with a dash line, and the overall trend of changes in 25 technical topics is depicted by a linear regression model, and some of the topics are shown in Figure 9. The overall slope of each technical topic is shown in Table 4, and the top 5 technical topics in terms of the slope of the linear trend were filtered out based on the slope of the linear trend: Topic 20: Safety, Topic 8: Installation Fixing, Topic 3: Cable Insulation, Topic 15: Vibration Reduction, and Topic 10: Pump Cooling.

4.4. Trend Prediction of Technology Topics Popularity Based on ARIMA Model

4.4.1. Time Series Smoothing Processing

As shown in Figure 10, the popularity of Topic 20 reached its peak in September 2018 and fell to the local lowest point in September 2019, with the overall popularity change trend being constantly rising. Since the ARIMA model requires the time series to be a smooth series, it still needs to perform the difference operation on the original data. The difference operation formula is as follows:

The timing diagram of sequence {} after the 1st-order differencing process is shown in Figure 11.

The long-term trend of the original series is obtained by 1st-order differencing, and it can be seen from the graph that the values of the series after 1st-order differencing fluctuate mainly around the upper and lower sides of the 0 scale. In addition, the stability is judged by the ADF (Augmented Dickey-Fuller) test, as shown in Table 5. The ADF test value of the original series is −0.125792, whose absolute value is less than the critical absolute value of −1.945389 at 5% significance level, and its corresponding value is 0.6370, which is greater than 0.05. It shows that the original hypothesis is accepted at 5% significance level and the initial time series has unit root; that is, the initial time series is a nonstationary series. The ADF test value of the time series with 1st-order difference is −10.66079, whose absolute value is greater than the absolute value of −1.945456 at the 5% significance level, and its corresponding , indicating that the original hypothesis is rejected at the 5% significance level and the 1st-order difference series does not have a unit root, so the 1st-order difference series is a smooth series.

4.4.2. Model Identification

From Figure 12, it can be found that the autocorrelation diagram after the 1st-order difference shows that the autocorrelation value fluctuates within the basic boundary value range; only the 1st-order and 5th-order autocorrelation values slightly exceed the boundary value, while the autocorrelation value fluctuates within the significant boundary in all other orders. In the partial autocorrelation graph, the 1st-order partial autocorrelation value is on the significant boundary value, while the 5th-order partial autocorrelation value is slightly beyond the significant boundary, in addition to the other basic within the boundary value; therefore, the autocorrelation and partial autocorrelation can be initially considered as 1st-order or 5^th-order trailing. In addition, since SPSS does not have the function of automatic identification of model parameters, it is necessary to calculate all possible parameter combinations, and the BIC indicator is selected as the evaluation indicator among many method indicators, and the optimal parameter combination is determined by the BIC indicator; that is, the parameter combination corresponding to the smallest BIC value is selected as the optimal ARIMA model parameters [64], and the calculation results are shown in Table 6, and the optimal model is ARIMA (0, 1, 1).

(a)

(b)

4.4.3. Model Testing

After determining the model parameters, the model was then fitted and tested by white noise test. R² = 0.893, the fitting effect was good, and the value of the Ljung-Box test was greater than 0.05, so the original hypothesis could not be rejected, and then the series could be considered as white noise. As can be seen from Table 7, the coefficient of MA is 0.282 and the significance level meets the test requirement (), so the coefficient is significantly not 0. Also, the autocorrelation and partial autocorrelation of the observed residuals are smooth states (see Figure 13), so it is reasonable to set the ARIMA (0, 1, 1) model.

(a)

(b)

4.4.4. Trend Prediction

Finally, the model is applied to predict the development trend of China’s new energy vehicles from 2021 to 2025 by plotting the time series prediction results of Topic 20: Safety, Topic 8: Installation Fixation, Topic 3: Cable Insulation, Topic 15: Vibration Damping, and Topic 10: Water Pump Cooling (see Figure 14), in which September 2020 is used as the dividing line, the left area shows the effect of model fitting, and the right area shows the trend prediction results. At the same time, by calculating the relative error between the fitted and true values to further judge the fitting effect of the model, the data show that the maximum relative error is less than 17%, which is within the acceptable degree, indicating that the fitting effect of the model is very good and therefore the model is more accurate in predicting the data results.(1)Topic 20: Development trend of security According to ARIMA model prediction, from January 2021 to December 2025, Topic 20 (Safety) has a good development prospect, and the popularity will increase from 63.58 to 113.07, with an increase of 49.49, which is the highest increase among all technology topics. There are two types of automotive safety, active safety protection and passive safety protection. Active safety mainly detects the condition of the car, environment, and people and takes active measures to avoid accidents in advance. At the technical level, it is mainly optimized by active suspension system (ASS), antilock brake system (ABS), electronic brake-force distribution (EBD), electronic stability program (ESP), traction control system (TCS), lane departure warning system (LDWS), panoramic surround view (PSW), and so forth. Passive safety is mainly in the car traffic accident through passive equipment to minimize personal injury to passengers or pedestrians in the car; in the technical design it is mainly through airbags, seat belts, head and neck protection devices, crash beams, and a variety of collision energy-absorbing collapse structures, as well as other aspects of improvement. The collision resistance of new energy vehicles is not as good as that of traditional fuel vehicles, and a collision or charging overcurrent may lead to major accidents of battery spontaneous combustion. The transition from fuel vehicles to new energy vehicles has put forward new higher requirements for safety technology. Along with the deep integration of the automotive industry and the Internet information industry, the safety technology of new energy vehicles will be extended to the fields of Internet information security protection. Therefore, new energy vehicle safety technology is one of the main factors affecting its development and solving this type of pain point can greatly promote the popularization of new energy vehicles.(2)Topic 8: Trends in mounting and fixing Topic 8 (Installation and Fixation) is mainly to improve the efficiency of new energy vehicle parts processing and production, enhance the quality of product production, facilitate future use and maintenance, and transform the processing equipment and parts structure of new energy vehicles. For example, the patent text “a new energy vehicle motor assembly device” proposes improving the efficiency of motor lifting and installation, changing the way of human eye positioning and installation in the past, and setting up a positioning device for the hanging machine, so as to effectively improve the efficiency of production and reduce the work intensity of workers. In the patent text “a new energy vehicle processing parts welding device” for the existing new energy vehicle parts processing welding device has welding angle adjustment difficulties, welding stability is not good, welding combination tools cannot coordinate with the operation and other problems, a new energy vehicle parts processing welding device is designed, and successfully deal with the existing problems, convenient for workers to operate while improving the accuracy of welding. In order to solve the limitation that the punching device cannot adapt to the existing demand in the production process of new energy vehicle chassis, which has low punching accuracy and complicated operation, Gu Guangzhao designed a new energy vehicle chassis positioning punching device equipped with laser displacement sensor, which greatly improves the product production quality. By reforming this kind of fixed-type technology and optimizing the process of product processing, automotive manufacturers can produce products that satisfy consumers more efficiently. According to the ARIMA model, it is predicted that, from January 2021 to December 2025, the popularity of Topic 8 (Installation and Fixation) will continue to rise, from 37.61 to 66.97.(3)Topic 3: Trends in cable insulation materials Topic 3 (Cable Insulation Materials) has a pivotal role in the whole technology field of new energy vehicles. It can be seen from Figure 12 that the research heat of Topic 3 decreased slightly in March 2020, but, from the second half of 2020, the research heat of cable insulation materials slowly rose. From January 2021 to December 2025, the research trend for cable insulation materials shows a gentle increase from 31.65 to 56.23. As a highly integrated platform for electrification, cables are as indispensable as the vascular system of the human body. From a broad perspective, cable insulation materials are widely used, which can be used not only as components of the power transmission system in the vehicle but also in charging gun, charging pile, or on-board charging. Since new energy vehicles mainly use unconventional automotive fuels as their power source, the system components are required to transmit currents as high as between 250 A and 450 A when performing their work, which is relatively rare compared to conventional power-driven vehicles with such high current transmission. It is worth mentioning that this cable insulation material with high heat resistance is of great significance to improve the performance of the motor and can obviously reduce the mass-power ratio, namely, kg/kw, in the technical index of the motor. The mass-to-power ratio of electric motors has been reduced from 533 kg/kw in 1900 to 6 kg/kw in 2020. The important reason for this change is the use of insulating materials with high heat resistance, which reduces the consumption of large amounts of metal materials and lowers the cost of motor production. In new energy vehicles, insulation materials account for about half of the cost consumption ratio of electric motors and electrical products, which illustrates the status and role of cable insulation materials research in the new energy vehicle industry. In addition, the development of highly flexible cable insulation materials facilitates the accommodation of more electronic components in a limited space. In the application of charging guns and charging posts, cable insulation materials need to overcome a variety of uncertain and unfavorable factors from sunlight, weathering, humidity, seawater, acid rain, freezing, and so forth and also meet the needs of bending and dragging in use, which puts new demands on cable insulation materials such as high and low temperature resistance, oil resistance, water resistance, flame retardancy, tear resistance, electrical insulation, and UV (ultraviolet) aging resistance performance put forward more and higher requirements.(4)Topic 15: Development trend of vibration damping Based on the ARIMA model, it is predicted that, from January 2021 to December 2025, the research enthusiasm of Topic 15 (Vibration Damping) shows a wave-like upward trend from 36.18 to 64.88. Topic 15 (Vibration Damping) seems to be less important and less technical in the automotive industry technology research. However, the reality is that damping technology can avoid damage to car parts by effectively suppressing the reciprocal rebound force generated after spring absorption, by reducing the vibration of the frame and body, and at the same time can effectively improve the smoothness and comfort during the driving process. Expanding to the field of new energy vehicles, the damping technology can have other new applications in addition to the above-mentioned roles. Through a careful reading of the patent text, it is found that the damping technology is generally applied to provide damping devices for batteries. This is because there are various precision electronic parts inside the battery module assembly, which have high environmental requirements in the process of use, especially in shock absorption. The vibration generated by the car driving on bumpy roads can easily affect the battery, and the internal components of the battery are likely to loosen, thus leading to failure and reducing the service life of the battery. The improvement of vibration damping technology can protect the parts of new energy vehicles to the greatest extent, reduce the failure rate of vehicles, ensure the smoothness of driving, and thus improve people’s riding experience.(5)Topic 10: Development trend of water pump cooling Although Topic 10 (Water Pump Cooling) is not as hot as Topic 20 (Safety), its research popularity follows the development of the overall technology field, as it plays an indispensable role in the field of new energy vehicle technology. Based on ARIMA model prediction, the popularity of topic 10 (Water Pump Cooling) grows from 31.43 to 55.68 from January 2021 to December 2025. Topic 10 (Water Pump Cooling) is mainly used in the process of driving new energy vehicles, the internal components of the motor continuously generate losses during the movement, and these losses are eventually dissipated to the outside through heat. These losses are eventually dissipated to the outside by way of heat, and it is necessary to take away unnecessary heat by effective cooling, so as to ensure the normal and stable operation of the motor in a stable hot and cold circulation system. According to the principle of cooling, it can be divided into three types: natural cooling, air cooling, and water cooling. Natural cooling mainly relies on the heat transfer from the motor body into itself to dissipate the heat generated by the motor. This type of cooling method is inefficient and only applicable to low-power generators. Air cooling is mainly through the coaxial fan to form the cooling air circulation, using the air around the motor as the medium, directly discharging the heat of the motor to the surrounding environment. Water cooling uses coolant as the medium and promotes the continuous circulation of coolant through pipes to take away the heat generated by the motor rotor and stator during operation. Among these three cooling methods, water cooling has the greatest impact on the efficiency improvement of new energy vehicles. At the same time, not only is the application of such technology limited to the cooling of the electric motor, but also it includes the cooling of the battery. During the high load and long-term driving of new energy vehicles, the continuous high temperature caused by the continuous discharge of the battery will inevitably lead to the high temperature of the coolant, which makes the efficiency of the water cooling system drop sharply and leads to the abnormal power supply of the battery at high temperature and also has a negative impact on the service life of the battery. Therefore, the transformation of water cooling technology can provide guarantee for the normal operation of new energy vehicles, which has an important impact on important components such as batteries and motors.

Looking at the above five topics with the fastest growing in research, it can be found that the overall development trend of each technology topic is relatively consistent, the development is relatively slow before the beginning of 2017, and the number of patent issuance is also relatively small. But, with the guidance of policies, the government increased the encouragement of automobile manufacturing enterprises to improve the quality of new energy vehicles through technological innovation, while providing a large number of preferential subsidy policies to consumers, which to a certain extent promoted the rapid development of new energy vehicles; as can be seen from the 5 technology topics after 2018, the technology of new energy vehicles has been rapidly accumulated. The results show that these five technology topics will continue to grow in popularity from 2021 to 2025.

5. Conclusions and Policy Implications

This paper identifies technical topics in the field of new energy vehicles in China and constructs ARIMA model to predict the trend of popularity of technology topics. The main conclusions are as follows.

Firstly, through the topic mining of patent texts in the field of new energy vehicles in China from 2000 to 2020 in the patent database of CNKI, it is found that there are 25 technology topics in the field of new energy vehicles. In terms of importance, the five most important technical topics in the field of new energy vehicles in China are as follows: Topic 8: Installation and Fixation, Topic 5: Heat Dissipation and Cooling, Topic 14: Vehicle Data Monitoring, Topic 9: Charging Piles, and Topic 15: Vibration Damping. Introducing the time dimension, the importance of Topic 5: Thermal Cooling, Topic 8: Installation and Fixation, Topic 6: Electric Drive Systems, Topic 9: Charging Piles, and Topic 15: Vibration Damping grows steadily from 2014 to 2020. In 2014–2020, Topic 7: Chassis, Topic 10: Water Pump, Topic 18: Electric Motor, and Topic 24: Module are relatively slow in development. In 2014–2018, Topic 12: Vacuum Pump and Topic 21: Wireless Charging steadily increase in importance, but, in 2019–2020, the importance decreases.

Second, the LDA topic model is trained unsupervised, and the evolution trend of each technical topic can be obtained by dividing the importance of each technical topic over the years by the overall importance of the technical topic. It was found that the importance of Topic 3: Cable Insulation, Topic 14: Vehicle Data Monitoring, and Topic 23: Circuit Control declined from 2014 to 2020, while the importance of Topic 5: Thermal Cooling and Topic 8: Installation and Fixation continued to increase.

Third, a linear regression model was used to portray the overall trend of changes in the 25 technology topics, and the top 5 technology topics in terms of hotness were screened according to the slope of the linear trend: Topic 20: Safety, Topic 8: Installation and Fixation, Topic 3: Cable Insulation, Topic 15: Vibration Damping, and Topic 10: Water Pump Cooling. Based on ARIMA prediction, it is found that 2021–2025 will be a booming period for new energy vehicles in China, and the popularity of Topic 20: Safety, Topic 8: Installation and Fixation, Topic 3: Cable Insulation Materials, Topic 15: Vibration Reduction, and Topic 10: Water Pump Cooling continues to grow over time. The popularity of Topic 20: Safety, Topic 3: Cable Insulation, and Topic 10: Water Pump Cooling shows a wave upward growth trend. Topic 20: Heat will grow from 63.58 to 113.07 with a value increase of 49.49, which is the fastest growth in heat among all technical topics.

Our findings not only contribute to advancing the study of China’s new energy vehicles innovation management and industrial development but also provide inputs for policymakers. Three policy implications are proposed based on the above analysis.

First, strengthen technology research and development in the field of safety to protect the development of the new energy vehicle industry. The study found that, among the 25 technology topics, Topic 20 (Safety) was the most popular. The safety of new energy vehicles is gradually attracting the attention of the industry, which is related to the safety of consumers and even the whole society, as well as the confidence of social groups in the development of new energy vehicle industry, which in turn affects the future development direction and space of new energy vehicle industry. Therefore, it is recommended to do the following: ① The subsidy policy of the new energy vehicle industry should be tilted to the field of technology research and development, focusing on subsidizing the research and development of core components such as batteries, motors, electric controls, and chips to enhance the safety of new energy vehicles. ② The subsidy policy of new energy vehicle industry should be directed to the R&D of battery and its related supporting technologies, support the R&D of graphene application in the field of battery, and improve the safety of new energy vehicle battery. ③ Strengthen the technology research and development of active safety, mainly involving active suspension system (ASS), antilock brake system (ABS), electronic brake-force distribution (EBD), electronic stability program (ESP), traction control system (TCS), lane departure warning system (LDWS), panoramic surround view (PSW), and so forth. ④ Strengthen the technology development of passive safety, mainly involving airbags, seat belts, head and neck protection devices, and so forth. ⑤ Improve the monitoring system of new energy vehicle enterprises and do a good job of after-sales quality tracking and inspection of products. ⑥ Strengthen the supervision of safety from core components to the whole vehicle, and implement the battery catalog system and battery coding system.

Second, accelerate the construction of new energy vehicle infrastructure. With the continuous development of the new energy vehicle industry, the charging problem has gradually become an important factor limiting the development of the new energy vehicle industry, and the lack of convenience of charging facilities has become a common concern for consumers, increasing the uncertainty of market promotion. As of 2020, the number of public charging piles in China was 807,000. Among them, 145,000 were in residential areas, accounting for 18%; the number of private charging piles was 870,000; the number of public charging stations was 63,800; and the number of exchange stations was 555. Despite the promising achievements of China's charging infrastructure in the “13th Five-Year Plan” period, there are still many shortcomings; mainly in the scale of charging facilities there is still a gap, the spatial layout needs to be optimized, charging efficiency still needs to be improved, the utilization rate of public charging facilities is low, and the social environment and synergy mechanism supporting charging facilities have not yet been formed. In order to accelerate the construction of new energy vehicle infrastructure, it is recommended to do the following: ① The financial subsidy policy for new energy vehicles should be gradually tilted toward the construction of supporting infrastructure. ② According to the market demand, appropriately increase the charging piles and other infrastructure, and encourage local governments to build charging spaces in public places such as paid parking spaces, temporary parking strips on the roadside, and parking lots in residential communities according to local conditions, so as to improve the utilization rate of charging spaces. ③ Ensure that residential communities, public institutions, and so forth are in accordance with a certain proportion of charging facilities to properly solve the difficult problem of charging piles into the community.

Third, to strengthen the technology development of Topic 8 (Installation and Fixation), Topic 5 (Heat Dissipation and Cooling), Topic 14 (Vehicle Data Monitoring), Topic 9 (Charging Pile), Topic 15 (Vibration Damping), and Topic 3 (Cable Insulation) to improve the overall performance of new energy vehicles, the study found that the above six topics have high scores in terms of importance and popularity, representing the technical dynamics and development trend of the new energy vehicle technology field. Therefore, new energy vehicle enterprises must strengthen the investment in the above technical fields and strive to gain the initiative to develop in the fierce competition. During the “13th Five-Year Plan” period, China's production and sales of new energy vehicles have grown rapidly, ranking first in the world for five consecutive years since 2015, with a total of more than 4.8 million vehicles promoted, accounting for more than 50% of the world, and new energy vehicles are entering thousands of households. However, we should also see that, compared with traditional fuel cars, the current market share of new energy vehicles is still very small, and to replace traditional fuel cars is a long way off. Historically, new energy vehicles have failed repeatedly in the competition with traditional fuel cars. Therefore, it is crucial to seize the technical dynamics and development trend in the field of new energy vehicle technology, strengthen the research and development of technical topics such as installation and fixation, heat dissipation and cooling, vehicle data monitoring, charging pile, vibration damping, and cable insulation materials, and improve the overall performance of new energy vehicles for the development of new energy vehicle industry.

Data Availability

All data in this paper are available upon request.

Conflicts of Interest

The authors declare that they have no conflicts of interest.

Acknowledgments

The support by Humanities and Social Science Research Projects of Ministry of Education (no. 18YJC630050), Guangdong Province National Natural Science Foundation of China (no. 2018A030313227), Soft science Project of Guangdong Province (no. 2019A101002057), and the National Natural Science Foundation of China (nos. 71874036, U1901222, 72074056, 71673064, and 71974039) is acknowledged.

References

L. Zhang, “Research on business model innovation and path evolution of china’s new energy vehicle from the perspective of sociotechnical system,” Beijing Jiaotong University, vol. 11, no. 6, p. 1711, 2020.
View at: Google Scholar
Y. Chen, L. Ni, and K. Liu, “Does China’s new energy vehicle industry innovate efficiently? a three-stage dynamic network slacks-based measure approach,” Technological Forecasting and Social Change, vol. 173, no. 12, Article ID 121161, 2021.
View at: Google Scholar
Q. Zhou, X. Gu, L. Zeng, and S. Wu, “The collaboration mechanism of patent implementation in new energy automobile industry,” Studies in Science of Science, vol. 38, no. 7, pp. 1236–1243+1259, 2020.
View at: Google Scholar
G. Zhang, W. Xie, R. Hu, W. Ma, and Y. Liu, Disruptive Innovation: SNM Perspective, Science Press, Beijing, China, 2016.
Z. Wu, Q. Shao, Y. Su, and D. Zhang, “A socio-technical transition path for new energy vehicles in China: a multi-level perspective,” Technological Forecasting and Social Change, vol. 172, no. 11, Article ID 121007, 2021.
View at: Google Scholar
B. Sun, X. Xi, and H. Yu, “A study on technology niche’s from the perspective of network relationship: empirical analysis of new energy vehicle industry in northeast three province perspective,” Studies in Science of Science, vol. 31, no. 4, pp. 518–528, 2013.
View at: Google Scholar
T. Yang, X. Chen, and X. Li, “Evaluation and analysis of new-energy vehicle industry policies in the context of technical innovation in China,” Journal of Cleaner Production, vol. 281, no. 1, Article ID 125126, 2021.
View at: Google Scholar
R. Wang, X. Zhao, W. Wang, and L. Jiang, “What factors affect the public acceptance of new energy vehicles in underdeveloped regions? A case study of Gansu Province, China,” Journal of Cleaner Production, vol. 318, no. 10, Article ID 128432, 2021.
View at: Google Scholar
L. Liu, T. Zhang, A. P. Avrin, and X. Wang, “Is China’s industrial policy effective? An empirical study of the new energy vehicles industry,” Technology in Society, vol. 63, no. 11, Article ID 101356, 2020.
View at: Google Scholar
Z. Zhang and F. Zhao, “The comparative study on the development strategies of the new energy automobile industry between China and America: based on the differences of goal orientation,” Studies in Science of Science, vol. 32, no. 4, pp. 531–535, 2014.
View at: Google Scholar
Y. Huang, Y. Hou, T. Liu, and Z. Gao, “Research on BEV industry competitive profile under patent information field of vision,” Science & Technology Progress and Policy, vol. 34, no. 4, pp. 72–77, 2017.
View at: Google Scholar
Z. Wang, “Characteristics and utilization of patent documentation,” Journal of Modern Information, vol. 2008, no. 9, pp. 151–152+156, 2008.
View at: Google Scholar
H Xiao, Patent Map Methods and Applications, Shanghai Jiao Tong University Press, Shanghai, China, 2011.
S. Jun, “Central technology forecasting using social network analysis,” Computer Applications for Software Engineering, Disaster Recovery, and Business Continuity,” Communications in Computer and Information Science, Springer, Berlin, Heidelberg, vol. 340, 2012.
View at: Publisher Site | Google Scholar
R. N. Kostoff, R. Boylan, and G. R. Simons, “Disruptive technology roadmaps,” Technological Forecasting and Social Change, vol. 71, no. 1-2, pp. 141–159, 2004.
View at: Publisher Site | Google Scholar
Y. Zhang, W. Zhang, and J. Bi, “Recent advances in driverless car,” Recent Patents on Mechanical Engineering, vol. 10, no. 1, pp. 30–38, 2008.
View at: Google Scholar
K. Suzuki, J. Sakata, and J. Hosoya, “An empirical analysis on progress of technology fusion,” in Proceedings of the 2008 IEEE Third International Conference on Digital Information Management, no. 11, pp. 937–939, London, United Kingdom, July 2008.
View at: Publisher Site | Google Scholar
S. Jeong, J.-C. Kim, and J. Y. Choi, “Technology convergence: what developmental stage are we in?” Scientometrics, vol. 104, no. 3, pp. 841–871, 2015.
View at: Publisher Site | Google Scholar
W. S. Lee, E. J. Han, and S. Y. Sohn, “Predicting the pattern of technology convergence using big-data technology on large-scale triadic patents,” Technological Forecasting and Social Change, vol. 100, no. 11, pp. 317–329, 2015.
View at: Publisher Site | Google Scholar
B. Huang and L. Huang, “Patent-based association analysis for technological convergence of IT and BT,” in Proceedings of the 2015 International Conference on Logistics, Informatics and Service Sciences, pp. 1–6, IEEE, 2015.
View at: Google Scholar
J. Park and D. Kang, “Trend analysis of unmanned technology using patent information,” Journal of The Korea Society of Computer and Information, vol. 22, no. 3, pp. 89–96, 2017.
View at: Google Scholar
F. Narin, “Patent bibliometrics,” Scientometrics, vol. 30, no. 1, pp. 147–155, 1994.
View at: Publisher Site | Google Scholar
O. Kwon, J. Seo, K. Noh, J. Kim, J. S. Kim, and S. Y. Shin, “Categorizing influential patents using bibliometric analysis of patent citations network,” Information-an International Interdisciplinary Journal, vol. 10, no. 3, pp. 313–326, 2007.
View at: Google Scholar
C. Choi and Y. Park, “Monitoring the organic structure of technology based on the patent development paths,” Technological Forecasting and Social Change, vol. 76, no. 6, pp. 754–768, 2009.
View at: Publisher Site | Google Scholar
C. C. Hsueh and C. C. Wang, “The use of social network analysis in knowledge diffusion research from patent data,” in Proceedings of the 2009 International Conference on Advances in Social Network Analysis and Mining, pp. 393–398, IEEE, Athens, Greece, July 2009.
View at: Publisher Site | Google Scholar
Y. Geum, C. Kim, S. Lee, and M. S. Kim, “Technological convergence of IT and BT: evidence from patent analysis,” ETRI Journal, vol. 34, no. 3, pp. 439–449, 2012.
View at: Publisher Site | Google Scholar
D. Zhai, L. Cai, J. Zhang, and X. Zhang, “The study of patent-based model for identifying technology fusion innovation trajectory illustrated by the case of cloud computing technology,” Journal of the China Society for Scientific and Technical Information, vol. 34, no. 4, pp. 352–360, 2016.
View at: Google Scholar
C. Kim and H. Seol, “On a patent analysis method for identifying core technologies intelligent decision technologies,” in Proceedings of the 4th International Conference on Intelligent Decision Technologies, pp. 441–448, Springer, Gifu, Japan, May 2012.
View at: Publisher Site | Google Scholar
H. Small, K. W. Boyack, and R. Klavans, “Identifying emerging topics in science and technology,” Research Policy, vol. 43, no. 8, pp. 1450–1467, 2014.
View at: Publisher Site | Google Scholar
B. Li and X. Chen, “Identification of emerging technologies in nanotechnology based on citing coupling clustering of patents,” Journal of Intelligence, vol. 34, no. 5, pp. 35–40, 2015.
View at: Google Scholar
J. Lee, J. Kim, J. Lee, S. Park, and D. Jang, “A study on analysis of R&D intensity based on patent citation information: case study on self-driving car of Google,” Journal of Korean Institute of Intelligent Systems, vol. 26, no. 4, pp. 327–333, 2016.
View at: Publisher Site | Google Scholar
D. M. Blei, A. Y. Ng, and M. I. Jordan, “Latent dirichlet allocation,” Journal of Machine Learning Research, vol. 2003, no. 3, pp. 993–1022, 2003.
View at: Google Scholar
X. Qin and X. Le, “Topic evolution research on a certain field based on LDA topic association filter,” Data Analysis and Knowledge Discovery, vol. 31, no. 3, pp. 18–25, 2015.
View at: Google Scholar
E. Yan, “Research dynamics, impact, and dissemination: a topic-level analysis,” Journal of the Association for Information Science and Technology, vol. 66, no. 11, pp. 2357–2372, 2015.
View at: Publisher Site | Google Scholar
H. Yang and J. Yang, “Quantitative analysis of policy text merged with LDA model: based on the field of international climate as demonstration,” Journal of Modern Information, vol. 36, no. 5, pp. 71–81, 2016.
View at: Google Scholar
L. Liao and F. Le, “Research on patent technology evolution based on LDA model and classification number,” Journal of Modern Information, vol. 37, no. 5, pp. 13–18, 2017.
View at: Google Scholar
W. Wang and J. Ma, “Topic detection and evolution analysis of research project based on LDA ———a case study of projects on ocean acidification supported by NSF,” Journal of Intelligence, vol. 36, no. 7, pp. 34–39, 2017.
View at: Publisher Site | Google Scholar
S. Fan, X. An, G. Yan, and Y. Li, “Study on the recognition method of Frontier topic in the medical field,” Journal of the China Society for Scientific and Technical Information, vol. 37, no. 7, pp. 686–694, 2018.
View at: Google Scholar
H. Yi, H. Wu, and J. Chang, “Technology topics evolution of graphene based on topic life cycle and technology entropy,” Journal of Intelligence, vol. 38, no. 7, pp. 64–70, 2019.
View at: Google Scholar
J. Shen, Q. Yan, Q. Sun, and Z. Wan, “Research on technology fusion recognition and technology opportunity prediction based on patent data mining,” Library Journal, vol. 38, no. 10, pp. 95–106, 2019.
View at: Google Scholar
H. Ma, L. Kong, C. Lin, and X. Yang, “Key generic technology identification based on patent mining,” Journal of the China Society for Scientific and Technical Information, vol. 39, no. 10, pp. 1093–1103, 2020.
View at: Google Scholar
M. Feng, F. Wei, J. Li, and L. Zheng, “Research on topic identification and technology evolution from the perspective of industrial Chain: a case study of 3D-printing,” Journal of Intelligence, vol. 39, no. 8, pp. 46–52, 2020.
View at: Google Scholar
H. Li and H. Xuan, “Multi-attribute mining method for technology innovation subject from the perspective of paten: the case of chip patents,” Library and Information Service, vol. 64, no. 11, pp. 496–107, 2020.
View at: Google Scholar
H. Tang and Y. Qiu, “Emerging technology topic identification based on multi-source information: intelligent connected vehicle as an example,” Journal of Intelligence, vol. 40, no. 3, pp. 81–88, 2020.
View at: Google Scholar
Y. Lianga, Y. Jua, L. Martínezb, and T. Yan, “Sustainable battery supplier evaluation of new energy vehicles using a distributed linguistic outranking method considering bounded rational behavior,” Journal of Energy Storage, vol. 48, no. 4, Article ID 103901, 2022.
View at: Publisher Site | Google Scholar
B. Lin and L. Shi, “Do environmental quality and policy changes affect the evolution of consumers’ intentions to buy new energy vehicles,” Applied Energy, vol. 310, no. 3, Article ID 118582, 2022.
View at: Publisher Site | Google Scholar
K. Song and Y. zhu, “Patent Frontier technology topic identification and trend prediction: a case analysis of artificial intelligence,” Journal of Intelligence, vol. 40, no. 1, pp. 33–38, 2021.
View at: Google Scholar
L. F. Yang, J. H. Xu, and P. Neuhäusler, “Electric vehicle technology in China: an exploratory patent analysis,” World Patent Information, vol. 35, no. 4, pp. 305–312, 2013.
View at: Publisher Site | Google Scholar
Z. Xie, Y. Zhang, C. Zheng, and M. Zhang, “Research on patent trend of new energy vehicles industry,” Enterprise Management, vol. 2015, no. 4, pp. 127–141, 2015.
View at: Google Scholar
S. Liang, X. Ji, and Y. Li, “Application of patent scientometrics methods in technology foresight: take the new energy automobile as an example,” Journal of Intelligence, vol. 34, no. 2, pp. 73–78, 2015.
View at: Google Scholar
J. Wang, Y. Liu, and K. Ari, “Research on technology innovation of Chinas NEV industry based on patent analysis,” Journal of Intelligence, vol. 35, no. 1, pp. 32–38, 2016.
View at: Google Scholar
G. Li and X. Fan, “The choice of the key new energy vehicle technology of SAIC basing on the patent analysis,” Journal of Modern Information, vol. 37, no. 10, pp. 122–127+136, 2017.
View at: Google Scholar
X. Shi, R. Jing, G. Zheng, and G. Hou, “Analysis of regional distribution characteristics of new energy vehicle technology innovation in China based on patent data,” Journal of Industrial Technological Economics, vol. 37, no. 08, pp. 60–67, 2018.
View at: Google Scholar
H. Sun, Y. Geng, L. Hu, L. Shi, and T. Xu, “Measuring China's new energy vehicle patents: a social network analysis approach,” Energy, vol. 153, no. 6, pp. 685–693, 2018.
View at: Publisher Site | Google Scholar
S. Feng, H. An, H. Li et al., “The technology convergence of electric vehicles: exploring promising and potential technology convergence relationships and topics,” Journal of Cleaner Production, vol. 260, no. 7, Article ID 120992, 2020.
View at: Google Scholar
H. Miao, C. Tang, and L. Luo, “Sales forecast of new energy vehicles based on ARIMA model,” Enterprise technology and development, vol. 2020, no. 10, pp. 97-98, 2020.
View at: Google Scholar
J. Wu, M. Hao, and J. Hang, “The Construction of enterprise innovation ecosystem in the “Internet+” environment: the case study of NIO NEV,” Soft Science, vol. 35, no. 5, pp. 70–77, 2021.
View at: Google Scholar
X. Fang and L. Zhan, “Prediction and analysis on new energy vehicle charging pile technology at home and abroad: based on lifecycle and international patent classification,” Science and Technology Management Research, vol. 41, no. 8, pp. 148–155, 2021.
View at: Google Scholar
X. Xu and M. Gui, “Research on identification and layout of core technologies of new energy vehicles based on machine learning,” Science and Technology Management Research, vol. 41, no. 9, pp. 96–106, 2021.
View at: Google Scholar
X. Tang and J. Liu, “Automatic indexing of questions in Q&A community based on BERT and TF-IDF——taking the CNGOLD Q&A community as an example,” Information Science, vol. 39, no. 3, pp. 3–10, 2021.
View at: Google Scholar
P. Guan and Y. Wang, “Identifying optimal topic numbers from Sci-Tech information with LDA model,” New Technology of Library and Information Service, vol. 2016, no. 9, pp. 42–50, 2016.
View at: Google Scholar
D. Qian and J. Zheng, “Research on subject extraction and evolution of information service literature based on LDA topic model,” Digital Library Forum, vol. 2019, no. 10, pp. 16–22, 2019.
View at: Google Scholar
J. Zhu and L. Hu, “Comparative analysis of RMB exchange rate forecast based on ARIMA and BP neural network: take the exchange rate of US dollar to RMB as an example,” Journal of Chongqing University of Technology (Natural Science), vol. 33, no. 5, pp. 207–212, 2019.
View at: Google Scholar
J. Qian and Y. wang, “Application comparison of several statistical software to build ARIMA mode,” Statistics and Decision Making, vol. 34, no. 16, pp. 80–84, 2018.
View at: Google Scholar

Copyright

Copyright © 2022 Renjie Hu et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

PDF Download Citation

Download other formats

Order printed copies

Views

1115

Downloads

1118

Citations