Abstract
HRMS is a very critical tool for companies. The recruitment text contains rich information that can provide strong information support for the company’s recruitment work and also improve the efficiency of job seekers in finding job opportunities. To this end, for the problem of multilabel text classification of recruitment information, this paper provides two algorithms for multilayer classification based on supported SVM. First, the same learning subclass method is used for text sorting subclass acquisition, and then, the class of the text is determined. Second, the hemispherical support SVM is used to find the smallest hypersphere in the feature space that contains the most text of that class and segment the text of that class from other texts. For the text to be classified, the distance from it to the center of each hypersphere is used to determine the class of the text. Experimental results on recruitment data demonstrate that the algorithm in this paper has a high check-all rate, check-accuracy rate, and F1. And, the relationship between HRM activities and corporate performance is discussed.
1. Introduction
More and more data are generated by the development of enterprises, and with such a huge amount of data, it is simply impossible to implement human efforts to manage and analyze the data, thus giving birth to information management systems. Information management system [1–3] is a kind of management software with functions such as data collection and database creation and data management, which can manage and analyze a large amount of data in a unified way. HRMS combines human resource management and information technology, which is not only an information processing tool but also a resource management specification whose main purpose is to standardize the business processes of HR departments, centralize HR information, and improve the transparency of HR management through the system [4]. Such a huge amount of data has exceeded the processing capacity of traditional HR management systems, thus making it impossible to properly manage and analyze data, which is described as “data-rich, but information-poor” [5, 6]. This situation is described as “data-rich but information-poor.” Moreover, most of the enterprise-level HR management systems have not been given full play, only collecting and storing data, but not effectively managing and analyzing the data and information, which not only wastes the limited resources of the enterprise but also loses the opportunity to use big data for rapid development [7, 8].
In the context of the traditional HR management system cannot meet the needs of enterprise big data, in order to make the system work properly, it is then necessary to artificially perform certain screening process on the massive data, so as to reduce the data volume level of big data, but this processing method seems to be back to the old era of manual data processing, which seriously hinders the improvement of work efficiency. For this reason, data mining [9] has been proposed to reduce the amount of data to be processed. The purpose of data mining is to analyze the large amount of raw data and discover the important data patterns and finally convert the raw data into useful information and knowledge.
Data mining is the use of various techniques and statistical methods to mine useful information and knowledge from large amounts of data [10].
2. Related Work
Human resource management system is an important tool for enterprise management and plays an important role in enterprise development, which has been generated at the present time. With the advent of the era of big data in the 21st century, it has brought some impacts and changes to HRM [11–13].
Jiang et al. [14] analyzed in detail the important role of data mining in HRM for talent recruitment, talent management, talent loss prevention, and improving HR precision management. Zhang et al. [15] demonstrated the capability of data mining in improving the quality of decision-making process in the HRM system. Rai et al. [16] analyzed and implemented the application of data mining in separation management. Xia et al. [17] proposed a knowledge discovery model that helps to support companies in predicting the future performance of their employees and then assigning the right people to the right positions and projects. Martínez-Ferrero et al. [18] proposed a fuzzy clustering method based on adaptive genetic algorithm, which provides effective data analysis for employee performance appraisal in modern corporate HRM. Bajo et al. [19] used artificial neural networks for turnaround prediction, knowledge-based search engine for candidate search, genetic algorithm for staff scheduling, text mining for HR sentiment analysis, resume data collection and information extraction, and employee self-service with interactive voice response. Trauth et al. [20] designed an improved algorithm based on support vector regression for HR demand forecasting in enterprises with good practical reference value. Walmsley [1] applied extreme gradient boosting (XGBoost) technique to predict employee turnover in order to solve the turnover problem and demonstrated its higher accuracy in predicting employee turnover compared to other supervised classifier algorithms. Brito and Oliveira [21] analyzed the factors affecting employee retirement based on work-life profiles and used machine learning methods to create a retirement prediction model that could provide a solution to the human resource outflow problem.
2.1. Data Acquisition
2.1.1. The Nature of Recruitment Content Classification
It is the process of classifying information about a job set in a text form and the process of automatic classification of documents using text, which is one or more predefined categories in the document. In this paper, text-based classification techniques about machine learning and statistical methods have become the main techniques in the area. The idea of classification is based on various labeled texts, on a relational model between features and document categories, and then on a relational model to determine the category to which the test text belongs [6, 22–25]. Text classification is divided into two processes: learning and testing. As shown in Figure 1, the learning process begins with preprocessed text; then, using any mathematical description, the previously processed text is represented as a computerized form of recognizable data, which is then represented with the help of a classification algorithm that receives training text data and then receives the appropriate classifier. As shown in Figure 2, the testing process is the process of preprocessing the text and the mathematical description after recognizing it by the classifier to obtain the text category.
2.1.2. Data Acquisition
The data used in this article is for the 58 Tongcheng open source data [11, 26], as well as other open source data on the site; the link is https://download.csdn.net/download/huazi99/11964169https://download.csdn.net/download/weixin_%2044028406/15801929.
2.1.3. Data Preprocessing
This section will introduce some knowledge about text classification, presenting the techniques used in the following experiments.
2.1.4. Chinese Word Splitting
In natural language, words are a relatively important presence, and most words have semantic information. At the same time, words are also the main components of a complete sentence. Because in English, an English sentence is composed of English words separated by the use of space characters, the division is relatively simple; whereas Chinese sentences are composed of each individual word as a unit, and Chinese phrases consist of a series of consecutive words, each without a separator. Therefore, the division into Chinese is more difficult than that of English [26–31]. There are three ways: string matching-based segmentation, statistics-based Chinese segmentation, and comprehension-based Chinese segmentation.
2.1.5. Removing Deactivated Words
Deactivators are words in Chinese text that do not contain actual semantic meaning and do not have a certain degree of distinction. In the text information, with the presence of stop words, the whole text looks more fluent, but in the text analysis, the presence of stop words interferes with it; these stop words not only do not contain effective information and cannot accurately express the meaning of the sentence but also may have a suppressive effect on other characteristic words.
2.1.6. Feature Selection
After the text data are divided into Chinese words and deactivated words are removed, many words will be obtained and feature selection is needed. In order to further improve the model and algorithm, good feature selection has the most important value. The methods of feature selection are cardinality statistics [32], mutual information method [33], TF-IDF method [34], and bag-of-words model [35].
2.2. Multilabel Text Classification
2.2.1. Hyperspherical SVM
Given a training sample set of class and a kernel function , where , corresponds to the inner product in the feature space, i.e., If there are some examples of remote mappings, then we find the minimum HILER that contains most of the examples. However, when we are not sure if there are remote mappings of samples,
To solve the above constraining optimization problem, the following Lagrangian function can be constructed:where is the Lagrangian multiplier of the class. To obtain the minimum of equation (2),
Substituting equations (1)–(3) into equation (4), the pairwise optimization problem is
According to equation (4), a weighting factor is linearly composed of
If , then the corresponding sample is called the support vector.
If it is , the corresponding sample is located near the surface of the pole.
If , the corresponding sample is called an outlier or noise sample.
2.2.2. Improvement Algorithm
For the sample x to be classified, then determine whether the sample x belongs to class m according to and :
If satisfies one, then calculate the affiliation of x belonging to class m from
According to equation (7), the category to which sample x belongs can be determined:
Our algorithm is described in detail as follows： Step 1: for X, we calculate according to equation (6), . Step 2: if one or more classes meet , sample x is marked as class m goes to step 4. Step 3: calculate use equation (7) and R use equation (8). X is labeled class . Step 4: end.
3. Experiments
The experiments were conducted on collected 58 Tongcheng data with five categories and 685 job postings, each with a class number less than 4. Four hundred and forty five articles were used as training samples (Table 1). The feature dimension is reduced by using information gain and feature weight of each word.
Precision, recall, and F1 are used to evaluate the classification performance. The definitions are as follows:
Nr is the actual number of classes for each sample:
If n is the number of all test samples, then the macro is the mean average accuracy (MAAP) [36]. If n is the number of test samples with the same number, then it is called microaverage accuracy (MIAP):
The implementation of the algorithm refers to [12], and the parameters of the algorithm are set as shown in Table 2.
The classification results are shown in Tables 3–5.
The experimental results demonstrate that, in this paper, the same approach applies to smaller training modules. Our method technique applies to a wider and broader range of educational materials.
In order to obtain a good human resource management effect, it is also necessary to make the effect of human resource management to be maximized through an appropriate performance appraisal system; it can be said that performance appraisal also has the following important significance in human resource management, as shown in Figure 3, and it can be concluded that having different academic qualifications and performance appraisal is an important basis for personnel appointments. Through performance appraisal, the situation of employees can be evaluated, so that the working attitude and status of each employee can be clarified, and good support for the arrangement of personnel positions can be provided.
As shown in Figure 4, the average salary of employees can be used to control the application of HRMS for staff scheduling and job promotion, so that the role of HRM can be fully utilized. In order to give full play to the role of talent, enterprises are required to have a detailed understanding of the work of employees in the process of hiring, through the performance appraisal system to clarify the attitude and work situation of employees during the work period, to conduct an objective comparative evaluation of employees, and to provide sufficient reference basis for employee promotion, so that enterprise human resource management can be reasonably realized. Therefore, performance appraisal has a very important significance in the process of human resource management job scheduling and also has an important significance for the optimal allocation of human resources.
As shown in Figure 5, such a performance appraisal is an important mechanism for rewarding salary, and in the modern enterprise management project, it is necessary to follow the principle of fairness and efficiency in salary allocation. Through the application of the performance appraisal system, it is possible to reasonably allocate the salary based on the actual working conditions of the employees and to guarantee the reasonableness of the salary allocation, which is also an important means to guarantee the employees’ motivation and enthusiasm for work.
4. Discussion and Analysis
In the process of constructing the performance appraisal system of modern enterprises, the purpose of the appraisal needs to be clear first, and on this basis, a targeted appraisal system should be constructed. The object of performance appraisal in enterprises is employees, so it is also necessary to combine the performance and work attitude of employees on the basis of a reasonable calculation of performance appraisal results and, on this basis, to carry out targeted management of employees so that the effectiveness of performance appraisal can be given full play. In the modern enterprise, human resource management work also needs to strengthen the importance of the assessment process to ensure the fairness and reasonableness of the performance appraisal. It can be said that the purpose of performance appraisal in enterprises is not only to do a good job in the management of employees but also to improve the business level of employees as an important management goal. Therefore, human resources managers can also give targeted training guidance according to the actual assessment situation of employees to improve their working ability and comprehensive quality. Therefore, enterprise managers also need to clarify the purpose of the performance appraisal.
Enterprises can build a reasonable performance appraisal system according to the existing market conditions and their own development, which can lay a good foundation for the follow-up performance appraisal. In the process of building the performance appraisal system, enterprises are required to clarify the status qua of other management systems outside of human resource management, to ensure mutual promotion and compatibility among the management systems and to avoid isolation of the performance appraisal system in the application process.
The fairness of performance appraisal is an important principle to play the role of performance appraisal, so the enterprise also needs to fully follow the principle of fairness in the process of construction and implementation of the performance appraisal system, so as to avoid the influence on the employees’ work emotion and protect the employees’ work enthusiasm. In the process of performance appraisal, managers are required to disclose the evaluation criteria of performance appraisal to employees, which can not only make employees clear the work requirements of their positions but also enable employees to clarify their own shortcomings in the work process and correct them by ways of reference and comparison so that the role of performance appraisal can be fully played. In addition, there are more performance appraisal methods in the process of performance appraisal, and the scope of application of different appraisal methods also has certain differences, which also requires enterprises to combine their own development characteristics based on a reasonable choice of performance appraisal methods, to ensure the fairness of the appraisal process to achieve the expected performance appraisal effect.
5. Conclusions
This paper provides two algorithms for multilayer classification based on supporting SVM. First, the same learned subclass method is used for text sorting subclass acquisition, and then, the class of the text is determined. Second, a hemispherical support SVM is used to find the smallest hypersphere in the feature space that contains the most text of the class and segment the class from other texts. HEM can anlayse the actual assessment of employees to give targeted training guidance to improve the work ability of employees and comprehensive quality. Therefore, enterprise managers also need to clarify the purpose of performance appraisal so that they can give full play to the role of performance appraisal in the financial management work.
Data Availability
The dataset used in this paper are available from the corresponding author upon request.
Conflicts of Interest
The authors declared that they have no conflicts of interest regarding this work.
Acknowledgments
This work was supported by National Social Science Fund Project: Research on Inter-organizational Knowledge Sharing Behavior and Its Dynamic Incentive Mechanism in Differentiation Context, under Grant no. 20BGL126.