Abstract
The performance function expressed by the expected value of the discrete event dynamic system usually does not have a closed analytical expression, so it can only be estimated by the Monte-Carlo method. The inherent limitations of the Monte-Carlo method (low simulation efficiency) and the complexity of the system bring a lot to the simulation optimization problem. A huge challenge has come. A reasonable combination of mathematical methods and computer simulation technology is an important direction to study the optimal design of DESD. On the basis of linear systems on maximal algebras, this article studies a class of multimodel systems on maximal algebras. This article presents a non-Markov DEDS time-synchronized parallel simulation algorithm and proves that the common random number method can maximize the simulation efficiency of an existing event-synchronized algorithm. We take advantage of the inherent parallelism of genetic algorithm, and combine the idea of parallel simulation and DEDS order optimization to give an optimization algorithm for discrete parameter space DEDS. This article discusses the relationship between ultimate control, internal control, and accounting information quality. This article analyzes the relationship between ultimate control right, internal control, and accounting information quality. We use the manipulated accruals calculated by the modified Jones model and the earnings response coefficient of the earnings-return model to measure the quality of accounting information, and analyze the impact of the ultimate control ratio and internal control level on the quality of accounting information. Based on the difference in the nature of the controller and the ratio of ultimate control rights, this article examines the difference in the effect of internal control on the quality of accounting information in companies with high and low ratios of ultimate control rights. The robustness tests of the two accounting information measurement models are carried out, respectively, and the results obtained are basically consistent with the empirical research conclusions. The situation is the most important to its own development. Among them, the asset turnover rate has a greater effect on the early warning of corporate financial crisis, and the status of the current asset turnover rate and the total asset turnover rate in the financial early warning is also crucial. This article studies the correlation between the overall level of internal control and the quality of accounting information, avoiding the one-sidedness caused by using a single index to measure the level of internal control.
1. Introduction
Facing the downward pressure of the macro-economy, my country’s large-scale discrete manufacturing industry has strengthened its cost control and improved the competitiveness of its products [1]. Large-scale equipment also needs to be assembled and debugged on-site to complete a real product delivery. Facing the domestic overcapacity and the implementation of the national “One Belt One Road” strategy, more and more large-scale equipment discrete manufacturing companies have chosen the “going out” strategy, and the EPC general contracting model is a commonly adopted form [2]. For large-scale discrete manufacturing groups, the cost control chain is not only limited to the equipment manufacturing link, but also reflected in the equipment installation, debugging, and construction stage. The challenges it faces are not only from customers but also from challenges in the same industry [3]. In this situation, if the manufacturing industry wants to achieve better performance, it uses a variety of methods, including enterprise resource planning software, to achieve cost control and optimization, thereby maximizing the economic benefits and management level of the enterprise [4]. The products of large-scale discrete manufacturing companies are mostly large-scale equipment and components. These characteristics determine the importance of their product project costs to the company. Before organizing production, large-scale discrete manufacturing requires a long period of preparation work, mainly technical preparation, production preparation, and material preparation [5].
A single product produced by discrete manufacturing has many parts, huge body position, and long manufacturing and assembly processes. Costs can be managed in accordance with the work breakdown structure, which provides strong independence [6]. In addition, it also highlights the cost of discrete manufacturing companies. Precisely because of the complex structure of large-scale discrete equipment, multiple levels such as product sets of parts and components generally need to be subdivided according to the progressive relationship of production cycle and product structure, and the bottleneck of corporate resources is mainly concentrated important parts (or main parts) [7]. These components and parts are the key control points in the cost control process of large-scale discrete equipment. Therefore, large-scale discrete manufacturing enterprises must start from every key point in the entire process and pay attention to the cost management of product projects [8]. In the context of the full implementation of the current key audit matter standards, the perspective of audit practice and related theoretical research can be enriched, and it is helpful to improve the key audit matter theory and related system. The multidimensional statistical practice of key audit matters can expand statistical methods, and provide reference and reference, and expand the scope of research in this field. There are also innovations in empirical research methods [9]. Chi-square and expandability tests are carried out to verify the impact of specific key audit matters on auditing [10]. The impact of quality and the quality of accounting information disclosure enriches research methods and provides a documentary basis for follow-up research [11]. The number of key audit matters disclosed has a positive effect on the quality of auditing and accounting information disclosure.
Disclosure of key audit matters has practical significance for users of audit reports. Higher quality audit reports can provide more effective information, better display the work content and results of auditors, meet the needs of information users, and improve information relevance and decision-making usefulness; effective accounting information disclosure can help companies. To obtain economic and social benefits, good accounting information disclosure quality can play a positive role, helping financial report users use financial reports more effectively, enhancing information perception and reducing ineffective investment. In addition, it provides practical guidance to stakeholders after the implementation of the key audit matter standards.
This paper analyzes the current research status of financial crisis early warning models at home and abroad, as well as the reasons for the formation of financial crises, combined with panel data composed of financial indicators and corporate governance indicators, and through the discretization of panel data to form time-period genes, delete those that have little impact on financial crises. Indicators established a financial crisis early warning model and analyzed the influence of weights on the model. In the process of index selection, this model needs to determine the decision threshold value V and the decision threshold value D when predicting. This is a continuous learning process, so the model has strong reliability and robustness.
2. Related Work
The key audit matters mentioned in the audit report have affected the key information concerns of information stakeholders and enhanced their understanding of accounting information. Researchers pointed out that the implementation of key auditing standards has improved the readability of financial reports and strengthened auditors’ responsibilities; at the same time, it has helped information users gain better information perception, which will help the public rebuild their confidence in audit work [12]. The changes in the report found that the new reporting systems in the United Kingdom and the United States are related to the improvement of financial reporting quality, mainly due to the significant decline in the absolute abnormal accrued profits of companies and the tendency to just meet or exceed analysts’ expectations, and the profit response coefficient has increased significantly [13]. The results show that the implementation of the new audit report can significantly improve the quality of the financial report information, but it has not been found that the implementation of the new reporting system causes a significant increase in audit costs. The researchers investigated the potential impact of key audit matters in the audit report in accordance with the requirements of the new international auditing standards. In the survey, professional investors and nonprofessional investors were fully considered [14].
Domestic scholars have begun to use financial data to conduct research on financial crisis early warning. The introduction of nonfinancial indicators into the financial crisis early warning model has also begun to develop. Many scholars have published academic results on the financial crisis early warning model. Researchers used discriminant analysis methods to establish a prediction model for my country’s corporate bankruptcy earlier. Among them, the univariate prediction model uses the method of mean interval estimation [15]. The results show that although the method is simple, easy to understand, economical, and applicable, and when appropriate indicators are selected as predictors, it can also obtain a fairly high forecast accuracy rate, but how to choose suitable indicators is quite difficult [16]. Therefore, the forecast accuracy rate of this method is generally low. This method is suitable for the situation where a complete financial system has not been established in the early stage; the multivariate model adopts a linear discriminant analysis method; this method can effectively reflect the comprehensive quantitative characteristics and differences of bankrupt companies and nonbankrupt companies, and generally has high predictive ability, but it requires more index data and other information, and high computational complexity, and the cost is greater [17].
Studies have shown that as the time of discretely driven approaches, the influence of financial indicators on discretely driven is gradually increasing. When the company has not yet formed a financial crisis, the influence of nonfinancial indicators such as corporate governance is more obvious than that of financial indicators [18]. The financial crisis early warning model can greatly improve the company’s ability to predict whether it will be discretely driven in the future. It has shown that the introduction of nonfinancial indicators of financial crisis early warning models can greatly improve the accuracy of the model’s predictions [19]. They proposed a KMV model based on option pricing theory, introduced the concept of default distance, and constructed a financial crisis early warning model using financial data and transaction data as sample data. The results showed that the model has strong explanatory and predictive capabilities [20]. But the stock transaction data have no great influence on the information contained in the model, so the influence on the prediction accuracy of the model is also small. Randomly select 30 financially normal companies as samples, among which 5 discretely driven companies and 10 financially normal companies are used as test samples, first standardize the indicators, and then use the mixed method of principal component analysis and discriminant analysis to obtain the company’s own financial [21]. The situation is the most important to its own development. Among them, the asset turnover rate has a greater effect on the early warning of corporate financial crisis, and the status of the current asset turnover rate and the total asset turnover rate in the financial early warning is also crucial.
3. Discretized Panel Data Financial Analysis Model
3.1. Discrete Event Data Classification
In large discrete manufacturing companies, each situation is different. Make cost project budget before bidding, implement cost analysis, and control plan in the whole process of design and process, and finally perform total cost accounting. MTO (Make to Order) mode is generally widely adopted by large discrete manufacturing industries. Production revolves around orders, and it is determined that it must be able to organize production according to orders at any time to meet customer needs. After the announcement of the tender, discrete manufacturing companies begin to design procurement costs (material procurement and outsourcing parts), manufacturing costs (including tooling preparation cost), design cost and design target cost, service cost mainly for product packaging, transportation, installation, and commissioning, quality cost and period cost, and all-round cost control, as shown in Table 1.
The principles of discrete data segmentation include equal distance, equal frequency, and optimization. Equidistant division is to divide the range of continuous variables into N segments on average, and the group distances for each segment are equal; equidistant division means to divide sample points into N parts, each containing the same number of sample points. Among them, the ordering method of the samples is to arrange them in time or other order of size; the optimal discrete method is to comprehensively consider the independent variables and the dependent variables, and the dividing point is the point that causes the dependent variable to change significantly; that is, the information increment is large. Isometric division is discretized according to the original distribution of the data. The more discrete the paragraphs, the better the original appearance of the data, but at the same time, it will increase the computational complexity as shown in Figure 1.
Equal frequency distribution treats the data as an even distribution, but this method of segmentation will result in the same value in different segments. Although optimized discreteness satisfies the characteristics of the original data and the target variables, this method is more troublesome, and the increment between some target variables is not obvious. Although the discretization of continuous variables has been criticized by scholars, researchers have pointed out that if the two ends of the distribution that the variables obey are also meaningful, it is easy to ignore the influence of the two ends of the distribution that the variables obey. And the distribution of the variable does not truly reflect the actual situation of the variable, so it is wise to discretize the value into several points. The results show that discretizing the data into 3-4 segments can get better results. In short, the discretization of continuous data not only facilitates the understanding and use of the model, but also increases the discriminative ability of the model.
Discretization processing will cause part of the information to be lost, eliminating the differences in the same paragraph and also eliminating the direct connection between different paragraphs. The discretized characters do not have the meaning of actual numerical value, and they do not have the relationship that continuous data can be calculated. For example, A certain index value Ai of a certain company is greater than the index value Bi of another company. If Ai and Bi are discretized as the same character “01,” the discretized characters indicate that there is no difference between Ai and Bi; if Ai and Bi are discretized as different, the characters “00” and “01” can only indicate that the value in the segment represented by “00” is higher than the value in the segment “01.” Because the specific value represented by the two characters after being separated cannot be expressed, only the Hamming distance can be used for analysis.
3.2. Classification of Financial Data
The index value is continuous, but it is difficult and complicated to consider the continuous data of multiple periods at the same time. In order to make the research more convenient, this article deals with the continuous index data according to the appropriate discrete reduction method. = {“00,” “01,” “10,” “11”}. The data discretization method used in this article is to discretize a certain indicator of a certain enterprise; first of all, the industry m of the enterprise must be obtained, and then discretize the data into S according to the following formula and the mapping relationship shown in Figure 2:
Observing the data in the annual financial statements, each industry has some indicators that are more or less large or small. If an average is used, the large or small data of the indicator will cause the average of the indicator. Impac:, you need to eliminate large or small data. However, these data also represent these indicators to a certain extent, so the use of the median not only eliminates the influence of large or small data, but also represents the normal situation of the indicator. For example, in 2019, there were 39 listed companies in industry A. For the operating income indicators, the median was 517,984,967.95, the median of 3/4 was 1060297192, and the median of 1/4 was 257,465,463.96.
In 2019, the maximum value of industry A’s operating income was company, the value is 15387490907, the second largest value is shares, the value is 6749444707.07, and the maximum value is about 2.3 times the second largest value; using the median, we can well eliminate the maximum value of the discrete result. Impact: especially for industries that contain fewer companies, the maximum value of an indicator can easily affect the average value of the indicator in the industry. Using the formula, it can be concluded that the operating income indicators were discretized into the string “11” in 2019.
This article takes holdings, a company named discretely driven by the securities regulatory commission in 2020, as an example. The company’s return on equity (operating profit) indicators are 0.0112, −103.7798, and −103.7798. The medians of 3/4 are −36.762, 8.026, and 146.685 respectively, so according to the formula, the indicator was discretized into “01,” “00,” and “00” in the three years. The trend is shown in Figure 3. If only the four indicators in the figure are selected to form the company’s time-period gene, it is 010000110101000101101011. So the author of this article believes that weight analysis is not necessary. However, for the sake of scientific rigor, this article still conducts a weight analysis, because the financial crisis does not appear suddenly.
This article uses statistical analysis methods to calculate the j-th index in financial crisis companies and healthy companies in tp year. In a financial crisis company, the probability of occurrence is
Among them, Nst represents the number of companies in financial crisis and P represents the time-period genes of the financial crisis companies in the sample. It means that in a financial crisis company, the discrete string of the indicator in year is the number of companies. In the same way, in a health company, the probability of occurrence is
Among them, NF discretely driven represents the number of health companies and it represents the time-period genes of healthy companies in the sample:
The difference between the probability of occurrence in discretely driven companies and nondiscretely driven companies is as follows:
When the indicator j is selected into the financial crisis early warning model, then discard this indicator and finally get the optimal financial crisis early warning model as shown in the following formula:
The choice of the V value in this model is also very important. If the V value is smaller, more indicators will be selected into the model, which will increase the accuracy of the model’s prediction and at the same time increase the complexity of the model’s prediction, and make some. The selection of unimportant indicators into the model will also affect the prediction accuracy of the model; on the contrary, if the value of V is large, many important indicators will not be selected into the model. Although the complexity of model prediction is reduced, it will also affect the prediction accuracy of the model. The prediction result has an impact, so the value of V requires repeated learning to get the result. The financial crisis early warning model indicates that if a company’s time-period genes are closer to this model, the company is more likely to fall into financial crisis.
3.3. Event Weight Analysis
The weight represents the position occupied by an indicator in the entire model. The previous regression analysis methods and clustering methods used weights to optimize the model. On the other hand, perform correlation analysis between the indicator and the dependent variable, and retain the indicators with higher correlation with the dependent variable. Although these methods retain the main indicators, they are all. It weakens the degree of influence between different indexes on the dependent variable and makes the index with higher influence on the dependent variable not particularly important. At this time, it is necessary to use the weight method to distinguish the critical degree of different indexes. Since this article did not perform correlation analysis and principal component analysis, and used the indicator with the largest model difference between discretely driven and nondiscretely driven companies as the selected indicators, the contribution of different indicators to the financial crisis was distinguished, but the original contribution of the indicator was retained. Based on experience, the indicator of the most recent year should be given a greater weight relative to the Hamming distance of the model, and the earlier should be given the lower weight, and then, the prediction accuracy rate is used to analyze the influence of the weight on the model, as shown in Figure 4.
The traditional clustering method uses coefficients to reflect the difference in cross-sections, so data mining cannot be performed from a large number of cross-sections, and only static problems can be solved. When performing cluster analysis on panel data containing multiple indicators, it is necessary to consider not only the distance between all indicators contained in each sample, but also the change characteristics of each indicator at different points in time. At present, many scholars have improved the distance calculation method in the clustering algorithm to make it suitable for solving the problem of multi-index panel data clustering. Most of them can achieve relatively satisfactory results, but due to the current large society. In the data age, a large amount of data are generated every second, which leads to a sharp increase in the amount of calculations, and it becomes more and more difficult to find the optimal clustering results in cluster analysis. Therefore, further research is needed to investigate whether the introduction of the genetic algorithm in cluster analysis is effective or not.
4. Case Analysis
According to the principle of the abovementioned financial crisis early warning model based on discretized panel data, the general process of empirical analysis in this article is shown in Figure 5. Empirical analysis first requires sample selection; then after multiple studies for index selection, the model is finally obtained, and then, the model is used to predict the verification sample; the reliability and robustness of the model are judged according to the prediction accuracy; and the weight analysis is performed. Determine whether the weight has an impact on the model, compare the financial indicator model and the model that contains corporate governance indicators, get the importance of corporate governance indicators, and get the final model. The regression analysis method requires correlation analysis between the indicators, which is relatively high. Retain only one of the indicators, or use the method of principal component analysis.
According to the learning process of the distance judgment threshold D, it can be seen that it may be because D trained based on the learning sample is not suitable for the verification sample, and the value of D may be too small, or it may be because the verification sample capacity is too small, resulting in discrepancies. This article first selects 28 financial indicators (including current ratio indicators, cash debt ratio indicators, return on net assets indicators, and net profit rate indicators) and 15 nonfinancial indicators (the two positions of chairman and general manager indicators). The Z index, the chairman’s shareholding ratio index, etc. are used as alternative indicators for forecasting. Among them, 39 indicators are discretized into binary strings according to the data discretization method, but because there are 4 indicators (whether the chairman of the board of supervisors has changed, whether the general manager has changed, the setting of the two positions of the chairman and the general manager, and the type of actual controller of the listed company). It is its own index and cannot be discretized, so the corresponding discrete string is assigned according to its specific value. For example, the indicator of whether the chairman of the board of supervisors has changed has only 0 (representing no change) and 1 (representing change). When it is 0, it is discretized as 00, and if it is 1, it is discretized as 01. According to the selection indicators and discretization rules and the generation method of time-period genes described in 4.3, the time-period genes of each company are generated. For example, for companies that were discretely driven in 2010, as shown in Table 2, first of all, the panel data composed of 43 indicators in the three years of 2019 and 2020 refer to the median of the indicators corresponding to the production and supply industries, and are discretized according to the formula to form the company’s time-period genes.
This article uses the improved Hamming distance to calculate the distance between each company’s model and the optimal model. According to the judgment threshold of the set distance D, predict whether each company will fall into a financial crisis. The prediction accuracy rate obtained by different D is different. This article uses learning samples to determine the value of D and obtains the results. It can be seen from Figure 6 that when D = 17, the average prediction accuracy rate of the discretely driven company and the nondiscretely driven company is the highest, which is 86.8%. When D is getting smaller and smaller, the prediction accuracy rate of the discretely driven company is getting lower and lower. When it gets bigger and bigger, the forecast accuracy rate of nondiscretely driven companies gets bigger and bigger. Only when D = 17 can the prediction accuracy of the discretely driven company and the nondiscretely driven company reach equilibrium. After this learning process, the decision threshold for determining the distance is set to D = 17.
This article uses learning samples and verification samples (the verification samples are 17 listed companies that were discretely driven in 2011 and 17 listed companies that were not discretely driven, and the index data from 2008 to 2010 is the sample data), and the index values of the selected indicators are used as samples. According to the data, the period genes of the learning sample and the verification sample are obtained. According to the improved Hamming distance di and the distance judgment threshold D = 17, it is predicted whether the sample company will be discretely driven in year t; if > D, it is predicted that the company will not be discretely driven at time t; otherwise, it is considered to be discretely driven. The prediction accuracy rates of the obtained learning samples and verification samples are shown in Figure 7. Under the entire logical chain, internal control has effectively alleviated this shock effect by improving the comparability of accounting information. Therefore, this article provides empirical evidence for the logical chain of internal control-accounting information comparability-economic consequences.
The results show that the prediction accuracy rate of the discretely driven company and the nondiscretely driven company in the learning sample is 86.8%, which means that the prediction accuracy rate is high. In the verification sample, the prediction accuracy rate of the discretely driven company is lower, which is 82.4%. The accuracy rate is 88.2%, which is relatively higher than the prediction accuracy rate of nondiscretely driven companies in the learning sample. Because the formation of a financial crisis is an accumulation process, for financial indicators, the performance of financial indicators in recent years is more prominent, and financial indicators in more distant years are not very obvious to the formation of financial crisis; similarly, because the corporate governance situation will suddenly lead to it. The company is in a financial crisis, so the Hamming distance of recent years should be given a higher weight.
After weighting, the prediction accuracy of the learning sample has not changed compared to the optimal model. In the verification sample, the prediction accuracy of discretely driven companies is obviously higher, but the prediction accuracy of nondiscretely driven companies is significantly lower, and the overall average prediction is accurate. The rate does not change, indicating that the weight cannot increase the prediction accuracy of the model, so the model is not affected by the weight. This may be because the correlation analysis or principal component analysis of each indicator in the model is not performed, so the indicator pair is not changed. The contribution of the model, but retains the original contribution of each indicator to the model, or because the sample size is too small, the specific reasons need to be analyzed.
5. Conclusion
This article combines panel data to define the symbols that appear in this method and then proposes the discrete method of this article-the median method, and then proposes the characteristic theory of this method-period gene. The period gene combines discretized panel data. The time series is integrated into the indicator value, which shows the basic characteristics of the data well. After that, you need to learn to determine the V value by analyzing the difference between the indicators of the discrete dynamic event-driven company and the nondiscrete dynamic event-driven company, and obtain the selected indicators to form a financial crisis early warning model, and then determine the company’s time series and financial crisis. The Hamming distance between the early warning models is predicted by determining the D value multiple times and then performing a weight analysis to analyze whether the weight affects the model. The increase in the comparability of accounting information can reduce the information impact of IPOs on incumbent companies and ease the impact on their stock prices. Under the entire logical chain, internal control has effectively alleviated this shock effect by improving the comparability of accounting information. Therefore, this article provides empirical evidence for the logical chain of internal control-accounting information comparability-economic consequences and finds that comparability has an incremental explanatory power compared to reliability. Through empirical research, the multiple regression analysis method is used to clarify the impact of carbon information disclosure and accounting conservatism of enterprises with different property rights on financing constraints, and combined with the empirical research results to reduce the financing differences of entities with heterogeneous property rights and improve the level of corporate carbon information disclosure. In the future, the performance of financial indicators in recent years is more prominent, and financial indicators in more distant years are not very obvious to the formation of financial crisis.
Data Availability
The data used to support the findings of this study are available from the corresponding author upon request.
Conflicts of Interest
The authors declare that they have no known competing financial interest or personal relationships that could have appeared to influence the work reported in this article.