Research on Enterprise Financial Management and Decision Making based on Decision Tree Algorithm

Transcription

1 Research on Enterprise Financial Management and Decision Making based on Decision Tree Algorithm Shen Zhai School of Economics and Management, Urban Vocational College of Sichuan, Chengdu, Sichuan, China Abstract In the current big data background, data mining and decision making algorithms are widely used in financial management. In this paper, the author analyse enterprise financial management and decision making based on decision tree algorithm. The enterprise can establish the financial management model by the result of data mining, and then establish a whole set of financial decision system. The empirical analysis shows that the decision model proposed in this paper can well analyse the financial status of enterprises, and can realize the functions of enterprise financial analysis and decision-making management. In the enterprise financial diagnosis decision stage, the enterprise managers should not only pay attention to the financial management status reflected by the financial statements, but also need to comprehensively consider the environment of the receiving enterprise, as well as the changes in the specific operating conditions. Key words: Data mining, enterprise financial analysis, decision model, information retrieval 1. INTRODUCTION In the era of information explosion, people get more and more ways to obtain information, and obtain all kinds of information from the network become the main source of information, investors try to obtain valuable information, so as to provide data support for future investment decisions. Enterprises have accumulated a large amount of data in the operation and management, and every financial website also publishes the letters of every large listed company every day. Managers need to analyze these data and make various decisions. Mining useful information from massive data is a fundamental way to use data mining technology, managers can build the object model using the data mining results, and establish a set of decision-making system, is conducive to the effective control of management decision-making information to make the right decision, so as to improve the competitiveness of enterprises. Powerful data mining technology is widely used in various fields. In the field of scientific research, modeling and simulation technology for data mining research, help realize the intelligent retrieval of data mining and simple information. Decision makers can also dig out the characteristics of consumer behavior habits, and then in the marketing activities for consumer consumption habits, so as to restructure the product. In the field of finance, securities analysts use data mining technology to track the changes of stock price. In the enterprise diagnosis, financial diagnosis is an important link, is the basis of the enterprise diagnosis, for example, provided higher are required to assist in the management of the financial diagnosis, improve management loopholes, eliminate the efficiency of resource development, in order to achieve the enterprise the new management model to guarantee the survival and development of enterprises. Therefore, the financial problems of diagnostic enterprises become more important. Financial diagnosis not only diagnoses and analyzes the internal production and sales reports of enterprises, but also puts forward suggestions for improvement, which is of great help to the investment decisions of all stakeholders. Though each enterprise s financial characteristics of cost management, financing channels, intangible assets and profit distribution are different from other industries, relatively speaking, the characteristics are not obvious, so the results obtained through the research selecting universal relevant indicators can be applied to comprehensive research. At present, the definition of China's financial distress is from the point of view of the financial crisis, including two aspects: one is that funds cannot repay loans and mortgage; the other is that funds are unable to repay the matured debt; and also a state between the two aspects. The continuous existence of this state will lead to the bankruptcy, acquisition and liquidation of enterprises. Also, there are scholars from the profit point of view regarding the financial distress as a process of gradual accumulation. When the result of bankruptcy appears, enterprises face the phenomena of defaults, losses, and insolvency in the early phase. The financial distress is divided into failure in the process of business operation, business failure of the creditors being unable to recycle all claims, and technical failure caused by improper enterprise scheduling and other conditions. In fact, this division does not meet the actual requirements, as it does not consider the financial balance phase appeared in the operation process. In other words, it cannot judge which type the financial situation should belong to. Therefore, with the financial situation as the reference basis, the company is divided into the listed company and the non-listed company. Among them, listed companies with the financial situation as the foundation are categorized into non-ideal, relatively ideal and ideal state, so do non-listed companies. In this context, China's research on financial situation has two main aspects. One is from a quantitative point of view, using the analysis method to quantitatively treat influencing factors, so as to obtain the degree of influence 166

2 of the financial indicators on listed companies. The other is from the perspective of qualitative analysis, using the analysis and evaluation of enterprise managers to predicate the financial distress well, so as to set up programs for the development of enterprises. 2. RELEVANT THEORIES 2.1. Financial distress In the 1980s, crisis management theory was generated, which studies various difficulties of enterprises before, during and after the crisis take measures to avoid and reduce the damage of the crisis. The financial crisis of listed companies is mainly reflected in inability of the capital to mortgage loans, excessive reliance on short-term fund-raising and borrowing, the presence of large amounts of the overdue profits, and inability to obtain the normal profit. In this context, research on the plight model of enterprises tends to be more diversified and comprehensive. Analysis and prediction of financial distress has achieved a rapid development. Many scholars and researchers have also made some contributions to financial distress prediction, but the research level is not perfect due to small sample sizes, a long model running time and a low prediction accuracy. Under the influence of the situation, the operation of enterprises now shows the trend from rise to decline. Especially under the background of anti-pollution and haze governance, financial forecast is one of the important means of enterprise research. The financial situation of enterprises is still severe. In order to achieve the fundamental solution, this paper makes analysis from the angle of financial management combined with the current status of work. Analysis and predication of financial distress is the scientific forecasting and decision making based on financial indicators. Based on the overall financial situation analysis, financial distress will show a state of gradual deterioration and continuous appearance. Figure 1. Asset Management 2.2. Financial model In China, the research achievements gradually become mature. Achievement of the financial distress warning is based on the market economy cycle theory and enterprise early-warning theory. This involves the consideration of the cycle of the recession and prosperity of the market outside the enterprise, and also the phases inside enterprises from initial establishment to recession. The research method involved is the financial early-warning model based on regression analysis. Data mining technology is realized through enterprise application in the aspects of data generation, collection, storage, and processing. Moreover, data mining technology including association technology, clustering technology, classification technology and deviation analysis technology is one of the important means of artificial intelligence. Now, the basis of financial theorem is MM theorem proposed based on the capital structure. MM theorem realizes the development and application of economics in the financial field through the discussion on whether the hypothesis is true. Later, through the improvement of the theory, the financial embarrassment cost theory, the agency cost theory and the information asymmetry theory were proposed. Genetic algorithm, neural network and other algorithms have been applied in this field. Treatment of financial crisis in foreign countries changes from the single prediction model to the mixed model. The involved models include single-variable model, multi variable model and artificial intelligence model. Especially with the current development of information technology, genetic algorithm and neural network in the artificial intelligence model has been widely used in more and more financial forecasting. 167

3 Figure 2. Financial management index 3. MODEL PREPARATION This paper through the literature review and inductive analysis grasps the research dynamic and establishes the research theme. Then, it illustrates the basic principle of financial risk, discusses the enterprise risk control methods, and identifies and evaluates enterprise financial risk. Through summary, the results of the study are obtained. The following further researches the prediction of financial difficulties based on the SVM model Classification standard The financial situation is the basis for distinguishing whether the company is in the ideal state. In addition, the financial situation is also a sign which measures whether the enterprise can continue to operate. When financial situation is not ideal, enterprise development situation will be in a dilemma until the final collapse and bankruptcy. The financial situation is actually composed of a variety of indicators. Figure 3. enterprise financial model The financial situation is affected by a variety of indexes, among which the most influential one is profitability. Profitability is not only the core index among all indicators, but also the basis of other influencing factors. Therefore, this paper uses the profitability as the general financial situation evaluation index. In order to ensure the accuracy of the research, the net profit and the net assets constituting the profitability are used as the research basis to judge the financial situation of the company. 168

4 3.2. Data selection When the research achieves some results, at the same time, it still face the factors of difficult index screening, prediction theory lacking basis and the lack of real data and so on. The abnormal financial situation is actually the abnormal financial data phenomenon appeared at the time of financial distress. Abnormal data will bring certain influence to the financial forecast. This kind of influence is mainly reflected in leading the index selection and determination of the financial distress prediction model. The selection of enterprise financial data should base on the characteristics of the listed company data: easy to obtain, true and scientific, and representative. At the same time, the scale of the enterprise itself is large, with a wide range and strong expansion capability. In order to study the listed company's financial difficulties, this article starts from the perspective of objective reality, rather than just limited in the present, before and future level. Based on the principle of comprehensiveness, representativeness and sensitivity, preliminary frame of the financial distress status of listed companies is constructed. This paper altogether selects 65 Hubei's listed enterprises with a precarious financial condition as the research object. Among them, the financial situation of 15 and 30 enterprises is non-ideal and ideal respectively, and the remaining 20 belong to the relatively ideal state. In order to avoid the contingency and instability of data, improve the accuracy of model predictions, provide stakeholders with timely and effective financial information, and achieve real and effective decision making, this paper determines to use the first three years financial data of the enterprises during their financial instability as the data specimen and constructs the model to achieve predication. At the same time, in order to meet the requirements of computer simulation, in the following research process, some abnormal data will be removed Determination of the initial index Research on the prediction of corporate financial difficulties is determined by both the financial indicators reflecting the actual financial condition of enterprises and non-financial indicators. To highlight the contrast effect of the model, according to a Hubei listed company's financial situation, the indexes determined in this paper adhering to the principles of being comprehensive, representative, operable, sensitive and measurable. Figure 4.Accounting calculation Screening of enterprise financial distress prediction model indexes is divided into 9 categories and 21 sub-classes. And the indicators chosen should be able to be compared with previous literature as far as possible. Among them, solvency refers to the ability of enterprises to repay the debt of the next year. This ability can be divided into short-term and long-term aspects. Operation ability refers to the efficiency of the flow of funds, mainly including the flow efficiency of liquid capitals, receivables and inventories. Profitability is the ability of the enterprise to obtain profits. Profitability is not only related to equity returns, but also asset returns and operating profit. Growth ability actually reflects the growth situation of assets, shareholders equity, operating income and profit Difference test In order to accurately study the results, this paper makes accurate the indicators. Through the test of enterprise indicators differences, this paper finds factors with significant differences as indicators of the model. First of all, the above 21 indexes will go through the variable K-S test under the simulated condition of 5% by the computer. The result shows that the ratio of current assets, fixed assets ratio and ownership concentration conform to normal distribution and the single variable variance can be used to analyze indicators. However, the other 18 indicators do not accord with normal distribution, so the K-W H non-parametric test method is used to 169

5 test the significant degree of variables mean. The specific results are shown in Table 1: Table 1. index test results Variables K-S Single variable Non-parametric P Z statistical quantity P value F value P value chi-square value In Figure1, on the premise of the significance of 10%, indexes including cash flow, debt ratio, equity concentration degree, the number of the board of directors and that of the board of supervisors conform to the original hypothesis. Therefore, the above 4 indicators are deleted, and the other 17 indicators are researched further. Figure 5.Financial model 4. MODEL ESTABLISHMENT AND TEST 4.1. Introduction to SVM model SVM model, as a kind of artificial intelligence models, is the classification technology which in the pre-condition of a limited number of indicators obtains the global optimal solution through the local optimal 170

6 solution, and then to solve the overall problem. It involves algorithms like standardized SVM model, LS-SVM model, and LS-SVM growth memory model based on the entropy. In addition, it can be used as the early-warning model to predicate financial distress Model establishment The SVM model is the algorithm dealing with multi-class problems. The classification forms taken by SVM model in handing financial situation can be divided into direct methods and indirect methods. The direct method is realized by assembling the parameters needed by multi-class into an optimization problem, but in the operation process of computers, the computational complexity is high, and the duration is long. There is another kind of algorithm, the indirect method. This method makes calculations through several multi-class SVM classifiers, and realizes sample classification through single to single and single to multiple. where, in the selection of parameters, based on the parameter values which have been calculated, the classification index s accuracy is considered, as shown in Figure 6, and thus the parameter value is determined: c= , t= Figure 6. SVM model parameter selection contour line After that, the SVM classification function is established as follows: n n n f x sng bm ymh x, xni a m Parameter determination In this paper, the parameters c and t are determined by network search, in which c is the penalty parameter in the SVM model, and t is the kernel parameter. The cross validation method is used to determine the parameters of the financial distress prediction model. First of all, quadratic programming is conducted on the data set S, and we get: n L n n N n max 1 b W b w b 2 Where, W is the matrix of the remaining variables after the n-th sample is crossed. Determine the initialization parameter set n, Combine all variables of the study, and then get the parameter model: * * k k, max By running MATLAB software, H groups of data are trained, and we get the test parameters c=1, t= , and then through regression analysis, we get the mean square error M and correlation coefficient S: M , S Then, based on the U, the values are increased by 5%, and two samples, U1 and U2 are obtained. After that, the two samples test result P is subtracted, and the difference obtained is the variation value of the degree of influence. Finally, the average influence vale of indexes is calculated, and the absolute value is used to evaluate the indexes influence degree on the financial condition. Results as shown in Table 2: Table 2. average influence value results Variables Mean order Return on equity Return on assets Sale net margin n 171

7 Current ratio Quick ratio Inventory turnover ratio accounts receivable turnover ratio Fixed asset turnover ratio Net profit growth ratio Net asset grow ratio Sales cash ratio Total assets cash recovery ratio Current asset ratio ]fixed assets ratio Proportion of state-owned shares Proportion of corporate shares Audit mark Figure 2 shows that the influence degree of the rate of return on assets and fixed assets turnover effect is relatively close, up to 0.1. The influence extent of the proportion of corporate shares and audit mark is above As for the growth rate of net asset sales, sales cash ratio, inventory turnover ratio, liquidity ratio, rate of return on net assets, the recovery rate of total assets in cash, and the ratio of fixed assets, their influence degree is more than 0.5. While for accounts receivable turnover ratio, quick ratio, sales net profit margin, net profit growth rate, the proportion of state-owned shares, and the liquidity ratio, their influence extent is below The order of the indexes as to their influence degree on the financial difficulties from big to small is: rate of return on assets, turnover rate of fixed assets, the proportion of corporate shares, audit mark, growth rate of net assets, sales cash ratio, inventory turnover ratio, liquidity ratio, rate of return on net assets, the recovery rate of total assets in cash, the ratio of fixed assets, accounts receivable turnover ratio, quick ratio, sales net profit margin, net profit growth rate, the proportion of state-owned shares, and the liquidity ratio. 5. CONCLUSIONS AND RECOMMENDATIONS The industry has a great influence on our economy. Achievement of reasonable prediction of the industry, in fact, lays the foundation for the whole economic situation. Through the simulation the SVM model in this paper, the order of the indexes as to their influence degree on the financial difficulties from big to small is: rate of return on assets, turnover rate of fixed assets, the proportion of corporate shares, audit mark, growth rate of net assets, sales cash ratio, inventory turnover ratio, liquidity ratio, rate of return on net assets, the recovery rate of total assets in cash, the ratio of fixed assets, accounts receivable turnover ratio, quick ratio, sales net profit margin, net profit growth rate, the proportion of state-owned shares, and the liquidity ratio. At the same time, through verification, it is found that the accuracy of this result is high, and the conclusions are scientific and reasonable. Through the prediction of financial distress, this paper identifies the factors which the listed companies need to develop. In order to promote positive energy of the financial situation of enterprises in China, especially to give play to the central role of China's listed companies in the international market, it is suggested to give full play to listed corporate financial distress prediction, so as to achieve early prevention and timely treatment. The first is to establish the financial distress early-warning system. Before the default, the phenomenon of financial distress will appear. Through the publicity of real and timely financial information of debtors, the financial distress model suitable for our country can be set up. According to the current economic, social transition situation, financial system and capital market, we can establish the prediction model meeting the economic environment of our country. Especially, it is quite significant to improve the index selection. The listed companies are actually enterprises of a wide range, but the index selection should also base on the actual situation in other fields. When appropriate, the non-financial indexes can be increased. In the process of the financial crisis analysis, the method or attitude of one-size-fits-all should be avoided. Instead, it is necessary to have an overall consideration based on the current status. Particular attention should be paid to the differences between different industries. The difference between different industries should be analyzed, and based on these differences to improve and change the model, in order to improve the prediction accuracy and lay the foundation for future measures. References 172

8 Cai L J, Cai L J, Wen-Yong L I, et al. Quality Analysis of Soybean Oil based on Near Infrared Transmission Spectra and Artificial Neural Network Model, Hubei Agricultural Sciences, 2015, 48(5): Huang H, Jiang W, Wang S. Research of Financial Early-warning for Listed Companies Based on SVM,International Conference on Computational Science and Engineering Huang S H, Xia X S. Quality Cost Analysis of Typical Forging Based on SPC and DEA Model, Advanced Materials Research, 2015, 1095: Huang T, Li X Y, Xu M L, et al. [Non-destructive detection research for hollow heart of potato based on semi-transmission hyperspectral imaging and SVM]., Spectroscopy & Spectral Analysis, 2015, 35(1):198. Peng X, Zhang Q. Comprehensive Evaluation and Research on Enterprise Financial Ability Based on Factor Analysis and DEA-Taking A Listed Coal Company as An Example, Coal Economic Research, Shen Q, Dai H, Han P, et al. Analysis and Research of Grenadiers Comprehensive Quality Based on Fuzzy Comprehensive Evaluation, 2015, 356: Song B, Zhu J M, Xu L I. The Research of Enterprise Financial Early Warning Based on Big Data, Journal of Central University of Finance & Economics, Tang L, Zhou C, He L, et al. Research on Air Quality of Beijing-Tianjin-Hebei Region based on SVM and Regression Analysis,International Conference on Education, Economics and Management Research Tang L, Zhou C, He L, et al. Research on Air Quality of Beijing-Tianjin-Hebei Region based on SVM and Regression Analysis,International Conference on Education, Economics and Management Research Wang D. Research of raw water quality assessment oriented to the drinking water treatment based on SVM model, Water Science & Technology Water Supply, 2015, 16(3). Zhang Y, Mei L I, Gao Q Q, et al. Research on forewarning model of seawater quality based on ELMR-SVMR, Journal of Dalian University of Technology, Zhang Y. The Research of the Enterprise Financial Model Based on Information Entropy and Correlation Model Reorganization Theory,International Conference on Intelligent Transportation, Big Data and Smart City. IEEE, 2015: