SEPTEMBER 2011 VOL 3, NO 5. Mohammad Hasn Gholizadeh 14 Mohsen Mohammad Nourbakhsh Langroudi 15 Ali Bahmani 16 Behnam Shadi Dizaji 17

Transcription

1 INTERDISCIPLINARY JOURNAL OF CONTEMPORARY RESEARCH IN BUSINESS Corporate financial distress prediction using artificial neural networks and Using micro-level financial indicators VOL 3, NO 5 Mohammad Hasn Gholizadeh 14 Mohsen Mohammad Nourbakhsh Langroudi 15 Ali Bahmani 16 Behnam Shadi Dizaji 17 Abstract: many companies are going in financial distress In today's complex economic environment; Predict the probability of bankruptcy will help them to decide of appropriate internal and external stakeholders organizations. The main goal of this research has been to predict corporate financial distress Using artificial neural networks and internal factors affecting on company, Using financial micro variables. The study population was consisted of 444 companies listed in Tehran Stock Exchange for the years and The sample includes 144 company. MATLAB software to verify the hypothesis of the study and Neural Network Toolbox and using it in the code, is a neural network design and analysis. The results of this study indicate that the The use of micro-economics can play an important role to play in financial distress or further fractures. Keywords: financial distress, neural networks, financial indice Introduction Growth and development of technology and management decisions are difficult, especially after the 1950 And institutions seeking to increase trade, economic relations and trade are significantly more complex; Economic policies of governments in the developed countries, this complexity is added. In such circumstances the role of financial management in the strategic goals of the organization is excellent. In such an environment, many companies are in financial distress; The predicted probability of bankruptcy and its external stakeholders in decision-making, The company's customers to banks and credit institutions in takes place, has a significant impact And other stakeholders from the local managers of the company, but they are Also predicted that the crisis can take decisions as appropriate to deal with it. What was described as being of strategic financial decisions, Financial managers tend to use modern methods of analysis and forecasts are And the purpose of computers and sophisticated mathematical models are used to advance their goals. The most advanced models, "Artificial Neural Networks" that benchmarking and simulation of natural human neural networks are designed. Artificial neural network model, mathematical aspects of the process of biological neural networks represent the body. This model uses the computational speed of computers, complex relationships between the variables analyzed and then used to predict future values 14 -Assistant professor,department of management, Rasht Branch, quilan University, Rasht, iran 15 - Financial Assistance and Support of Regional Power, Part-time faculty of Islamic Azad University, Rasht Branch,iran 16 - Budget experts and Gilan Regional Electricity MA student - Finance, Islamic Azad University of Rasht 17 -Masters, Department of Business Management, ajabshir Branch, peyamnour University, ajabshir, Iran COPY RIGHT 2011 Institute of Interdisciplinary Business Research 595

2 are(komijani,1385,p11) The first study on the use of artificial neural networks began in But the use of financial models is very recent. The main application of artificial neural network models in economics, Forecasting economic variables, including variables related to the financial markets, monetary and macroeconomic variables such as Predict stock prices, exchange rates, oil prices, inflation and growth. Another application of these models in economics, "Classification of economic units" is. Of neural networks to predict bankruptcy in general economic units are used.(zhaf,1380,pp ;gadimi,1381,pp ) Artificial Neural Network: Structure (network) consisting of units connected together (processing units), each of which is the entrance and exit. Financial distress: In which case the power loss due to company profitability Likely inability to repay debt principal and interest has increased the company(heydari,1389,p65) Bankruptcy: In accordance with Article 412 of the trade, business or business bankruptcy at the end of his duties for the funds that are(heydari,1389,p213) Current ratio: This ratio indicates the company's ability to repay short term debt and is now(david,1386,p317) Debt ratio: Percent of total funds provided by creditors. (David,1386,p319) The total return on assets: Profit after tax per of an asset(david,1386,p320) Oxford Dictionary of the Hong-Far Distress means failure and frustration, pain, sadness, lack of monetary resources are. Various definitions have been provided for in the financial literature. Altman tells Financial ruin occurs when a company fails to pay its debt not, Therefore, of the business remains open( altman,1968,pp ). Gordon is one of the first studies of its bankruptcy case it is defined as loss of business profits Likely inability to pay interest and principal debt increases(gordon,1998,pp80-86) The reason or reasons for the bankruptcy and financial problems is never easy Dun&Bradstreet(Dun&Bradstreet,1998,pp21-41) Fractures around the main financial and economic problems are financial. But unlike the Gitman(1986) Believes first and foremost because it is the mismanagement of the bankruptcy. Whitaker(1999) Financial distress is a condition in which the cash flows of the company's total interest costs related to long-term debt is less. From an economic point of view, the fatigue can be financial loss as the company that this company has suffered a failure. In this case the company's rate of return is less than the cost of capital rate. Another case of financial distress that occurs when The company failed to comply with one or more provisions relating to the debt contract Such as keeping the current ratio or the ratio of equity to total assets is not under contract. This technique is called the default mode. Other scenarios include the case of financial distress Company to repay debt principal and interest cash flows be insufficient and the company's equity is negative(weston&compland,1992) In this paper, using artificial neural networks and the use of financial indicators at the micro level The company has been predicting financial distress. The internal factors affecting financial distress and then to identify the financial needs of these factors has been extracted; With this in mind as an input to the artificial neural network, ultimately to predict financial distress has been paid. Background of financial ratios and financial distress prediction The analysis compared to 300 years before Christ returns, oglidoss when this type of analysis used in studies(horrigan,1968,pp ) The scientific use of the funds were launched in This was in the banks for lending, as financial companies were seeking Using this ratio in the 1890s and intensified in the course of this variety was introduced. According Horigan, one of the oldest of the major financial impact on the financial analysis is done, than it is Current Ratio. The ratio of current assets over current liabilities of the division is obtained, which indicates the debt is short term. COPY RIGHT 2011 Institute of Interdisciplinary Business Research 596

3 Beaver states that the ratio analysis in the early 1900s began with the development of Current Ratio.( Beaver,1966,pp1-25) In 1919, Dupont Compant., the triangular model proposed. The triangle in the top return on investment (profit divided by total assets) And two at the base of the profit margin (profits divided by sales) And asset turnover (sales divided by total assets) were used. However, this approach will provide a framework for the development of different proportions, But these proportions until 1960 was used as a tool to predict bankruptcy. In the first decade of 1920, various financial ratios was created by commercial entities. In 1930 with the Securities and Exchange Commission in the United States, intensified development and use of financial ratios. These studies showed that the proportion of bankrupt financial companies, Significantly with the ratio of non-financial company is bankrupt.( Altman,1968,pp ) Winakor & Raymond, in 1930 and 1935 trends in financial ratios of bankrupt companies were analyzed. The 10-year trend of financial ratios with these companies21 use of financial ratio analysis and concluded that they Working capital to total assets ratio, the most appropriate as an indicator of the company is bankrupt. Major flaw of this study, the lack of a control group of nonbankrupt firms were(horrigan,1968,pp ) Beaver was the first company of statistical techniques for predicting bankruptcy of financial ratios used(beaver,1966,pp1-25) And the first multi-point analysis can be used to predict bankruptcy, is Edward Altman. The researchers found that some of the financial ratio approach to corporate bankruptcy, have changed significantly. The initial model was introduced in 1968 and Altman Z-Score is well known, was presented with a study of manufacturing companies. Altman of the 22 financial ratios, five of them chose to use this model. In 1993, he revised his original model and this time the four financial ratios used. Altman Z-Score model is revised, the coefficients are as follows. Z=6.56X X2+6.72X X4 Where: X1= Total assets / working capital X2= Total assets / retained earnings X3= EBIT/ Total assets X4= Total debt / equity Z= Index In this model, companies are classified according to the number obtained for Z took this way: Companies that had the potential to bankrupt the vote, less than 1 / 1 found; Bankrupt companies that were not expected to score more than 2 / 6 were And companies that score between 1 / 1 and 2 / 6 were obtained in the unpredictable.( Altman,1968,pp ) Altman was a theory that models such as MDA, can be used to distinguish between companies that are being used to bankrupt.( Altman,2000,pp ) Application of Artificial Neural Networks in Finance Neural networks in a wide range of issues such as aerospace systems, automated home appliances, banking, electronics, industry, defense, medical, audio and video, robots, telecommunications, and transportation systems Be utilized. What neural networks will become more popular in the future High speed computers and computational algorithms that learn faster than neural networks used in industrial problems requiring high computing. Neural networks can be best used to solve problems that Are largely unstructured and require a form of pattern recognition. It is also possible to solve these problems, the data are incomplete and distorted. Financial District, an area that is active in deal with routine matters, such issues are also COPY RIGHT 2011 Institute of Interdisciplinary Business Research 597

4 involved. In some applications of neural networks to solve problems that corporate financial managers, Financial institutions and professional investors are faced with it, is mentioned. Financial simulation: the operations of each company's financial structure, the environment is very dynamic and complex form. Although the functions of financial management A number of conceptual and operational tasks easier and to be divided into smaller, However, interaction between these smaller tasks is very complex. Neural network can be modeled in different parts of the company's financial environment. Such models can (1) is a specific company. (2) Due to changes in the company's financial structure over time, is dynamic, and (3) Reflection of the relations between modeling, and other financial and non financial corporate sector and the external environment. For example, the artificial neural networks can be used to simulate the behavior of a consumer credit company When changes in economic conditions of use. The data can consist of data and economic data for each customer is And output behavior can be expected by customers on the purchase or payment. Training data, based on actual customer behavior in the past. Planning for such a system in suspicious claims costs, Periodically increase and decrease in accounts receivable and assess the conditions and restrictions are intended for individual customers, it will be useful. The simulation using neural networks, can also be used in many other parts of the financial company; Such as cash management, investment evaluation, risk management (insurance) and Personnel,asset management, credit risk, exchange rate and projected costs based on financial data and other Kind of. Prediction: Other duties in financial management, financial forecasting is. These tasks using the computer software can be done more effectively than the use of neural networks. Especially those tasks that involve complex numerical calculations of the models are specified. However, financial analysts are always on the behavior of some investors are concerned. Information about the company's investors that there can not be analyzed separately But affected the entire information about the company and in fact there is a comprehensive, they should be. The neural network to emulate and simulate the response of investors Changes in financial position or policies of the Company occurs, the train. In other words, using the actual behavior of investors as a training model, networks can be devised to simulate the reaction of investors. For example, in response to changes in the dividend policies, accounting methods, the company reported earnings, capital structure and to simulate. Studies have been done in the past, changes in stock price as a criterion for measuring investor reaction is emphasized. Investors may react to Other than the shares they buy or sell. Neural networks can function as a financial analyst in forecast Investors react to changes in financial policy to improve. Evaluation of neural networks can be used in the evaluation of companies looking for acquisitions. This information is used by financial companies. Neural network training process, including the vectors that contain financial information about the company, it is. This process also includes an output that is determined The estimated value of the company, which is comparable to that by human experts. The neural network can be desirable to select companies for acquisitions, based on criteria other than a simple evaluation of the training; The only criteria for a human expert is understandable, I like the internal or personal preferences. In other words, this neural network can be trained through A specialist in human thinking and insight without relying on definable rules, or programmable logic, to modeling. Such neural networks have great advantages include(hawley et al,1990,pp28-46) Of such a system can be used to monitor Many companies and cull those that are less valuable Or for the acquisition conditions are used. Thus, given that the final decision The only conditions are that companies are examined, The time COPY RIGHT 2011 Institute of Interdisciplinary Business Research 598

5 savings are enormous. Because of such a system that relies on rules or knowledge base is not scheduled, The assessment techniques can easily decide to take a pattern. The system can automatically, so that changes The analytical methods and selection criteria are the decision maker over time, will adapt. The importance of research Since companies can provide needed financial resources From internal sources such as retained earnings and borrowing to use or refer to outside sources; Or the issuance of debt and equity to proceed. Which method of financing for companies that are a priority For managers, shareholders, creditors and future beneficiaries is important. This is important due to the effect that the method of financing the company's value(najati asl,1389,p4) In today's complex economic environment, many companies are in financial distress; Predict the probability of bankruptcy and help them to Decisions appropriate internal and external stakeholders will organizations. The predicted probability of financial distress can be prevented from wasting the national capital. This prediction can also select the method of financing companies to help. This study involved the use of capital markets, financial institutions and banks The credit rating companies, all companies active in the economy and government support or the amount of taxes to be imposed. The teachers and students and other researchers can benefit from the findings for future research. And demonstrated the importance of this issue is the economic decisions In this research can be done to get better results in the macro-economic decision making was used. Research Questions The main research question is posed as follows: Neural networks can be retained by the financial companies in Tehran Stock Exchange predicted? Research hypothesis Research hypothesis is expressed as: Accurately predict failure of financial firms Using the wisdom of the macro-economic financial ratio increases. The population and sampling Theories based on research, the research must have clear and defined limits to the researcher At all stages of the research surrounding it is sufficient and reliable. The results of the study can be generalized to the whole community(azar and Moemeni,1377) The present study is no exception and is almost certain. The study of corporate financial distress With the help of artificial neural networks and application of micro variables is expected. Spatial domain of companies listed on the Stock Exchange Tehran. Financial information on companies listed in Tehran Stock Exchange has features that are intended for research, From early 1384 until 1386 were studied. Given the recent global financial crisis on micro finance and macro-economic variables has been impressive, That were intended to prevent An unexpected effect on the results of recent research into the period before of that limit. The population consisted of 444 companies listed in Tehran Stock Exchange For the years and involved a sample of 144 companies considering the feature please consider below: 1- By the end of 1383 the company name in the list of companies listed on the exchange is accepted. COPY RIGHT 2011 Institute of Interdisciplinary Business Research 599

6 2- In terms of company information is available. 3- Holding companies and is not part of financial intermediation. 4- The study is divided into two experimental and control groups. Test group, 72 companies had accumulated losses have to be in 1386 And the absolute value of accumulated capital losses that are greater than or equal to 50%Was. With an equal number of controls tested were selected. First, the requirement that companies that had accumulated losses in 1386 And ratio that 50% smaller than the company had been selected And the rest of the companies that Retained earnings and accumulated profits to investors than it is among the smallest of companies. Data analysis methods Required data mining research through documents, observation and audit of financial statements has been collected, The use of published reports about the market By the Securities and Exchange databases and data is collected. Research to verify the hypothesis of MATLAB software is used. Using Neural Network Toolbox in the design and coding of neural network analysis has been. Test the research hypothesis To test the research hypotheses, a multi-layer perceptron neural network was designed. This model only considering the financial ratios as independent variables perform. The results of the model to accept or reject the hypothesis is shown below. The aim of the network, expected to be the best estimate, the lowest error is output to the original values. Ss and the matrix nett network and net_final are defined by default. Using the model of financial companies, were processed And results of the estimation of the Chart (1) (matrix of error = SS) are shown. Chart (1) error matrix ss The Classification error Time required for training number of neurons COPY RIGHT 2011 Institute of Interdisciplinary Business Research 600

7 INTERDISCIPLINARY JOURNAL OF CONTEMPORARY RESEARCH IN BUSINESS Results from the model using financial ratios of companies in the graphic (2) are: Chart (2) apply to financial firms VOL 3, NO 5 As can be seen, the networks and the output of both the "tansig" and the error function "mse" is intended. Repeat the calculation in the network (epoch) 24 have been stopped. Minimum error value is equal to and the amount of weight reduction (gradian) each time repeating the calculation error is estimated to equal Chart (3) Confusion matrix Confusion matrix, by definition, offer the power grid companies are forecasting. Accordingly, in this matrix indicate that it is ij How many networks (and what percentage) of the outputs envisaged in the original situation (bankruptcy, and health) has to offer. Thus the [1 1] and [2 2] demonstrated the accuracy of the prediction network and by [1, 2] and [1, 2] suggests that the number of forecast errors has been done. The final results including the final weights of the network processing, in [3 3] COPY RIGHT 2011 Institute of Interdisciplinary Business Research 601

8 is the confusion matrix. All of the above matrix, considering only the network of financial companies, 0.75% future of the company is able to accurately predict. Chart (4)Training Network Graph of weight training, network with error correction, which is seen as Chart (5) Error Condition Error status data to train test, validation error during training of zero epoch of data validation has been added. So stop the network training epoch 24 And circle in the epoch of zero means the weights are obtained in this epoch. Conclusion In this study was to predict accurately whether the financial failure of companies with Financial ratios made by using artificial neural networks will increase? Therefore, the current ratio, return on total assets and liabilities of companies in financial distress as factors were considered. Type of financial companies as the only independent variable and the second time into account all the variables implementation wisdom. The results of both runs were compared to test the research hypothesis. To achieve the above objectives, design and testing hypotheses were the COPY RIGHT 2011 Institute of Interdisciplinary Business Research 602

9 following: Accurately predict failure of financial firms Using the wisdom of the macro-economic financial ratio increases. Information necessary to test the research hypothesis of the text, explanatory notes accompanying the financial statements and corporate, through intensive Tablets and comprehensive database of companies and Web Site Development Bank and Islamic studies, Islamic Republic of Iran were collected. To test the hypothesis of the study was to use artificial neural network models. Model than a purely financial ratio company as an independent variable with Type And the second time into account all the variables implementation wisdom. The results of both runs were compared to test the research hypothesis. The economic wisdom of entering the network, Neural network accurately predicted a significantly increased The research hypothesis was proved correct and the The use of microeconomics can play an important role to play in financial distress or further fractures. The end point should be reminded that predict corporate financial distress can only be The first step in getting the company to bankruptcy and the consequences it did prevent the But what is the reaction of the senior managers Strategic plans and decisions for the organization. The proposals will be presented as follows: 1- The artificial neural network model for service companies like banks and trading companies 2- Comparison between models using artificial neural networks and traditional statistical models like regression 3- Topology employing other artificial neural networks 4- Using genetic algorithms in predicting corporate financial distress COPY RIGHT 2011 Institute of Interdisciplinary Business Research 603

10 INTERDISCIPLINARY JOURNAL OF CONTEMPORARY RESEARCH IN BUSINESS VOL 3, NO 5 References Altman, E. I. (1968, September). Financial ratios, discriminant analysis and the Prediction of corporate bankruptcy.the Journal of Finance, 23(4), Azar Adel,Momeni Mansour,(1377),Statistics and its application in management. Volume 1 and 2, Tehran, Published SAMT. David, Fred. R.., (1386), strategic management, translation Parsayyan, Ali. Arabi, Mohammed. Cultural Research Bureau, Tehran Dun & Bradstreet(1 998). Bankruptcy Insolvency Accountiong Practice and Procedure. Wiley, Hawley, D., Johnson, J. & Riana, y. (1990.). Artificial Neural systems: A New Tool for Financial Decision Making, in: Trippi Zhf, Marjan, (1380), "Neural networks and financial markets", Exchange, No. 30 Gadimi mohammadreza, (1381), "Modeling and forecasting economic growth in Iran using artificial neural networks", First Printing, Tehran's Allameh Tabatabai University. Komyjany, Akbar, Saeadatfar, Javad, (1385), Application of neural network models in predicting bankruptcy of financial company stocks. Two Journal - External Economic Studies, Year III, No. VI, autumn and winter Zahra Najati asl, (1389),theory test, "Building static" and "hierarchical" in Iran - Accounting Thesis,azzahra University. Altman, E. I. (2000). Predicting Financial Distress of Companies: Revisiting the Z- Score and Zeta Models. New Tork University. Altman, E. I., Haldeman, R.G.,& Narayanan, p. (1977). Zeta analysis: A new model to identify bankruptcy risk of corporations. Journal of Banking and Finance, 1,29-54 Weston J.Fred,Copeland Thomas E. (1992, Februry). Managerial Finance. Dryden Press,The edition. Beaver, W. H. (19PP). Financial ratios as predictors of failure. Journal of Accounting Research, 4, Empirical Research in Accounting: Selected Studies, (Supplement), Gordon, M.J. (1971). Towards a theory of Financial distress. The Journal of Finance, 26, COPY RIGHT 2011 Institute of Interdisciplinary Business Research 604

11 Whitaker. Richard (1999).The Early Stage of Financial (Distress. Journal of Economics andfinance, 23 (2), Horrigan, J. o. 1968, autumn). A short history of Financial ratio analysis. The Accounting Review, COPY RIGHT 2011 Institute of Interdisciplinary Business Research 605