Designing a stock valuation model based on capital market indicators using data mining methods

Purpose: The purpose of the study is to design a stock valuation model based on capital market indicators using data mining method. In most of the studies on capital market indicators, the indicators have been separated into structural, cash flow and expectation categories. In the existing research, these indicators and their effects on the price are presented at the same time, or several indicators are mentioned as influencing variables on the price.
In his book entitled “Stock Market Indicators”, William Gordon (2000) defined and classified the main indicators of capital market and studied their effects on the Dow Jones Index in a period of 70 years. He divided the market indicators into those related to the profitability of companies, cash flow indicators and structural market and stated the time series of decisions to buy and sell shares. Yardani (2003) investigated the basis of valuation in different periods by classifying cash and expectation indicators. Van Scheulein (2003), Fernandez (2004) and Ignacio (2003) investigated the effects of cash and discount indices on stock prices and compared different value-forecasting methods. Lina Dajilin and colleagues investigated the factors affecting stock valuation in terms of accounting and market factors. Duran (2005) tested the method of Conslim presented by O’Neill and explained the role of expectation and structural indicators in the valuation process. Roeback (2005, 2002 and 2004) presented the basics of value extraction. Kenny also divided the market indicators into structural, cash and expected indicators.
This research examines the relationship between stock valuation and financial ratios by using the financial information of the companies listed in Tehran Stock Exchange. It is expected that the investigation of this relationship and the clarification of ambiguous cases will increase the level of transparency of the market through increasing the knowledge of investors and better allocations. Therefore, the question the research seeks to answer is ‘What is the effect of financial ratios and market indicators on share price valuation?’
Methodology: Since the current research sought to design a model of capital market indicators using a data mining method and the experts in this field were limited, the statistical population of the research included all the companies in which investors generally have the opportunity to participate in shareholding processes. The data needed to analyze the relationship between the research variables were obtained from the Kodal system. They were collected from audited financial statements, explanatory notes, and the companies admitted to Tehran Stock Exchange. The new Rahvard software was used for data analysis. Finally, the panel data method was used to test the hypotheses. The independent variable in this research was the structural indicators of the capital market, the cash flow indicators of the capital market, and the expected indicators of the capital market. The stock value was the dependent variable. The information related to 157 companies admitted to Tehran Stock Exchange during a period of 11 years from 2012 to 2023 was collected in the form of daily, monthly and quarterly data. The collected data were analyzed according to the purpose of the research, which was to provide a comprehensive model regarding the main indicators of the capital market with a new method.
Findings and Discussion: In this research, structural equation modeling was used to investigate the effect of capital market indicators on the stock value. The results showed that those indicators have significant effects on the stock value; 52.7% of the stock value was explained by the fluctuations of the capital market indicators. In the next step, in order to test the hypothesis of the research, 32 machine learning models were first compared in terms of accuracy in order to predict the stock value. The results showed that the C5 model has higher prediction accuracy than the other machine learning models. Furthermore, in order to increase the accuracy and reduce the error of the C5 model, boosting and feature selection validation techniques were used. The results showed that the superior model (C5) has significantly increased accuracy and a noticeable decrease in error. Also, modeling was done separately for the capital market indicators.
Conclusions and Policy Implications: It was shown that cash indicators are more accurate than the other indicators in predicting the stock value. Based on the results, the following suggestions are made:
a) It is suggested that managers and investors use machine learning models and cash indicators, along with other relevant variables, to predict the value of company shares and make the necessary investment decisions.
b) Potential and actual investors should make investment decisions based on the purchase and sale of shares, managers are suggested to take preventive measures in order to prevent the occurrence of financial risk that may ultimately lead to the financial distress of the company, analysts should consider the information of ordinary investors (non-experts), the stock exchange is recommended to the stock exchange organization for the acceptance and evaluation of companies, and investment companies are advised to develop and expand their activities (through the purchase and merger of other companies). To predict the value of stocks, the C5 decision tree model is recommended.