Data Mining of Financial Statement for Predicting Common Stock Prices for SET50: A Case Study of Thailand

Abstract

                   The objective of this research aimed to apply data mining techniques to investigate the market price of common stock from SET50 in the Stock Exchange of Thailand in order to forecast the market price for investment. The population consist of fifty companies.
We collected the financial statements (form 56-1) all fifty companies in SET50 since 2016 to 2020. The factors or variables that we concerned are composed of seven variables: revenue, net profit, profit per shareholder’s equity, shareholder’s equity, number of common shares, registered capital, and dividend. We used three various data mining techniques including Generalized Regression Model, Decision Tree, and Support Vector Machine by using RapidMiner Go as a tool. The research results indicated that all seven factors or variables that we concerned significantly affect the market price. However, each data mining technique forecast different levels. We considered the values of RMSE (Root Mean Squared Error) importantly; less RMSE, less error. We found that RMSE from lowest to highest value are Decision Tree (28.19), Support Vector Machine (35.05), and Generalized Regression Model (40.78) respectively. It meant that Decision Tree is the best for forecasting in this research. Decision Tree showed the results in the form of decision tree visualization. Support Vector Machine and Generalized Regression Model showed the result in term of weight and coefficient consecutively.