Performance and Usage of Data Mining Techniques for Predictive Modeling: A Study of selected Stock Markets
Loading...
Date
2019
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Punjab Agricultural University, Ludhiana
Abstract
Data mining techniques can effectively deal with the nonlinearity of the stock market and allow search for hidden patterns, in large volumes of data. The present study is aimed at evaluating the performance of data mining techniques across the global stock indices and also to evaluate the performance of data mining techniques in the selected Indian stock market, i.e., National Stock Exchange. Further, an attempt has been made to explore the usage of data mining techniques among the investors and stakeholders. Secondary data based on daily values of stock indices of three developed markets (United States of America, United Kingdom and Japan) and four emerging markets (China, Brazil, India and South Africa) were collected i.e. Dow Jones Industrial Average (DJIA), FTSE 100, Nikkei 225, SSE 50, iBovespa, Nifty 50, JALSH for the period of 12 years ranging from 1st April, 2005 to 31st March, 2017. Further, secondary data related to technical and fundamental variables were collected for the constituents of CNX 500 of National Stock Exchange with time period ranging from 1st April, 1998 to 31st March, 2016. Further, primary data was collected from 167 retail investors and other stakeholders with the help of pre structured non-disguised questionnaire. Results based on the analysis of secondary data reveal that there was a significant difference in performance of data mining techniques in terms of Hit Ratio, Returns, Mean Absolute Error and Root Mean Square Error across different stock indices. ANN models based on data of all selected indices performed better in terms of Hit Ratio and annual returns as compared to SVM models based on data of all indices. Further, results reveal that both data mining techniques i.e. ANN and SVM have an ability to differentiate between high performance and low performance stocks. Performance of stocks improved with the usage of data mining techniques as compared to buy-and-hold strategy. Results based on primary data reveal that retail investors face various challenges in using prediction techniques for investment decision making. Generally, there is a lack of awareness and knowledge regarding prediction techniques among the investors.
Description
Keywords
null