Prediction of stock market index movements using machine learning methods: An application on the stock markets of developed and developing countries

Stock markets possess a complex structure influenced by numerous factors such as macroeconomic indicators, global developments, and human behaviors. Stock market indices, in particular, stand as crucial indicators reflecting the overall health and direction of these markets. Predicting the movement directions of these indices holds significant importance in terms of making investment decisions and managing risks. Thanks to machine learning methods capable of identifying relationships and trends within historical data, it becomes possible to predict the movement directions of stock market indices. The objective of this study is to compare the performance of machine learning methods in predicting the movement directions of stock market indices in both developed and developing countries, and to determine the best prediction method. In line with this aim, the stock market indices of developed countries were selected as NYSE 100, NIKKEI 225, FTSE 100, CAC 40, DAX 30, FTSE MIB, and TSX; while the stock market indices of developing countries included SSE, BOVESPA, RTS, NIFTY 50, IDX, IPC, and BIST 100. The movement directions of the stock market indices were predicted using decision trees, random forests, k-nearest neighbors, naive Bayes, logistic regression, support vector machines, and artificial neural networks. The analyses employed a daily dataset and input data consisting of technical indicators from January 1, 2012, to December 31, 2021. Based on the results obtained, artificial neural networks are identified as the most effective method during the examined period. Alongside artificial neural networks, logistic regression and support vector machines are recognized as methods accurately predicting the movement directions of all indices with an accuracy rate exceeding 70%. Furthermore, it is observed that the identified best method, artificial neural networks, is not universally applicable across all indices. In this context, it is concluded that the performance of the explored methods varies across countries and indices but does not significantly differ based on countries levels of development. In conclusion, artificial neural networks, logistic regression, and support vector machines are recommended as the most advantageous methods for predicting the movement directions of stock market indices.