A comparative analysis of the performance of machine learning models for predicting stock prices from the years 2012 to 2022: Evidence from the ASEAN 5 stock market indices

• Main Authors: Dhanani, Sameer D., Escaño, Hugh Leon B., Lim, Jasmine L., Pascual, Isaiah Franz Dominique L.
• Format: text
• Language: English
• Published: Animo Repository 2023
• Subjects: Stocks—Prices—Southeast Asia; Stock index futures—Southeast Asia; Finance; Finance and Financial Management
• Online Access: https://animorepository.dlsu.edu.ph/etdb_finman/83; https://animorepository.dlsu.edu.ph/context/etdb_finman/article/1063/viewcontent/A_Comparative_Analysis_of_the_Performance_of_Machine_Learning_Mod.pdf
• Institution: De La Salle University

The vitality of an economy is often reflected in the fluctuations of its stock market, emphasizing the necessity of studying and anticipating market dynamics. To assist some analysts, machine learning models have gained prominence in forecasting stock market behavior over the years. A few of these models are Artificial Neural Networks, Support Vector Machines, and Random Forest. The study focused on the ASEAN 5 indices and FTSE ASEAN All-Share Index as a benchmark for considering their benefits and limitations. The study forecasted the directional movements of the ASEAN 5 indices using these machine learning models and evaluated their performance. Historical price data from 2012 to 2021 were used to predict stock prices for 2017 and 2022 and were compared to actual prices to assess accuracy. The study also examined volatility and used metrics like the Sharpe Ratio, Jensen's Alpha, and Beta coefficients to assess model performance. Moreover, the study compared the performance of the models between 2017 and 2022. When assessed using the metrics, SVM exhibited the most consistency in accurately predicting the ASEAN 5 index prices for 2017 and 2022. The reason is that SVM is generally designed for smaller datasets and is moderately complex to optimally predict this study’s dataset prices. SVM’s performance was followed by the ANN then the RF model, underscoring the varying predictive power of these models when evaluated using financial metrics. The study recognized the constraints of historical data, the ever-changing dynamics of the stock market, and the possibility of external factors affecting predictions. Hence, the study emphasizes the importance of exercising caution and prudence when interpreting the results and the need to adopt a comprehensive approach to making investment decisions.