Leveraging machine learning techniques for stock price prediction of selected VN30 index companies and portfolio optimization: A case study o f Vietnam stock market

Abstract

Accurately predicting stock market returns is a very difficult undertaking because of how unpredictable and complicated financial stock markets are. However, programmed prediction approaches have become a potential strategy for achieving more accurate stock value projections because to developments in artificial intelligence and increasing computer power. This study focuses on predicting stock closing prices for four prominent companies, namely The Corporation for Financing and Promoting Technology (FPT), Hoa Phat Group Joint Stock Company (HPG), Vietnam National Petroleum Group (PLX), and Vietnam Dairy Products Joint Stock Company (VNM), listed on the VN30 Index. In this study, a variety of methods including Support Vector Regression, Artificial Neural Network, and Long Short Term Memory have been used to anticipate stock closing values for future time periods ranging from 1, 5, 10, 15 and 30 days. Technical indicators that have been calculated from the stock's Open, Low prices, High, and Close prices and are used as the input features for prediction. Using common strategic metrics, such as rRMSE, MAE, and MAPE, the models' performance is assessed. Low values of these indicators are displayed by the models, demonstrating how well they are able to forecast closing stock prices. In addition to price prediction, the Mean Variance Portfolio Optimization technique is employed to optimize the Sharpe ratio. The Sharpe ratio calculates an investing strategy's risk-adjusted returns, so optimize the Sharpe ratio is to optimize the risk-adjusted returns. And in order to optimize risk-adjusted returns, the best distribution of assets among the four businesses is identified via Mean-Variance Porfolio Optimization.