EVALUATING THE INFLUENCE OF TECHNICAL INDICATORS AND TEMPORAL FEATURES FOR STOCK MARKET PREDICTION USING MACHINE LEARNING

Predicting the stock market is an ill-defined and difficult task because of the time dependency, dynamic and non-linear characteristics of financial data. In this paper, the authors explore the impact of more technical indicators and temporal characteristics on the predictive power of machine learning models for stock market prediction. Preprocessing and enrichment of historical stock price data were carried out using technical indicators that are widely used in the finance industry, such as Relative Strength Index (RSI), Moving Average Convergence Divergence (MACD), Exponential Moving Average (EMA), Simple Moving Average (SMA), Average True Range (ATR), and momentum-based indicators. Further, time-based information like lag values, rolling statistics, trading day, and monthly trends were added to showcase sequential market behaviour. Performance of the various combinations of features was evaluated based on classification and regression performance metrics like accuracy, precision, recall, F1-score, Mean Absolute Error (MAE), Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE). Results from the experiments show that the incorporation of technical indicators with the temporal features can greatly improve prediction performance over the use of raw price data only. The proposed feature engineering strategy resulted in an accuracy and F1-score of 96.82% and 96.40%, respectively, which are very good, and an RMSE of 2.08, an AUC of 0.982, and a strong performance achieving good predictive capability. The results validate that judicious design of financial and temporal parameters significantly enhances machine learning-based stock market forecasting and can aid in building intelligent financial decision support systems.