Bridging Bollinger Bands and Machine Learning in SET100 Stock Trading Strategies

Effective stock selection plays a pivotal role in investment success, given the inherent risks and fluctuations tied to various factors. This paper aims to explore the experimental efficacy of combining Bollinger Bands (BB) and an Artificial Neural Network (ANN) in the context of stock transactions. Utilizing historical stock prices from diverse sectors within the SET100 index, the proposed hybrid predictive model, BB-ANN, was tested using ten selected features to forecast percentage profit and loss. Assessment metrics included Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and the coefficient of determination (R-squared). The study compares the best and worst ANN models with a baseline model using only Bollinger Bands without ANN. Results indicate that the top-performing models outperform the baseline across evaluation measures, emphasizing the substantial impact of ANN model selection on stock trading performance. However, it is noteworthy that the worst-performing ANN model falls short of the baseline model's performance.