Comparison of deep learning model for stock trading after peak pandemic: case study SET50 stocks

Wealth creation through stock market investment remains an enduring challenge. In recent times, researchers have been leveraging machine learning (ML) to generate stock trading signals, based on its potential to uncover patterns that may elude human analysis. However, algorithmic performance is contingent upon varying market conditions. The after peak pandemic era is riddled with uncertainties like bank runs, inflation, and geopolitical conflicts, potentially escalating market volatility. Deep learning models, renowned for detecting complex patterns, are examined in this paper. Three such models—ANN, LSTM, and MLP—are evaluated for their efficacy in stock trading during this turbulent period. The proposed LSTM model, in particular, has demonstrated the capacity to generate average positive returns, outperforming the traditional buy and hold strategy. Furthermore, all proposed three deep learning models show enhanced performance with locally-focused stocks, suggesting a nuanced strategy for navigating the complexities of the current financial landscape.