Enhancing stock market trend reversal prediction using feature-enriched neural networks

According to several previous studies, neural network-based stock price predictors perform better for plunging patterns of stock prices than normal stock price patterns. Focusing on this issue, this study proposes a novel method that uses a neural network-based stock price predictor to predict the upward trend-reversal of the plunging market itself. To achieve more consistent prediction results for plunging patterns, newly designed input features are added to improve the performance of traditionally used neural network-based predictors. The statistics of the prediction scores for past plunging markets and analyzed, and the results are used to predict the upward trend-reversal in the plunging market that occurred during the test period. We demonstrate the superiority of the proposed method through the simulation results of 3-year trading on KOSDAQ, a representative stock market in South Korea.