Abstract This study investigates the predictive power of informed traders’ activities in stock price movements by employing neural networks. Specifically, we examine whether informed investors’ trading activities can predict drastic changes in stock prices in the subsequent 5-day period. Our empirical results show that the probability of the model being correct can be as high as 74%. In addition, the simulated trading strategies based on our trained model lead to significantly positive risk-adjusted returns and show strong performance measures. Overall, we find that informed traders’ activities contain informational content and may provide actual investors with information that is useful for stock price prediction.
Paper
Full text
Predicting stock prices based on informed traders’ activities using deep neural networks
Semantic Scholar · Computer Science · 2021
Abstract
Abstract This study investigates the predictive power of informed traders’ activities in stock price movements by employing neural networks. Specifically, we examine whether informed investors’ trading activities can predict drastic changes in stock prices in the subsequent 5-day period. Our empirical results show that the probability of the model being correct can be as high as 74%. In addition, the simulated trading strategies based on our trained model lead to significantly positive risk-adjusted returns and show strong performance measures. Overall, we find that informed traders’ activities contain informational content and may provide actual investors with information that is useful for stock price prediction.