Utilizing Deep Learning architectures to predict stock price movements has been a research focus since the introduction of artificial intelligence, in search of a way to make better, well-informed, and rational investment decisions. Numerous models were used with the price history, from Deep Neural Networks (DNN), Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), and Long Short-Term Memory (LTSM), to hybrid models like the CNN-LSTM model, all have a fairly decent prediction accuracy. However, to achieve an even better model performance and prediction accuracy, one important element that plays a big role in price movements in the stock market requires more attention, which is information asymmetry (Noah, 2016).

This study aims to fill a research gap that currently exists, by comparing the model performance and prediction accuracy between predicting stock price movement using LSTM solely with stock price history and with the incorporation of insider transaction data.

Insiders are regulated by the SEC and must abide by certain rules. They must report to the SEC by filing an SEC form within two business days after they make a transaction. Every filing is publicly accessible on the SEC’s EDGAR database and is tracked by various well-known financial news pages such as Forbes, Finviz, etc. Through this comparative study, a conclusion can be reached on whether insider transaction data plays a role in stock price prediction and whether it can induce a positive effect on the model performance and accuracy.