Stock Price Movement Prediction using Attention-Based Neural Network Framework

Kartik Goyal, Nitin Bansal, Soumyabrata Kundu, Ayan Kundu, Nitish Jain

Abstract: There is a lot of scientific work going on NLP trying to predict the impact of news on a stock price, much of this uses basic features (such as bags-of-words, named entities etc. ), but fails to capture structured entity-relation, and hence lacks accuracy.1. Encoding the information like daily events, meta-stock information and stocks 50 days moving average using LSTM.2. Employing attention mechanism to rate the relevancy of all events for each stock.3. Using non-linear neural network on the weighted events to predict the stock movement. The model achieved an accuracy of around 72 % on test set.

Keywords: Stock Price Movement, Neural Network, Attention Mechanism, NLP