Stock Price Prediction Using Corporation Network and LSTM

Abstract

The problem of stock price prediction is addressed in this work by incorporating additional stock-related factors and using them to model relations between stocks. We have built a corporation network that displays the relation between stocks based on common shareholders and their shareholding ratio. The network is constructed by including all involved corporations based on investment facts from the real market. In this work, we have used a deep learning-based model, long short-term memory (LSTM) for the prediction of stock prices. We have considered node embedding methods that can store the properties of the nodes in the network, and use this information to make the model more accurate. The results produced by an initial and a revised LSTM model are compared, which have achieved a minimum mean average percentage error (MAPE) value of 4.121% for the initial LSTM model, and 1.788% for the revised LSTM model. © 2022 IEEE.