Stock Market Prediction and Portfolio Optimization

Abstract

The highly volatile nature of the stock market has made stock price prediction as challenging as weather forecasting. Consequently, as a hint of this dread, people don’t invest in the stock market. In this paper, we have discussed hybrid networks and a stacked LSTM network for stock price prediction. Additionally, it also focuses on portfolio optimization done using six different techniques, which focuses on creating best performing portfolios categorized on the basis of sectors. One hybrid neural network consists of 1D-Convolutional layers and LSTM layers, and the other is a combination of GRU and LSTM layers. The stock prices of SBI, Indian Bank, Bank of India are predicted using stacked LSTM and Hybrid Neural Networks and compared using the sliding window of time steps with variable width. The neural networks predict the following day’s closing price using a variable sliding window. The RMSE, MSE, and MAE are used to evaluate the efficiency of these neural networks. The hybrid network is proving to be more competent in various situations.