A comparative analysis of stock price prediction techniques

Abstract

Stock index futures are difficult to forecast due to unpredictable stock market conditions. Despite this fact, efforts are being made over the years to create an efficient forecasting tool. The current advancements in technology like machine learning and artificial intelligence have improved our performance with non-linear estimation. Here, the non-linear, RNN (Recurrence Neural Networks) -based stock index prediction model has been compared to the linear, technical indicator-based stock index prediction model. In our empirical research, ten years of day-wise close price data of Tata Consultancy Services Ltd has been used. The study explores two separate methods for predicting stock prices, each coming from a distinct specialty: In the linear model, MA (moving average) and EMA (exponential moving average) model, and the nonlinear model LSTM (long short-term memory) approach has been used. The analysis shows that when it comes to stock price prediction, the exponential moving average outperforms the LSTM.