STOCK PRICE PREDICTION USING STATISTICAL, MACHINE LEARNING AND DEEP LEARNING MODELS

Abstract

Forecasting stock price is a challenging topic for the researchers by the way of statistics or in newer version by the way of Machine Learning and Deep learning. There are researches that prove that the direction of time series for a stock price can be predicted with a good accuracy. Design of this kind of predictive models requires choice of appropriate variables, right models and methods, and tuning the parameters. In this research, the goal is applying different algorithms and approaches for stock price prediction then compare and evaluate them together. This research also aims to apply these models for short-term stock price prediction. The daily stock price data is used, from tsetmc for five biggest Iranian companies active in the Tehran stock Exchange. For each stock, the data is gathered from January 1, 2018 to January 1, 2021 to train, test and evaluate the algorithms. For evaluating performance of all models, three measures are used for performance evaluation including The Mean Absolute Error (MAE), The Mean Squared Error (MSE), and The Root Mean Squared Error (RMSE). Statistical, Machine learning and deep learning approaches that used are include Auto-Regressive Integrated Moving Average (ARIMA), Random Forest (RF), Support Vector Machines (SVM), and Long Short-Term Memory (LSTM) Networks. Results for applying those models are tested and evaluated and the LSTM model has the best performance concerning other approaches and prove outperformance.