Application of deep learning models on time series analysis based stock price prediction

Abstract

The stock market has been a promising investment market, full of potential, and yet a volatile market that requires the analytical skills of traders and investors. The nature of the stock price time series is always changing, and nonlinear, and hence it is considered to be unpredictable. Nevertheless, thanks to the application of machine learning algorithms and robust quantitative forecasting models, predicting stock prices is not as difficult as before. This paper aims to fulfill two main goals. Firstly, in the context that more and more young people are entering the stock market but do not have much experience and analytical skills that they more often than not hardly get the desired profits, the study proposes an approach of a deep learning model for daily stock price forecasting, which is Long-Short Term Memory (LSTM). This approach is widely considered as one of the most accurate forecasting approaches for dynamics time series. Secondly, the model's performance metrics shall be designed to evaluate the forecasting accuracy of each deep learning model and from there help the investor to make precise investing decisions. This paper uses three metrics: Root Mean Squared Error (RMSE), Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE). The forecast model is used to generate the stock closing price of Vietnam Dairy Products Joint Stock Company (VNM), Phu Nhuan Jewelry Joint Stock Company (PNJ), Petrovietnam Gas Joint Stock Company (GAS), and Asia Commercial Bank (ACB), which are four of the most popular companies in Vietnam. The results of the study show that the LSTM model is capable of generating accurate stock price forecasts.