Stock Market Prediction: A Time Series Analysis 

Abstract

Stock market prediction is one of the most critical tasks in the area of computation due to its high erratic nature. There are many factors involved in the stock market that are responsible for the rapid vicissitudes of the stock price. Continual unsettlement in the stock market is the primary reason for the investors to sell out their stock at a deceitful time that leads to failure to gain profits. Due to the continual unsettlement of stock price, it is not possible to predict stock market movement with full accuracy but by predicting the stock market movement more precisely based on the historical data analysis, we can reduce the loss of investors largely. To predict the stock market movement, several machine learning techniques are available. Here, we have employed several machine learning approaches on historical stock price data to understand future trends and patterns. To do this, we have applied five regression models namely linear regression, random forest, support vector regression (SVR), vector autoregression (VAR), and long short-term memory (LSTM). To determine the actual performance of the models, several metrics namely explained variance score (EVS), root mean squared error (RMSE), mean square error (MSE), mean absolute error (MAE), R squared score (R² Score), and adjusted R squared score (adjusted R² score) have been used here to evaluate the models’ performance. Among all these models, the LSTM model outperformed the other models in terms of the abovementioned metrics. The obtained EVS, RMSE, MSE, MAE, R² score, and adjusted R² score of LSTM are 0.9726, 0.6220, 0.3869, 0.4306, 0.9716, and 0.9697, respectively. The obtained result is very significant for stock market prediction.