Forecasting Stock Price Using Time-Series Analysis and Deep Learning Techniques

Stock investors must purchase only those stocks whose values are anticipated to grow soon and sell those stocks whose values are anticipated to drop if they want to make a profit. Due to the unstable and unpredictable behavior of stock prices, it is challenging to forecast the accurate value of stocks. With the invention of time series and deep learning, forecasting has become increasingly effective in all areas. This study compares time-series and deep learning models for forecasting stock prices, such as the ARIMA, Facebook Prophet, LSTM-RNN, and CNN. The standard metrics, including MSE, RMSE, MAE, and MAPE, are considered to study the performance of models. The small values of these metrics indicate that the model is good for forecasting stock prices. Results obtained disclosed that the CNN model can compete well with existing methods and is capable of forecasting stock prices for the long term.