An Efficient Ensembled Deep Learning Technique for Stock Price Prediction

Digitization has recently had a significant impact on the finance industry in terms of analyzing large amounts of time series data for research and forecasting purposes. Considering the high volume of data created by the stock market every second, performing an effective data analysis is critical for international markets. In order to recognize the significance of real-time stock market predictions, two common techniques are employed to capture time series patterns, they are neural network-based LSTM and the well-known ARIMA method. In order to overcome specific limitations, this research proposes an ensemble model by combining the ARIMA and LSTM with a weighted sum unit. The predictions from the two models are combined by utilizing selected weights to maximize performance on a testing dataset. The proposed system that combines the deep learning techniques with classical time series analysis has the potential to improve stock price prediction. Evaluation criteria such as RMSE, MAE, MSE, and MDA are used to measure the accuracy of the proposed model.