Recommendations on Financial Models for Stock Price Prediction

Stock market prediction greatly influences the economy of a country. Financial analysts estimate the subsequent future prices of stocks by analyzing the current stock market condition. However, it is challenging to select a unique financial model for future Stock Market Prediction (SMP) of any arbitrary sector. An SMP recommendation system is proposed for the selection of the most appropriate sector-wise financial model. Here, ARIMA (in-sample and out-of-sample) along with LSTM [with and without technical indicators (TI)], ANN, and CNN predictive models is used for determining the sector-wise preferences for stocks like INFY, WIPRO, and TCS of the IT sector, HP, RELIANCE, and ITC of the Manufacturing sector, PNB, SOUTHBANK, and YESBANK of the Banking sector. ARIMA (3, 3, 2) out-of-sample prediction model outperforms the LSTM model without TI for TCS and WIPRO stocks. LSTM model without TI performs best for PNB and YESBANK stocks and the ARIMA (3, 3, 6) out-of-sample model performs best for SOUTHBANK stocks. ARIMA in-sample prediction model achieves an MAE improvement of 44% and 74%, respectively over the LSTM model without TI for PNB and SOUTHBANK stocks. With TI, the LSTM model achieves 4%, 8%. 3.5%, 8.5%, and 19.1% improvement in directional accuracy for INFY, TCS, ITC, HP, and YESBANK stocks, respectively over the LSTM model without TI. Moreover, the LSTM model with TI achieves 19.3%, 44.4%, 40.9%, 38.7%, and 42.7% reduction in MAE for WIPRO, TCS, Reliance, PNB, and SOUTHBANK stocks, respectively.