Forecasting Stock Price using LSTM-CNN Method

Abstract

Foreseeing assumes an indispensable part in setting an exchanging methodology or deciding the ideal opportunity to purchase or sell stock. We propose an element combination long transient memory-convolutional neural organization (LSTM-CNN) model, which joins highlights gained from various presentations of similar information, i.e., stock timetable and stock outline pictures, to anticipate stock costs. The proposed model is created by LSTM and CNN, which extricate impermanent and picture components. We assessed the proposed single model (CNN and LSTM) utilizing SPDR S&P 500 ETF information. Our LSTM-CNN combination highlight model surpasses single models in foreseeing evaluating. Also, we track down that the candle graph is the most precise image of a stock diagram that you can use to anticipate costs. Subsequently, this examination shows that prescient mistake can be viably decreased by utilizing a blend of transitory and picture components from similar information as opposed to utilizing these provisions independently.