Abstract
Traders have become increasingly reliant on the stock market. However, humans find it difficult to predict the best stock script at the right time to buy or sell from thousands of stock scripts. So, by extracting features from both CNN and LSTM, we expect to forecast stock script selection based on market movement. To improve accuracy in forecasting stock prices, we propose a feature fusion model that combines a CNN and an LSTM to fuse features of different representations from financial time series data. This proposed model is known as a function fusion LSTM-CNN model. When we equate this model to individual CNN and LSTM models, we find that CNN-LSTM outperforms the other two.