Multi-layer Feature Fusion method based on Convolutional Neural Network for Stock Trend Forecasting

Stock trend prediction remains a crucial area of research in the financial domain, where the stock market is characterized by a plethora of indicators that describe various stock features. A key challenge lies in effectively capturing the interrelationships among these features and extracting them comprehensively from stock data. To tackle this, a novel multi-layer feature fusion method based on convolutional neural networks is proposed in this paper, which transforms the time series prediction problem into an image classification problem. Specifically, the representative indicators of stock features are first selected, and then the indicator data are generated into a two-dimensional matrix with different indicators and different times as two dimensions. Finally, the multi-layer feature fusion convolutional neural network model is constructed to classify the stock trends, in which the shallow convolutional features and deep convolutional features of this network are fused. Experiments on real data show that the proposed method can be more effective in fully capturing stock features to the extent of improving the prediction accuracy.