Stock Price Prediction Method Based On Multi-span Stock Feature

With the rapid development of the national economy,the stable but low-yield fixed interest rate savings model can no longer meet the people’s investment needs.With the characteristics of high yield and high return,stock investment is favored by the majority of investors.Because the fluctuation of stock price is affected by many factors,how to accurately predict the stock price and explore the law of market change has become the current research hotspot and difficulty.In recent years,the artificial intelligence technology has become increasingly mature.Researchers have tried to apply the emerging machine learning method to the field of stock price prediction to solve the problems of the traditional stock price prediction method based on mathematical statistics,such as the inability to capture nonlinear relations and low prediction accuracy.However,the existing stock prediction methods based on machine learning do not fully consider the impact of the correlation between different moving averages and K lines on stock prices,and the prediction accuracy is poor.To solve this problem,a stock price prediction method based on multi span stock characteristics is proposed.The method specifically includes:a multi-span stock feature extraction method based on convolutional self coder(CAE)and a stock price prediction method based on attention mechanism fusion multi-span stock features.The feature extraction method is responsible for dividing the original data according to different time spans,and extracting short-term,medium-term,and long-term depth K-line features using CAE;The stock price prediction method uses a gated circular neural network incorporating attention mechanisms to stack the depth K line features and mean square features learned by the CAE module as input to the model to learn the depth K line features and mean square features of the current trading day.Finally,its output,together with selected fundamental and technical features,is sent to the full connectivity layer to obtain the predicted stock price.In this paper,the above stock index is used as a data set,and the effectiveness of the proposed feature extraction method and stock price prediction method is verified through ablation experiments.Comparative experiments with convolutional neural networks and short-term memory networks have verified that the prediction error of the proposed model is low and the fitting effect is good.