Applying Clustering Algorithm to Improve the Futures Prediction Accuracy of Neural Network

Futures forecasting has always been a difficult problem for investors, and has been the subject of much research by domestic and international scholars. In previous studies, traditional analysis methods are far from being able to resolve the important information hidden in stock price movements. With the development of neural network prediction models in recent years, more and more scholars have tried to use neural network models to solve stock prediction problems. However, little research has been done to mine the hidden information of the historical stock movements from multiple aspects and dimensions. Therefore, the innovation of this paper is to improve the input of neural network models by extracting features of long-term price models through clustering algorithms during the training process based on traditional LSTM and RNN neural network prediction models, and mining dimensional information from multiple perspectives of the original stock data. The experimental analysis shows that the proposed network model has better prediction accuracy compared with the traditional LSTM model and RNN model.