Research On Deep Convolution Axiomatic Fuzzy Set System And Its Stock Price Prediction And Application

Stocks are part of the ownership of a corporation and a long-term credit instrument in the capital market.Stock trading not only provides investors with opportunities for high returns,but also brings high risks.Therefore,investors need to make a more accurate judgment on the trend of stock price.With the continuous development of science and technology,many researchers began to explore the method of stock price prediction,hoping to achieve objective scientific investment and reduce investment risk.At present,the commonly used stock price prediction methods include deep convolution neural networks,support vector machine and ARIMA time series prediction.However,the prediction models of deep convolution neural networks lack of interpretability.Interpretability occupies a very high position in the finance.Some prediction methods based on traditional fuzzy theory have certain interpretability,but the prediction effect is not satisfactory.The proposal of the deep convolution fuzzy system makes the prediction model interpretable and improves the accuracy of short-term prediction.To improve the accuracy and stability of multi-step stock prediction,the deep convolution axiomatic fuzzy set system is proposed based on axiomatic fuzzy set theory and the model structure of the deep convolution fuzzy system.Experiments show that the deep convolution axiomatic fuzzy set system has high accuracy and stability for multi-step prediction of stock price.Aiming at the limitation of improving prediction accuracy by using a single stock price training model,this thesis uses the Pearson correlation coefficient to screen the relevant indexes of stocks,and introduces the indexes with high correlation into the deep convolution axiomatic fuzzy set system.By comparing the multi-factor experiment of introducing multiple stock-related indexes with the single factor experiment of single stock price,it is verified that the introduction of stock related indexes into a deep convolution axiomatic fuzzy set system can greatly improve the accuracy of prediction.The deep convolution axiomatic fuzzy set system has the problems of long computing time and insufficient single machine performance.To solve this problem,this thesis first uses the sorting algorithm and search algorithm to optimize the time complexity of the coherence membership of the axiomatic fuzzy set,and then introduces the distributed framework to build the distributed computing system,and the distributed parallel computing method is used to complete the calculation of the system model,which improves the overall calculation speed of the system and reduces the configuration requirements of a single machine.The experiment proves that the distributed computing system greatly improves the computing speed of the system.In order to further simplify the process of system model training and prediction,this thesis develops a software system to train and predict the system model in the way of visual operation interface,which saves time costs and reduces repetitive workloads.