Multi-classification Integration Model Of Stock Index Based On Feature Selection

This paper first introduces the importance of stock index in financial market and emphasizes the scientific and effective application of machine learning in financial data analysis.In various domestic and foreign researches,machine learning methods show excellent results which are different from traditional methods.Secondly,due to the important position of csi 300 index in the financial market,this paper chooses csi 300 index as the object and carries out the research by combining theoretical derivation and empirical analysis,as well as quantitative experiment and qualitative analysis.Nowadays,the problem of multi-classification of stock index is still challenging,and the results of good classification can often provide a reference for investors to make decisions.In this paper,aiming at this kind of problem,the return rate of stock index and corresponding quantiles of different time intervals are calculated to determine the labels of the samples,and the data sets of three classifications are constructed.Then,in order to screen the features from different perspectives,three feature selection methods were adopted,namely,information entropy gain,Relief F algorithm,and principal component analysis(pca)to obtain the feature importance order.This paper proposes a multi-classification integration model of stock index based on feature selection,and makes an empirical analysis by writing Python code.In the process of building the model,the convolutional layer and cyclic layer are used to build the network structure,and then the support vector set is combined to form the model.The output values obtained by combining the three feature selection methods and the model are integrated through OWA operator to calculate the final prediction results.Finally,through fitting experiments on data sets with different time intervals,a better classification effect was achieved in various evaluation indicators,including accuracy,f1-score and AUC.In the multi-classification integration model of stock index based on feature selection proposed in this paper,three feature selection methods are used to reduce the dimension and improve the computing speed,while retaining the key information.The convolutional layer and the cyclic layer ensure that the continuous time series information is fully learned,and then the support vector machine is combined to enhance the nonlinear expression ability of the model,which reflects the high efficiency of the classification principle that only part of the support vector is used.According tothe classification results of the three feature selection methods combined with the model,OWA operator can integrate the output values of the model,realize the result of model optimization and improve the classification accuracy.