Research And Application Of Machine Learning Stock Selection Algorithm Fused With Behavioral Finance

At present,the related algorithms of machine learning are mostly used in professional fields such as image recognition and pattern recognition which have high classification accuracy.This mainly benefit from the strong similarity of data samples with the same category.The model fully learns the training set and achieves a more accurate prediction on the test set.However,in the financial field,most data does not have this feature,and the predictive accuracy of the machine learning model is far lower than the above.Therefore,if we want to explore the association rules between financial data,we should try to integrate financial theoretical logic with machine learning algorithms to improve the effectiveness of the algorithm in the financial field.With the market trading data of CSI300 Index as the data source,processing data according to the characteristic of time series of stock data,using the ensemble learning algorithm : the KNN Classifier,GradientBoosting Classifier,Adaboost Classifier,these classifiers are aggregated into a new classifier model by voting algorithm to predict stock market information of the stock index T+1 trading day,the improved voting algorithm considers the classification effects of weak classifiers,the final classification result has been improved.Researching stock selection algorithm based on the effective prediction of the CSI300index: by analyzing the company’s latest quarterly financial data,combining the Affinity Propagation with the Adaboost algorithm to construct the fundamental data classifier.The classifier classifies the stocks and forms a stock picking pool.Market-timing is carried out by the reversal strategy in behavioral finance combined with technical indicators--KDJ,which commonly used in the stock technical analysis field.Time to sell is determined by setting the profitability or stop-loss rate.Combine the market forecast and stock selection to formulate the stock selection algorithm model: conducting revenue detection and risk assessment on the quantitative trading platform "you kuang",the results show that the stock selection algorithm can surpass the market return and the various risk indicators have performed well.This is a relatively reliable investment algorithm,and has certain reference value for minority investors.