The Application Of Neural Network In The Research Of Persistence Of Stock Fund Performance

China's open stock fund started from the first fund in September 2001,but it is only 10 years since.Since its birth,because of its good liquidity,high transparency and strong incentive and constraints,it has gradually become the investment object of the investors.With the development of open stock fund and its increasingly prominent role in the securities market,it not only attracts the enthusiasm of investors,but also attracts the attention of domestic and foreign scholars to China's emerging securities market.Whether the fund market is effective or not,whether the fund manager can continue to defeat the market and whether the performance of the fund is sustained before and after the fund is the key and difficult part of the research.Spearman rank correlation test,scan statistics and cross-sectional regression methods are commonly used methods in the study of sustainability of fund performance,but these methods have failed to effectively classify th e overall fund sample,did not consider the extreme value interference,and failed to resolve distribution function not applicable.These deficiencies affect the accuracy of the measurement of the continuity of the fund's performance.The neural network algorithm can utilize the information transfer between neurons to integrate the continuous usefulness information for fund performance among the three methods of estimation,and the neural network algorithm uses the logistic function as the neuron processing function to better fit the non-linear relationship.Through the subordination function of the fund according to performance and ranking and other factors,neural network algorithms can effectively classify and reject outliers.This paper introduces the neural network algorithm and empirically measures the persistence of the performance of China's equity funds.Sample data range for measuring the sustainability of fund performance is from October 2014 to October 2017.As a result,it has been found that the sustainability of the performance of China's equity funds has fluctuated at different times,and that the funds with continuous performance have a 13% confidence level at a 90% confidence level.Based on a neural network algorithm that performs enough learning times,the accuracy of the continuous detection of the performance of the fund reached 70.37% during the inspection period.Therefore,this paper believes that the neural network algorithm can better solve the deficiencies in the current fund performance persistence measurement method,can improve the accuracy of the fund's performance persistence measurement and optimize the measured data structure.