The Research On Open-ended Fund Rating In China

In recent years,with the rapid development of China's economy and the innovation and improvement of financial market,people have paid more and more attention to the management of their own assets and investment.All kinds of asset management tools also have been developing rapidly in this environment.As a key family financial product,fund also turn to have larger quantity and capital scale.Up to the third quarter of 2017,the total number of funds has reached 4463,and the net assets of the fund also been up to 9956182 million yuan.Compared to the status of China's economy in the world,it seems that the development of fund doesn't match this situation,so there remains great potential for fund to develop.Only more investors participating in fund investment activities can both sides of fund market and investment achieve the gold of development.But the environment where investors live and invest and the features that investors have themselves make it difficult for them to choose which fund to take,leading them to rely more on the information that rating organizations and relevant regulatory authorities disclosure.This also reflects the function of fund rating that reducing the information asymmetry.Therefore,theoretical research work on fund rating is necessary,and it is meaningful to try to use more efficient way of rating or the method that can make information disclosure more transparent.Firstly,this paper describes the vigorous situation of China's fund industry.Secondly,it summarizes the theories and methods of comprehensive evaluation of fund performance both national and abroad,as well as the area of fund rating.Finally,through comparison and reflection,this paper combines funds rating with methods of machine learning to make research,trying to make the three-years funds rating through method of RF-SVM,and then tests the quality of this rating.Based on the rating information on March 31,2017,we finally choose 95 funds which are consistent with the rating results of four institutions which include Morningstar,Shanghai securities,Yinhe Securities and Wind as samples to study the feasibility and applicability of machine learning methods in the field of fund rating.Drawing lessons from the classic indicators of fund performance evaluation,the original index is made of three parts:income and risk,risk adjusted income and manager's capability.In the process of feature selection,we select feature subset by using the random forest method,and then leave the final indicators through the accuracy of each support vector machine on test set.Finally,the support vector machine is constructed under this feature set,and the results of it are compared with the test results of the random forest and support vector machine,also compared with Morningstar and Yinhe Securities on rating quality.From the research results,we find that the application of machine learning method in the field of funds rating is feasible and can be further studied.First,model of random forest,support vector machine and RF-SVM all have good performance on the training set.The prediction results of test set in random forest-support vector machine accuracy is greater than the random forest,then the accuracy of the random forest is greater than the support vector machine.Second,the three methods all do not have the phenomenon of under fitting or over fitting,which shows that machine learning method is feasible in the field of fund rating,and can be further studied.In addition,the final model contains four important features,which are Sharpe ratio,information ratio,Jansen ratio and the coefficient Alpha,covering parts of income,risk adjusted returns and the ability of manager,which fits the real situation.Finally,the performance of the RF-SVM method is slightly better than that of Morningstar and Yinhe securities on inspection of rating quality.