Research On FOF Fund Investment Strategy Based On Momentum Effect And Machine Learning Scoring Model

By further selecting funds on the basis of traditional investment portfolios to better diversify investment risks,FOF funds have been attracting the attention of Chinese and foreign investors since their inception.Although the emergence of domestic FOF funds in the foreign financial investment market is relatively late,in 2016,with the release of the China Securities Regulatory Commission's ‘ Guidelines for the Operation of Publicly Offered Securities Investment Funds No.2-Guidelines for Funds in Funds' and the domestic securities investment fund market With the continuous development of scale,FOF funds have developed rapidly in China in recent years.Under such circumstances,how to accurately and efficiently select suitable investment targets from the huge fund products and obtain stable investment performance through a reasonable investment portfolio has become a common concern of securities investment institutions and ordinary investors.This paper first proposes a method to test the momentum effect of FOF funds based on machine learning.Compared with the traditional method,which mainly examines the single return rate index of the fund in the past,this method takes the risk return index of the fund as the model variable of machine learning.Machine learning dynamically models predictive effects to verify momentum effects.In order to solve a large number of data labeling requirements in dynamic modeling,this paper constructs an automatic data labeling method,which is used to label the data of the fund in the current month based on the relative performance of the fund in the next month,which greatly improves the labeling efficiency.In the process of data processing,this paper uses the method of variable binning and WOE coding to map the original data according to the probability distribution,so that the nonlinear data also exhibits linear characteristics.The build lays the foundation.Finally,the momentum effect of the FOF fund pool under the monthly rebalancing frequency is preliminarily verified through dynamic modeling of six machine learning models.After comprehensively comparing the accuracy and interpretability of the models,the results of binary classification using logistic regression The momentum strategy winner and loser combinations are constructed,and the momentum effect is further verified.In the construction of the quantitative investment strategy of FOF funds,this paper continues to use the logistic regression algorithm to map the probability value output by the model to the scores,and constructs a scoring model to further screen funds on the machine learning binary classification results.The investment sequence data of the strategy is obtained by scoring the fund on each rebalancing day.In the aspect of fund investment portfolio,this paper proposes the scoring weight investment portfolio and the scoring weight risk budget portfolio through the constructed scoring weight calculation formula.Since most of the existing quantitative investment platforms only provide backtesting functions for stocks and exchange-traded funds,in order to verify the investment performance of the strategy,this paper builds a set of fund quantitative investment backtesting framework based on Python object-oriented programming technology.Finally,the asset allocation ability,fund screening ability and position management ability of FOF fund investment strategy based on momentum effect and machine learning scoring model are analyzed by comparing with different investment returns.The study found that the performance stability and risk-benefit ratio of the scoreweighted portfolio and the risk-budgeted portfolio based on the scoring model were better.By comparing with the mean variance and risk parity models,it is found that the performance of this quantitative investment strategy in asset allocation is more in line with the investment needs of FOF funds to diversify their positions.By comparing the performance with the CSI300 index and the random investment sequence,it is found that the performance of the Wind five-star rated fund in the backtest range is better than the market benchmark return in the same period.The effectiveness of the scoring model's base selection ability.Finally,the fund's position information in different states of the market is analyzed through the constructed offensive index of fund positions,and it is found that this strategy has significant position management ability during the backtest period.In general,the FOF fund investment strategy research based on the momentum effect and machine learning scoring model has achieved the expected investment effect,and has shown a relatively stable ability to obtain excess market returns under different market conditions during the backtest period.The strategy also enriches the research practice of applying machine learning and quantitative investment methods to the fund investment market.