FOF Sub-Fund Selection Strategy Based On Machine Learning

In recent years,the development of Fund of Funds(FOF)in China has been very rapid,and FOFs have gradually become an important category of funds with broad prospects for future development.Because the goal of FOF’s asset allocation needs to be achieved through the performance of specific sub-funds,the selection of sub-funds is extremely important in FOF fund management.Currently,there are numerous funds in China,and the momentum effect of domestic funds is not obvious,and the performance of funds fluctuates greatly,especially when facing systemic risks caused by external factors.The selection of equity sub-funds presents significant challenges.Researching FOF sub-fund selection strategies,especially how to make FOF sub-fund selection strategies still effective in achieving FOF investment goals when facing extreme market risks,is an unavoidable issue in the development process of FOF.Scholars have gradually conducted research on FOF funds in these years,with the focus of most studies on FOF’s asset allocation and sub-fund composition,forming the theoretical basis for how FOF conducts asset allocation and selects sub-funds.Existing research has also explored various algorithms for using machine learning.However,there is still relatively little discussion on evaluating and selecting equity sub-funds with large return fluctuations using machine learning methods,especially in the market environment of extreme risks,and how to use data-driven strategies to select sub-funds.This thesis sorts out the common methods and evaluation indicators of FOF funds in selecting equity sub-funds,as well as the common algorithms of machine learning methods for fund screening,and attempts to use machine learning methods to improve the selection strategies of existing equity sub-funds,and selects the time window from early 2018 to the end of 2022 to conduct empirical research on domestic funds.It is worth noting that in the three years from the beginning of 2020 to the end of 2022,the financial market has experienced the impact of various extreme risks,including the global outbreak of large-scale the Covid-19 epidemics,The RussoUkrainian War,energy crises,and repeated domestic industrial policies,and domestic mutual funds have fall down repeatedly comprehensively and sharply,which has brought unprecedented difficulties to the selection of FOF sub-funds.The extreme environment caused by these external factors to the market is an inevitable problem for FOF managers when choosing sub-funds,and this market data is also a very rare sample for studying the investment strategy of FOF funds.This thesis finds an effective way to dynamically sense market trends and set the forecast targets of FOF sub-funds in a data-driven manner to effectively respond to the extreme risk environment of the market.This thesis attempts to propose a data-driven approach to optimize sub-fund selection strategies in different market environments by setting different sub-fund selection goals and using a multi-level and multi-objective machine learning prediction method.In the face of different market situations,especially extreme market situations,this strategy can self-adjust and respond in time so that the investment portfolio of sub-funds can closely track the performance benchmark of equity assets and strive to obtain excess returns.This strategy provides a new idea for investors to construct FOF sub-fund selection strategies based on machine learning methods and has some practical guidance significance.