An Enhanced Index-based Investment Strategy Design Using Machine Learning

Index enhancement strategy,that is,enhanced index strategy,is a strategy that can have stronger performance based on anchored indexes(such as CSI 300,CSI 500,SSE 50,etc.).To reflect the relevance with the anchor index,the enhancement strategy needs to control the degree of deviation from the index to ensure that both the trend of the index and the excess return exceeding the index return are tracked.Therefore,index enhanced products can ensure stable returns,and make the investment style and investment strategy more transparent and stable.Compared with index funds,they can obtain higher returns,and are favored by investors at home and abroad in the asset management product market.The traditional index enhancement strategy is mainly to realize the above two basic points by analyzing the macro-economy and studying the industry sector and by the idea of "core + satellite".However,quantitative investment relies on the continuous deepening of Computer Science in the field of artificial intelligence.It uses machine learning and in-depth learning methods to process financial time series data and mine the internal relationship between data,so as to achieve the purpose of forecasting data,judging the rise and fall of stock prices and constructing investment portfolios.Firstly,this paper makes a comprehensive review of the research status at home and abroad,which lays a foundation for the rationality and innovation of the strategy design.Then the classical asset pricing theory of finance is studied,and the quantitative multi factor model is used to select the stocks that are more likely to rise in the future.Then the basic contents of index enhancement strategy and multi factor model are explained in detail.Finally,the empirical tool of this paper machine learning algorithm is introduced in detail.The second part is the processing stage of factor data.Factor data,also known as factor exposure value,is also a feature of stocks.The factors in this paper are divided into three categories: fundamental factors,technical analysis factors and market behavior factors.The work of this stage is to preprocess,feature engineering and timing the above factors.Preprocessing is mainly to clean the data,while feature engineering is to perform linear analysis on the factors to improve the interpretation of each factor,and neutralize the factors to achieve the purpose of model tracking index.In addition,the factors with high efficiency in each period are selected to participate in the multi factor model and predict the stock price,so the factors affecting the factors are classified.The last part is the training and back testing of the model,showing the training steps of the factor timing model and the multi factor stock selection model,as well as the detailed comparison before and after the model optimization.By optimizing the fitting degree of the model,increasing the trading frequency and reducing the number of shares held,the risk of the portfolio is further reduced and the return is improved.The performance of the final optimized version of the enhanced strategy design in this paper achieved 24.34% excess return during the back test period;The maximum pullback rate was 9.87% and the sharp ratio was 1.96,which controlled the investment risk.Based on effectively tracking the trend of the CSI 300 index,it has achieved a100% winning rate.