Commodity Futures Trading Strategy Design Based On Catboost Model

With the continuous integration of domestic futures market with the international market,the variety of domestic designed futures price index is becoming more and more abundant,which marks the development of commodity futures has entered a new era.The emergence of commodity futures price index not only makes China's financial product system more abundant,but also helps investors to select commodity futures price index through the design of trading strategy,which makes the risk secondary diversification,and helps investors to conduct asset allocation and prevent the risk of unilateral market decline.As the futures price index gradually appear,the domestic scholars for the research of futures price index is also more and more,the current domestic scholars studies are mostly limited to commodity futures,especially the single commodity futures varieties,and few people study the futures price index,and the futures price index choice for the study of commodity futures and hedge risk is very important.Is adopted in this paper,according to current situation of the Catboost model of machine learning to choose investment targets and combined with minimum variance method,risk parity method and weight method to construct the futures price index such as asset portfolio,and by setting different stop control risk,for small and medium investors and investment institutions try to provide an alternative investment targets and the allocation of weight.The theoretical part of this paper mainly describes the quantitative investment theory and portfolio theory,and then introduces the Catboost model used for prediction,as well as the risk parity method and minimum variance method used for weighting.In the empirical analysis part,this paper refers to a large number of literatures and collects existing data,and selects 20 indexes that affect the futures price index to establish the model.Then,the data from January 2014 to December 2017 were selected as the training set,and the Catboost model was adopted to predict the rise and fall of the sample futures price index in 2018,and the ten commodity futures with the highest probability of rise in 2018 in the verification set were selected as the investment targets.Then,this paper USES the minimum variance method,the equal weight method and the risk parity method to assign different weights to the selected commodity futures,and selects the varieties that are predicted to rise to buy and hold.After adjustment and the addition of stop loss points,the backtest results show that the optimal performance of this strategy for the centralized annualized return is 9.79%,2.98% maximum retractable,and 0.57 sharpe ratio when the minimum variance method is adopted after adjustment and the stop loss point is 2.25%.At the end of the paper in a test set is compared with Catboost model fitting score model,2 it is,and common Lightgbm in machine learning and Xgboost,GBDT,logistic regression were compared with random forest model,comparison proved to measuring results,the optimal model based on Catboost commodity futures trading strategy,in the test set an annual return of 9.46%,2.72% retracement,sharpe ratio of 0.82.This paper hopes that the results of empirical analysis can give the majority of investors some reference for futures price index investment.