Research On Arbitrage Trading Strategy Of Black Commodity Futures Based On Catboost Model

The price fluctuation of black commodities has an important impact on the operation and development of the entire economy.In the futures market,futures contracts for black series commodities have become an important tool for manufacturers to control costs.In the futures market,risk management can be carried out by buying and selling futures contracts to reduce the impact of price fluctuations on the real economy.Machine learning technology has been applied to the financial investment market and achieved certain results.Therefore,this paper constructs an arbitrage trading strategy for black series futures commodities based on the Cat Boost model.This thesis takes three representative varieties of black series futures commodities: rebar,iron ore,and coking coal as the object,and studies the construction of arbitrage trading strategies using Cat Boost model.Firstly,the mechanism and theory of cross breed arbitrage trading are studied;Secondly,we conducted a cointegration analysis of the daily price series of the three black series futures varieties.By verifying the cointegration relationship among the three varieties,it is found that there is a stable cointegration relationship among the three varieties,which can result in arbitrage trading.Thirdly,based on the method of mean regression,a trading strategy suitable for black series futures is constructed,and the shortcomings of this strategy are summarized through backtesting data.Fourthly,using Cat Boost algorithm,a prediction model for black series futures price difference series is constructed,and a trading strategy is constructed based on the prediction results.Through back testing,it verifies the price difference prediction trading strategies of different models,and proves that the trading strategy based on Cat Boost algorithm has better trading effects.The conclusion of this article is as follows:(1)The trading opportunities generated by the mean regression trading strategy are limited,the average holding time is long,the overall profit level is low,and the trading risk is high.(2)The Cat Boost model has a lower RMSE for predicting the price difference of the three major black series products,resulting in better predictive performance;(3)In intraday trading,using Cat Boost prediction resulted in better trading results using the intraday trading strategy,with an annualized return of 18.5%,higher than other models,and a maximum pullback rate of 18.5%,lower than other models,while the return of the mean regression model was less than 0;(4)In terms of price difference trend trading,using Cat Boost prediction results in better price difference trend trading,with an annualized return of 24.6%,higher than other models,and a maximum pullback rate of 22.4%,lower than other models,while the return of the mean regression model is less than 0.