Research On Short-term Trend Forecast Of Non-ferrous Metal Futures Price Index

In the current financial industry,quantitative strategies have been widely used in various investment activities.The progress of various technologies centering on big data has promoted great changes in the business model of the financial industry,making the methods involved in quantitative strategies more abundant.Big data,intelligent investment management,algorithm-based trading and other fintech means have been frequently applied in the securities field,while the application of related technologies in the futures market lags far behind that of other financial peers.The identification of trend is an important part of investment trading and risk management based on futures market.The application of the data mining technology represented by machine learning in the prediction of the price trend of the domestic futures market and the verification of its applicability and limitations is not only a result of the development requirements of the futures industry,but also an active exploration.Based on the applicability of machine learning algorithms in predicting the price trend of domestic futures market,this paper selects Industry Metal Commodity Index(later referred to as “IMCI”)compiled and released by Shanghai futures exchange as the empirical research object,and uses Logistic model and XGBoost model as the research tools.Firstly,the historical data of IMCI is obtained,then various technical indicators are designed,and the technical indicators are discretized and validated.In addition,on the basis of the smoothed price,the short-term trend of the index is classified as two kinds of patterns,upward and downward,according to the 5-day change of the smoothed price.After the construction of feature factors and classification tags,the Logistic model and XGBoost model were trained and applied to predict the trend category of the next period at the daily frequency.This paper firstly introduces the theoretical basis involved in the empirical process,including the relevant research results of price trend prediction,Logistic algorithm and XGBoost algorithm.After that,the empirical analysis process is carried out.In particular,it is necessary to mention that the time period interval involved in trend classification should correspond with the date parameter involved in the construction of technical index characteristics,and feedback processing should be carried out according to the empirical results.Only in this way can we obtain better empirical results.In this paper,the Logistic model,as a single model,is optimized mainly based on the results of coefficient significance test,model significance test and model goodness of fit test,without considering the problem of super parameters.In the tuning process of XGBoost model,we use the cross validation function of R language,and select the optimal parameter combination by traversing different parameter combinations in combination with the grid search algorithm.The empirical results of this paper show that,on the basis of the more reasonable and effective index characteristics,the Logistic model and Xgboost model have a better prediction effect on the price trend of IMCI in a short period.It can be seen that the machine learning method has good applicability in predicting the price trend of the futures market.