Prediction And Analysis Of Copper Index Of Shanghai Futures Based On Markov State Transition Model

Copper as an important industrial metals, is widely used in electrical, light industry, machinery manufacturing, construction industry and other fields, has important position in the development of national economy. But, the price of copper is fluctuating frequently, copper production and sales link on the risk of the smooth running of the economy in our country has a huge impact. At the same time, this article not only to apply Markov state transition model price forecasting process, and the data mining algorithm and Markov state transfer model, this paper proposes a new analysis method, therefore, combined with data mining algorithm based on Markov state transition model of Shanghai copper futures price index to forecast has very important practical and theoretical significance.In this paper, we use Markov state transition model combined with the traditional prediction method and data mining algorithm analysis and forecast the cooper price data from January 1, 2001 to October 2015. First using Markov state transition model of 150 session of Shanghai copper futures index logarithm yield were analyzed, then divided the 150 session of Shanghai copper futures index logarithmic yield into two states, combined with the two kinds of state data of the next trading day logarithm yield of Shanghai copper futures index to predict and estimate. In the process of prediction and estimation, this paper adopted in the process of the analysis of the Markov state transition model of scholars often take forward prediction method, as well as the data mining algorithm, linear regression, regression, support vector machine(SVM) regression, decision tree to bagging, boosting returning method respectively to forecast the logarithm yield of Shanghai copper futures index and estimates.The research results show that using the Markov state transition model can well predict index futures copper price fluctuations. Among them, the use of the Markov state transition model based on forward prediction method for investment, annual average of 7.06%, and the average transaction number nor more than 5 times a year; Markov state transition model based on data mining algorithm performance than traditional Markov state transition model based on forward prediction method of performance, more excellent yield in the direction of the mean and variance, the rise and fall is superior to the traditional model in prediction accuracy index, based on data mining methods, boosting the regression Markov state transition model on the annual average yields the indexes, the optimal performance, yield is as high as 19.63%, the yield significantly better than other models;(support vector machine(SVM) regression) based on data mining method of Markov state transition model in yield variance on the test, the optimal performance, yield variance is only 2.22%, at the same time, the Markov state transition model based on data mining methods can improve to a certain extent forward prediction method of Markov state transition model for the problem of shortage of volatile market analysis ability. This suggests that, by further combining with data mining algorithms, we can further improve the Markov state transition model in sword sequence analysis in the process of prediction ability, improve the practicality of the models and explain ability, further development and improvement of Markov state transition model.