Font Size: a A A

The Research And Empiescal Analysis Of Commodity Futures Price Forecast Based On Hidden Markov Model And Markov Switching Autoregressive Model

Posted on:2018-11-29Degree:MasterType:Thesis
Country:ChinaCandidate:C D LiuFull Text:PDF
Abstract/Summary:PDF Full Text Request
Investors are involved in futures trading for hedging,speculation or arbitrage.If we can make a timely and accurate forecast of futures price volatility,we can effectively reduce the risk of trading,and even get excess returns.Based on Hidden Markov Model(HMM),this paper studies the forecast of futures prices and the prediction of futures market state.We use the relatively active futures varieties,RB,in the domestic futures market,as an empirical research object.Firstly,the processed closing price sequence is used as a single feature factor for empirical analysis under Gaussian Hidden Markov Model(GHMM).Then relaxed the hypothesis that the price sequence is not relevant,and use Markov Switching Autoregressive Model(MS-AR)to further empirical test.Finally,we use Gaussian Mixture-Hidden Markov Model(GM-HMM)for data pattern matching and price forecasting.Firstly,this paper introduces the development of HMM,and then introduces the theoretical basis of the model.Finally,we conduct empirical analysis based on the RB888 data.The empirical process mainly includes three steps,model parameter estimation,decoding and prediction.The model parameter estimation is also the model training process,using the Baum-Welch algorithm under the GHMM model and the GM-HMM model.By setting the initial parameter value and passing the training sample data into the model,the optimal parameter estimation is obtained by iterative estimation.The forecasting process is to use the model that has been learned to predict futures prices and market conditions over the next period of time.In the price forecast,this paper uses the GHMM model to predict the probability distribution of future prices.The GM-HMM model is used to forecast the future price trend based on the 'one-day forecast'and 'weighted forecast'.The state forecast is to predict whether the market status and market conditions will change,'cattle to bear' or 'bear to cattle'.In the MS-AR model,the parameter estimation of the model is mainly realized by Hamilton filtering process.The empirical results show that the GHMM,MS-AR and GM-HMM models have certain accuracy for price prediction.It makes sense for investors predict market trends and trade in the market.
Keywords/Search Tags:commodity futures, price forecast, Hidden Markov Model, Markov Switching Autoregressive Model, Gaussian Mixture-Hidden Markov Model
PDF Full Text Request
Related items