Research On Stock Index Futures Strategy And Optimization Based On HMM

At present,the market conditions for the development of quantitative investment in China have gradually matured,and quantitative investment transactions have developed rapidly due to their low cost and excellent performance.The core element of quantitative investment is stock price trend forecast.The stock price is unpredictable,but it will fluctuate in the form of trend.Once the trend is established,the price will continue to rise or fall until there is a reversal.How to accurately predict the stock price trend,choose the appropriate forecasting method for the stock price development strategy has become a hot spot in the investment field research.Traditional stock price forecasting methods,such as ARIMA,GARCH,etc.,require time series to be stationary and the data to be normally distributed.The stock market is often affected by a variety of random factors,and the data presents an unsteady state.When using the traditional stock price forecasting method,some valid information may be ignored,and the results may be biased.This paper chooses the innovative stock price forecasting method-HMM forecasting method to improve the data stability of traditional forecasting model,considers the influence of stochastic process on stock price,avoids the model over-fitting problem due to too many training parameters,and improves the model.The stability,dynamic portrays the price-driven process.Based on the research background and related literatures,this paper proposes the idea of constructing a strategic simulation experiment based on the research hypothesis.Firstly,the traditional HMM prediction model is constructed.The training set and test set samples are modeled by sliding window training method.The Baum-Welch algorithm is used to continuously train the model until the best parameter estimation is obtained.The Viterbi algorithm is used to decode the prediction set data.The data pattern closest to the likelihood value in the history is identified,and the one-day forecasting method is used to predict the rise and fall of the trading day,and the trading is performed according to the rising and falling signals predicted by the model.Secondly,the HMM Quantitative Timing Strategy based on wavelet denoising is constructed to optimize the analysis.Because the stock market is affected by many random accidental factors,there are noises,and at the same time it has non-stationary and nonlinear characteristics.By introducing wavelet denoising analysis The original stock price signal is optimized.Finally,a comparative study of the performance of the strategy before and after optimization is conducted.The research shows that:(1)HMM has higher prediction accuracy for stock index futures.The quantitative investment strategy based on HMM for stock index futures is significantly higher than the broader market,indicating that HMM prediction research is effective.(2)Since the wavelet denoising has good denoising effect,the original signal after denoising by wavelet can significantly reduce the noise,verifying the saliency and effectiveness of wavelet denoising.In the case of similar risk,the optimized HMM Compared with the traditional HMM strategy,the timing strategy of stock index futures has better investment performance,the model effect is significantly improved,the performance of the strategy is more obviously improved,and it is also proved that the HMM based on wavelet denoising has a broad scope in quantitative investment.Application prospects and potentially huge benefits.