Market Timing Strategy Based On HMM

We apply the hidden Markov model to the China stock market,from other scholars,we can see that the previous research on HMM mainly has two modes;one is to divide the sample data into two parts,one for training another for testing the result.It is not difficult to see that the above method is obviously unreasonable,the market behavior is likely to change significantly in a long period,therefore,the model should be updated continuously.In the formation of the transaction,we will choose the most profitable hidden statue to trade.Every transaction will be traded by open price and settled by close price.We choose Shanghai and Shenzhen 300 index,the CSI 500 index and SSE 50 index as our targets,according tote back test result the model operates quite well in sample data.Net value curve shows that the model earns a profit in 2009 while the market experienced a huge retracement but in the second half of 2015 the model also experienced a huge retracement,this shows that the model cannot response very well in complicated market situation.When we remove the 2015 rapid decline caused by withdraw,it is not difficult to see that the net value curve will be greatly improved.In order to verify the stability of model,we test different parameters,the results for different update rate and model training period shows a vast Parameter Plateau,so we can think less likely to model overfitting in the parameter optimization aspect;in addition,we also test the Influence of different special.features of our model input,from the results we can see that the model can achieve good results under different input features.