Mining Of Momentum Pattern In A–Share Market

A-share market is a very complex system,its external performance is series of time sequences,including price,volume,book orders at every seconds.How to carry out reasonable investment strategy with useful information,mined from various and multifarious time series,is the origin of this paper.Momentum effect is a long standing term within the stock market.Taking advantage of momentum effect to guide our investment decision is an important study direction.We proposed the ?Momentum pattern‘ based on the momentum effect in this paper and described this pattern quantitatively and statistically.The result shows that this pattern we proposed do exit in our A-share market and there is some relationship between those different momentum patterns.Then,we used the stock market trading data and the machine learning algorithms trained our model to try to predict this pattern,and then evaluated the training performance with out-sample data set,thus describing the predictability of momentum pattern.The outcome is that the momentum pattern can be predicted and great differences occurred when applying different algorithms.Lastly,we chose the model that had relatively better performance,and applied it into the investment back test with historic data,finally achieving desired result in some degree.