Modeling The High Frequency Financial Data Based On The Information Of Limit Order Book

Using high frequency financial data in Shanghai Stock Exchange,we study the market microstructure.We find the financial market sophistication is growing with limit price differences becoming smaller.In 12/2007,the bid-ask spread was 0.0589954,but in 10/2014,the bid-ask spread is 0.01011087,which means the market friction is decreasing.Based on the work of Cao, Hansch, and Wang(2009)[1],we define an indicator of net buy-sell pressure which is used in our final model. Our core idea is to use the Regression-Markov Chain Model to study the high frequency financial data,intra-day mid price.First,we study mid price with regression model in which the indicator of net buy-sell pressure is included.By doing this,we get a new sequence of price,then we test the Markov Property of the new sequence. Based on this two steps,we can now model the mid-price with Regression-Markov chain.Finally,by changing the threshold value and transition steps we have defined, we compare the prediction results of different parameter values.We find that when threshold value is fixed,most of the time,precision rate of prediction is first getting bigger and then getting smaller with transition step decreasing.And when the transition step is fixed,most of the time,precision rate of prediction is also first getting better and then getting worse with the threshold value increasing.When transition step equals 200,threshold value equals 0.8,the precision rate reaches the maximum,55.1%.Actually,during the whole process of modeling the high frequency financial data with Regression-Markov Chain,we have demonstrated an idea for high frequency arbitrage.when some transition probabilityexceeds the threshold value,we send a limit order on sell side,in the mean time,we send a market order on buy side.