Research And Application Of Corrected Weighted Markov Model By Method Of Stochastic Optimization Burnishing

In the stochastic process and theory of statistical prediction, time series models, such as grey theory and Markov process and so on, have been widely used. But, only several close data of time-correlated have higher accuracy of grey model prediction, the farther it is away from the reality of time, the more inferior prediction effect becomes. Markov process is based on the transition probability between the states to conjecture the future development of system, which requires the states have no after effectiveness and stationary process and so on. It can fully and reasonably use information that weighted sums of Markov chain which use various steps's autocorrelation icoefficient as weight is applied prediction. The prediction of corrected weighted Markov model based on slipping average grey prediction can also remedy defects of poor predicted accuracy which grey prediction model bring about to data sequence of large stochastic volatility. So, slipping average grey- weighted Markov combined model combined with theory of slipping average grey and theory of weighted Markov, has improved prediction of global data, while the effect which local data on the model don't eliminate. Therefore, the paper modifies the weighted Markov model with random optimization polishing method, which eliminates the effect of local data on the whole and get the corrected weighted Markov model by method of stochastic optimization burnishing.In nature, as the variability ,diversity and complexity of weather conditions, there are plenty of uncertainty and randomness with the precipitation process, therefore, after the statistical analysis of JingDeZhen's historical data of precipitation , weighted Markov model's maximum relative error is 66.14%, the maximum relative error of slipping average grey-weighted combination Markov model is 60.29%, and the maximum relative error of corrected weighted Markov model by method of stochastic optimization burnishing is 12.78%.The results shows that: the prediction accuracy of corrected weighted Markov model by method of stochastic optimization burnishing has been more significantly improved than the two other models, which have strong research value and practical function and can be widely applied to the other subject areas.