Fuzzy Clustering Diagnosing and Forecasting on Subtropical High Intensity Index

In order to analyze and diagnose the correlative characters and phase relation between subtropical high and the east-Asia summer monsoon circulation and thermal factors as well as southern hemispheric pressure field, an idea of synthetically combining FCM, GA with FSC to make clustering analyze and diagnosis on subtropical high and its influence factors is presented. Based on observational and statistical facts, some initial subtropical high influence factors in east-Asia monsoon system are chosen to form a high-dimensional feature spaces. Firstly, the FSC technique is used to objectively estimate the clustering number of samples in the feature space, then the GA technique is used to make global optimization search and the FCM technique is used to make further local optimization searching and adjusting on the GA global optimization results. Finally, the synthetical optimized clustering results of subtropical high indexes and its influence factors in the feature mapping space are taken as the classification criterion for subtropical high intensity index classification. By the clustering analyze and diagnosis of using above synthetical techniques, some important portentous characteristic facts of subtropical high changing are revealed and discovered, and a corresponding clustering diagnosis and prediction are carried out.On the basis of above researches, some examples of subtropical high index and its monsoon influence factors are analyzed, diagnosed and predicted by using the synthetical clustering technique, the experiments results show that the classification correctness and forecast precision of 5 grades subtropical high indexes are superior to 60%, especially, the correctness and precision in the case of the weakest and the strongest subtropical high indexes are better than 80%, and similar better subtropical high classification results are also gotten by different examples clustering experiments, the validity and reliability of the synthetical clustering technique are well supported and approved by above experiments.The idea and technique presented not only can overcome the shortcoming of GA/FCM short of sufficient local/global optimization, but also can objectively determine the number of clustering centers. In the actual application, by calculating and judging the mapping location of monsoon influence factors in their high-dimensional feature spaces, the intensity grade of subtropical high index are accordingly determined, and that the subtropical high index diagnosis and forecast are also carried out.