On new effective integrated PI algorithm to automatic processing of space-time images in earthquake activities

The PI algorithm is a the seismic physical statistical prediction models based on statistical mechanics of complex systems, which were raised by the seismic activity forecast studies and obtained rapid development in the applications. But its general stability of earthquake statistics, the good and bad of the results has a close relationship with the sensitivity of some parameters and the specific parameter settings of the PI algorithm, while using the ROC icon scoring method, including the separate two parts of the prospective rate and false rate, this is contrary to the customary understanding of quantitative score, and tend to score high, inconsistent with the actual situation. For these above reasons, we will use the R value with quantitative score to replace the ROC method with qualitative score, serves as the sole quantitative evaluation criteria of good and bad of PI predicted results, and on this basis, we also proposed multi-scale grid search strategy to realize the automatic optimization of optimal PI parameters. Through this technological transformation and integration, the new obtained PI model has some good with fast computation, the global optimum and the appropriate score. Accordingly, in order to analyze space-time evolution images of seismic activity and test the integrated the new model, study on seismic activity in Yunnan as an example, using the grid search algorithm to obtain the best parameters of PI method, and after based on the PI algorithms, the retrospective prediction tests of seismic activity in Yunnan region since 1970 were carried out, and achieved remarkable results, raise a certain amount of prediction accuracy and the actual capacity. It is ultimately proved that the above improvement and integration of PI algorithm are reasonable and the practical application is feasible.