48potential energy surfaces (PES) based on LEPS molde have been designed, which are parameterized with 6 geometry characters and 240 samples are calculated by QCT method (1 000 trajectories in each sample), according to the different reaction conditions, which had been parameterized with 7 dynamical characters. Then, the methodolgy of pattern recognition is used to analyze the relationship between the observables and the PESs and the reaction conditions. After this analysis, some known rules and new informations about the formation of long -lived trajectories are obtained. By using the discriminate equation obtained through pattern recognition, the behavior of unknown samples can be predicted.

Key words: PATTERN RECOGNITION, POTENTIAL ENERGY SURFACES