Abstract: Towards the regular fluctuation and warning fluctuation in the social order situation，an exponential attenuated crime distribution profiling vector to describe the comprehensive social order status is defined and a pre-process algorithm to classify raw data into optimized classes based on Poisson distribution is put forward.Besides，a complex social system tendency analyzing algorithm is proposed to compute the system tendency anomaly index combined with experts’ advice.

Key words: statistic data, attenuation, profiling vector, exception value

摘要： 针对社会治安形势的“规律性的变化”和“警示性的变化”，定义了基于指数衰减的犯罪分布特征向量来描述社会治安综合状态，设计了原始数据预处理算法获取对犯罪数据进行Poisson逼近的最优分类。提出基于统计数据的复杂社会系统趋势分析算法，分析各个统计特征量中的分布特性，归一化后综合专家意见对趋势的异常指标进行计算。

关键词: 统计数据, 衰减, 特征向量, 异常值