基于时空序列混合模型的犯罪情报预测分析

A Model of Crime Intelligence Prediction Based on a Hybrid Model of Spatio-temporal Sequence

[目的/意义]犯罪行为的分布和发生在时间上和空间上具有一定的规律性。犯罪情报预测分析对于获取未来的犯罪变化动态具有重要意义。传统的犯罪情报预测分析,要求熟悉政治、人文、经济、地理等多方面因素和社会犯罪动态变化规律,具有一定的局限性。因此需要探究新方法。[方法/过程]针对犯罪数据构建了神经网络和STARM A(时空自相关移动平均模型)的时空序列混合模型,根据历史犯罪数据预测未来发生犯罪的数量变化。首先利用神经网络提取犯罪数据中非线性特征,然后对残差建立STARMA模型,整合出最终的预测结果。[结果/结论]既弥补了传统STARMA模型无法挖掘非线性关系的不足,又满足了模型所需数据的平稳性的要求。通过实验验证了该方法可减小预测误差,在犯罪情报预测方面更加准确。

[Purpose/Significance]The distribution and occurrence of criminal behaviors have certain regularity in time and space.The a nalysis of crime intelligence prediction is of great significance to obtain the future crime dynamics.Traditional crime intelligence prediction analysis requires police officers to be familiar with politics,humanities,economy,geography and the law of social crime.Because the tra ditional analysis has certain limitations,there is a real need for new approaches exploration.[Method/Process]A hybrid model of spatial-temporal sequences of neural networks and STARMA(spatio-temporal autoregressive and moving average)was constructed for crime data,and the number of future crimes was predicted according to historical crime data.Firstly,the neural network was used to extract the nonlinear characteristics of the crime data.Then,the STARMA model was established and the final prediction result was integrated.[Re sult/Conclusion]It can not only make up for the deficiency of the traditional STARMA model in nonlinear relationship digging,but also satisfy the data stability requirement of the model.The experimental results show that the method can reduce the prediction error and im proves the prediction accuracy of crime information.