基于模糊信息粒化的支持向量机在犯罪时序预测中的应用

犯罪时间序列一般具有随机性和波动性强的特点。传统的时间序列建模方法利用犯罪时序数据之间的相关性建立预测模型;但对细颗粒度下的信息利用不足。相比之下,基于模糊信息粒化的支持向量机能够在对时间序列的细颗粒度数据进行粒化预处理的基础上建立拟合回归模型,实现粗颗粒度下的时序预测。利用基于模糊信息粒化的支持向量机方法对S市的侵财类案件数据进行分析预测,并与ARIMA模型进行了比较。结果表明该方法在预测精度上要显著优于时间序列预测模型。对公安部门的警务指挥与情报研判具有较高的实用性。

The application of fuzzy information granulation and support vector machine in crime forecasting

Usually, the time series of crime incident shows the features of randomly distributed and oscillated which caused it is hard to be predicted. The time series model which was frequently used to predict crime count fails to take advantage of the sublevel information of time series because it models the data by using the autocorrelations. In contrast, the Fuzzy Information Granularity based Support Vector Machine (FIG-SVM) could more effectively use sublevel information by granulating the time series data and then model it. In this paper, the FIG-SVM method is proposed to analyze and forecast the crime data in S city. The outcomes show that the FIG-SVM method could detect and predict the crime count in higher accuracy, which is much better than time series model ARIMA. The work done in this paper demonstrate that using FIG-SVM to forecast crime could play significant role in police commanding.