犯罪量动态优化组合预测方法  

Dynamic Optimization Combination Forecasting Method for Crime Quantity

作　　者：李明[1] 薛安荣[1] 王富强[1] 吴正寅[1] 

机构地区：[1]江苏大学计算机科学与通信工程学院,江苏镇江212013

出　　处：《计算机工程》2011年第17期274-275,278,共3页Computer Engineering

基　　金：国家自然科学基金资助项目(60773049);江苏大学高级人才启动基金资助项目(09JDG041)

摘　　要：单一预测模型在预测犯罪量时难以协调拟合和泛化关系,从而影响预测结果的准确性。针对以上问题,提出一种数据驱动的可动态优化组合预测方法。以分析自回归求和移动平均模型、向量自回归模型及支持向量机模型的优点为基础,使用后验概率为每个模型赋予权重,结合误差最小原则动态调整权重。实验结果表明,该方法具有较高的预测精度和稳定性,能满足短时犯罪量预测的需要。When single forecasting model forecastes the crime quantity,it is difficult to coordinate the fitting and generalization.The result of forecasting is not accuracy.Aiming at the problem presented above,this paper proposes a data-driven dynamic optimization combination forecasting method based on each advantages of the model virtues of Autoregressive Integrated Moving Average（ARIMA）,Vector Autoregressive Model（VAR） and Support Vector Machine（SVM）.The method gives the weight to each model by using the posterior probability,then dynamic adjusts the weight on the principle of minimum error.Experimental results show that the method has high prediction accuracy and stability to meet the needs of short-time crime forecasting.

关 键 词：犯罪量预测 组合预测模型 支持向量机模型 向量自回归模型 自回归求和移动平均模型 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]