Dynamic Optimization Combination Forecasting Method for Crime Quantity
WU Zheng-yin · Jisuanji gongcheng · 2011
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.