An SVM-based method for predicting cigarette sales volume

Not satisfied with the accuracy of cigarette sales volume prediction with linear regression method,an SHPM(SVM-based hybrid prediction method) was proposed based on SVM(Support vector machine) by taking the sales volume of China Tobacco Hunan Industrial Company Limited as objects.Municipal level cigarette sales volume predicted separately by SHPM,linear regression,ARIMA(autoregressive integrated moving average) and SVM were compared and analyzed.The results showed that it was feasible to predict cigarette sales volumes with SVM method.The prediction errors against SVM,linear regression and ARIMA reduced by 9.58%,11.83% and 45.79%,respectively;SHPM prediction method was more effective.