Forecasting Model of Mass Incidents in China - An Explorative Research Based on Suppport Vector Machine

[Purpose] Mass incidents have emerged as a serious social problem concerning national security in China. So, it is necessary to construct a forecasting model to predict such public events. In this paper, support vector machines are applied to the model. [Method] Based on the social surveys conducted in 119 counties of Shanxi, Gansu and Hubei provinces, 3 multi-class classification problems were proposed, and then 3 multi-class support vector classification forecasting models were constructed. [Results] Preliminary experiments have proved that our method, compared with multiple cumulative logistic regression, should be more effective and accurate(enter method as well as the stepwise one). [Conclusion] It can be concluded from the results that irrationally behavioral intentions can be predicted more accurate than those rational ones. When the collective attitudes are applied to the forecast of the collective behavioral intentions, SVM method was approved to be the most effective approach. This paper represents an originally explorative research.