Mass Incidents Prediction Based on ID3-SMOTE Algorithm

The current domestic mass incidents are shown as the characteristics of organized, complex, political and violent, seriously affect the social's harmony and stability. It is an effective way to prevent the occurrence of mass incidents by means of scientific methods. In the past, the early prediction methods of mass incidents are mainly about the qualitative analysis or simple quantitative analysis methods, not including scientific and reliable data as support. In this paper, an ID3 decision tree is introduced for mass incidents prediction with China's mass incidents real data at present, realizing the function of predicting hazard level of each group event. In order to improve accuracy rate and overcome the data imbalance problerms, a further optimization of the algorithm called SMOTE is attempted. The experimental results have been greatly promoted through this approach. The idea of machine learning is introduced into the field of mass event early warning, has subversive meaning in the incidents mass prediction area, which implies more accurate decision can be made by the government.