Construction and Optimization of Emergency Prediction Model based on random Forest algorithm

Abstract

Based on the massive emergency data set, the emergency early warning model is established by using machine learning method to predict the realization risk of emergency attack target, and the importance characteristics of emergency early warning can be found. The 135-dimensional emergency features are screened, normalized, single thermal coding and chi-square test. Random forest algorithm and other machine learning algorithms are tested and evaluated. It is proved that random forest algorithm is better than other machine learning algorithms in performance. Based on the global emergency data from 2001 to 2019, this paper uses the random forest algorithm to reduce the dimension of the factors affecting the number of emergency casualties, selects the more important event characteristics, and forecasts the number of emergency casualties in the target country. finally, the international community overseas investment risk assessment and emergency early warning model construction optimization.