基于随机森林的犯罪风险预测模型研究

A forecasting model of crime risk based on random forest

犯罪预测是犯罪预防的前提,也是公安部门亟待解决的问题.随机森林作为一种组合分类方法,具有准确率高、速度快、性能稳定的特性,且能够给出指标重要性评价,本文将其应用于犯罪风险预测中.实验证明,随机森林方法选出的指标集可以显著地提高预测准确率,基于该方法构建的预测模型相较于神经网络与支持向量机具有更高的准确性和稳定性,能够满足犯罪风险预测的需求.

Crime prediction has always been an outstanding issue for public security department. Random forest is a combined classification method with high accuracy, high speed, and stable performance, which is suitable for solving the problem of predicting crime risk. In the meantime, this method can choose the index group for predicting crime risk more objectively. As proved by studies, the index group chosen by random forest method can significantly improve the accuracy of prediction, and the predictive model based of this method is more accurate and stable, so it can meet the demand of crime risk prediction.