基于随机森林的犯罪风险预测模型研究

发布时间：2018-04-01 08:18

  本文选题：随机森林　切入点：犯罪风险预测　出处：《华东师范大学学报(自然科学版)》2017年04期

【摘要】：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.
[Abstract]:Crime prediction is the premise of crime prevention and an urgent problem to be solved by public security departments.As a combined classification method, stochastic forest has the characteristics of high accuracy, fast speed, stable performance, and can give the evaluation of index importance. This paper applies it to crime risk prediction.The experimental results show that the index set selected by the stochastic forest method can significantly improve the prediction accuracy. The prediction model based on this method is more accurate and stable than the neural network and support vector machine.Can satisfy the demand of crime risk forecast.
【作者单位】： 华东师范大学地理科学学院;
【基金】：国家自然科学基金人才培养项目(J1310028)
【分类号】：D917;TP18
，

本文编号：1694719