华东师范大学学报(自然科学版) ›› 2017, Vol. ›› Issue (4): 89-96.doi: 10.3969/j.issn.1000-5641.2017.04.008

• 计算机科学 • 上一篇    下一篇

基于随机森林的犯罪风险预测模型研究

王雨晨, 过仲阳, 王媛媛   

  1. 华东师范大学 地理科学学院, 上海 200241
  • 收稿日期:2016-06-28 出版日期:2017-07-25 发布日期:2017-07-20
  • 通讯作者: 过仲阳,男,教授,博士生导师,研究方向为数据挖掘和遥感图像处理.E-mail:zyguo@geo.ecnu.edu.cn. E-mail:zyguo@geo.ecnu.edu.cn
  • 作者简介:王雨晨,男,硕士研究生,研究方向为数据挖掘.E-mail:wangyc_ecnu@qq.com.
  • 基金资助:
    国家自然科学基金人才培养项目(J1310028)

A forecasting model of crime risk based on random forest

WANG Yu-chen, GUO Zhong-yang, WANG Yuan-yuan   

  1. School of Geographic Sciences, East China Normal University, Shanghai 200241, China
  • Received:2016-06-28 Online:2017-07-25 Published:2017-07-20

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.

关键词: 随机森林, 犯罪风险预测, 指标集选择

Abstract: 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.

Key words: random forest, crime risk prediction, index group selection

中图分类号: