基于XGBoost算法的非法经营罪量刑预测模型构建及应用  

Crime of Illegal Business Operations Sentencing Prediction Based on Text Mining and XGBoost Model

作　　者：王为久 徐敏亚[1] 徐博希 孟思雨 魏钊 Wang Weijiu;Xu Minya;Xu Boxi;Meng Siyu;Wei Zhao(Guanghua School of Management,Peking University,Beijing 100871)

机构地区：[1]北京大学光华管理学院,北京100871

出　　处：《情报探索》2022年第9期20-28,共9页Information Research

基　　金：国家自然科学基金重点项目“家庭购买决策过程与机制研究:基于‘匹配’和‘社会比较’的视角”(项目编号:71632001);北京大学统计科学中心以及数量经济与数理金融教育部重点实验室(北京大学)项目研究成果。

摘　　要：[目的/意义]旨在利用文本挖掘技术和XGBoost算法构建量刑预测模型,实现量刑预测的智能化。[方法/过程]首先设计了量刑预测模型的构建路径,然后采用XGBoost算法构建了预测模型,最后将模型与多种算法(随机森林、决策树、KNN等)进行了比较,探讨了量刑预测模型的特征选择问题。[结果/结论]基于3208条司法判例数据所构建的量刑预测模型,对量刑严重程度和有期徒刑时长的预测准确率达到了80.02%和78.46%,优于其他算法。研究表明,XGBoost算法在非法经营罪的量刑预测中具有实践应用价值。[Purpose/significance] The paper aims to use text mining technology and XGBoost algorithm to build sentencing prediction model,so as to realize the intellectualization of sentencing prediction. [Method/process]The paper firstly designs the construction path of prediction model,then uses XGBoost algorithm to build the prediction model. Finally,the paper compares the model with many algorithms(random forest,decision tree,KNN,etc.),discusses the feature selection of sentencing prediction model.[Result/conclusion] The sentencing prediction model based on 3208 judicial case data has an accuracy of 80.02% and 78.46% for the severity of sentencing and the length of imprisonment,which is better than other algorithms. The research shows that XGBoost algorithm has practical application value in the sentencing prediction of the crime of illegal business.

关 键 词：非法经营罪 量刑预测 XGBoost 文本挖掘 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]