摘要
通过求助类警情风险分析,可以实现风险预测和特征分析,为公安机关提供决策支撑。针对文字、风险两种不同特征的求助类警情风险分析方式进行对比研究。基于文字特征的分析方式,使用Word2vec提取警情文字特征,通过随机森林—粒子群优化模型进行训练;基于风险特征的分析方式,从主体风险、行为风险、场所风险、要素风险四个维度构建求助类警情风险评估指标体系,使用BERT-MRC提取警情风险特征,设计随机森林—粒子群优化模型进行训练。实验证明,基于风险特征的随机森林—粒子群优化模型能基本实现风险预测和特征权重分析功能,具有实践意义。
Through the risk analysis of help-seeking police alerts risk prediction and feature analysis can be achieved providing de-cision-making support for public security organs.A comparative study is conducted on two different approaches for risk analysis of help-seeking police alerts text-based and risk-based.In the text-based analysis method Word2vec is used to extract textual features of police alerts and a random forest-particle swarm optimization model is used for training.In the risk-based analysis method a risk assessment index system for help-seeking police alerts is constructed from four dimensions subject risk behavioral risk venue risk and key factor risk.BERT-MRC is used to extract risk-related features of police alerts and a random forest-particle swarm optimization model is applied for training.Experimental results show that the random forest-particle swarm optimization model based on risk features can basically a-chieve risk prediction and feature weight analysis demonstrating practical significance.
作者
杜镐
Du Hao(Zhejiang Police College,Hangzhou 310053)
出处
《西部学刊》
2025年第15期60-63,共4页
Journal of Western
基金
浙江警察学院2024年度校级课题“警情舆情风险评估及智能化决策模型研究”(编号:2024QNY002)的有关成果。
关键词
警情
求助
风险分析
人工智能
police alerts
help-seeking
risk analysis
artificial intelligence
