基于集成学习的毒品违法异常行为预测模型研究

Research on the Predictive Models for Abnormal Behavior of Drug Violation Based on Ensemble Learning

通过机器学习和深度学习,可以在复杂和动态环境中提取毒品违法异常行为数据中的隐藏关系,在预测警务和打击犯罪领域具有广阔的应用前景。然而,传统的机器学习算法在“打击毒品犯罪、预防毒品犯罪”中不能达到很好的预测效果。为此,提出了一种基于k-fold bagging集成学习的融合模型,首先优化了机器学习和深度学习模型作为基学习器,其次选用k-fold bagging方法进行集成训练,最后采取硬投票策略,得到预测性能最佳的模型。测试结果表明:该集成学习模型在两组训练集上的准确率较单一基模型的平均准确率分别提高了9.82%和6.96%,同时将该模型成功应用于毒品违法异常行为预测,为集成学习在预测警务和打击犯罪中的深度应用指明了新方向。

Through machine learning and deep learning,hidden relationships in drug violation abnormal behavior data could be extracted in the complex and dynamic environment,which had broad application prospects in the field of predictive policing and crime combating.However,the traditional machine learn⁃ing algorithm could not achieve good prediction effect in“combating drug crime and preventing drug crime”.Therefore,a fusion model based on k⁃fold bagging ensemble learning was proposed.Firstly,ma⁃chine learning and deep learning models were optimized as the base learner.Then,k⁃fold bagging meth⁃od was selected for ensemble training.Finally,hard voting strategy was adopted to obtain the model with the best predictive performance.The results showed that the accuracy of the ensemble learning model on the two training sets was 9.82%and 6.96%higher than the average accuracy of the single base model.Meanwhile,the model had been successfully applied to the prediction of abnormal behavior of drug viola⁃tion,indicating a new direction for the deep application of ensemble learning in the prediction of police and crime combating.