基于随机森林和SVR的阿片类药物危机分析

Opioid crisis analysis based on random forest and SVR with sliding window

阿片类药物危机严重威胁着社会的稳定与发展,美国作为危机发生的严重地带,当地政府已下达了针对阿片类药物危机的紧急戒严。依据由NFLIS提供的2010-2017年统计美国五个州的药品事件与人口分布情况,由层次化分析,建立随机森林模型计算特征重要性,随后再通过回归模型,对美国部分城市未来的受影响程度作预测。在回归模型部分,分别使用了深度学习中的Seq2Seq与滑动窗口机制的SVR两种方法分析,通过实验结果对比二者在该问题上的优劣,对五个州内城市2017年的受影响程度作预测。结果表明模型对其中的四个州的预测效果较好。

Opioid crisis is a serious threat to social stability and development.As a serious area of the crisis,the local government has issued an emergency martial law against the opioid crisis in the United States.According to the statistics of drug incidents and population distribution in five states provided by NFLIS from 2010 to 2017,a stochastic forest model was established to calculate the importance of characteristics based on hierarchical analysis,and then a regression model was used to predict the future impact of some cities in the United States.In the part of regression model,Seq2Seq in depth learning and SVR in sliding window mechanism are used to analyze two methods respectively.Through the experimental results,the advantages and disadvantages of two methods on this issue are compared,and the impact degree of five cities in the state in 2017 is predicted.The results show that the model has a good prediction effect for four of the states.