基于机器学习的北京PM2.5预测算法

发布时间：2018-01-06 03:12

本文关键词：基于机器学习的北京PM2.5预测算法　出处：《天津工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：随着细颗粒物(PM2.5)和其他空气污染物在空气的增加,空气污染的问题仍然是中国,特别是污染严重的东北地区的一个很大的问题。本论文基于2015年收集的PM2.5浓度数据和气象数据,使用机器学习方法预测北京市PM2.5的浓度水平。本论文使用了机器学习方法中五种分类算法:Naive Bayes,Multinomial Logistic Regression,Sequential Minimal Optimization,k-Nearest Neighbor(k-NN)和 Random Subspace。为了构建PM2.5预测模型,我们通过WEKA对上述分类算法进行实现。预处理和可视化的工作,我们使用工具R Studio完成。通过研究,我们发现使用Logistic Model Tree(LMT)作为Random Subspace算法的基树分类算法有最好的结果:预测精度为70.92%。同时,我们也发现湿度是影响PM2.5浓度的最大因素。这项研究通过实现对PM2.5的预测,可以及时地告知公众并使其采取适当措施,以保护他们的健康。
[Abstract]:With the increase of fine particulate matter (PM2.5) and other air pollutants in the air, the problem of air pollution is still China. This paper is based on the PM2.5 concentration data and meteorological data collected in 2015. The concentration of PM2.5 in Beijing is predicted by using machine learning method. In this paper, we use five classification algorithms:: naive Bayes. Multinomial Logistic Regression,Sequential Minimal Optimization. K-nearest neighbor k-NN) and Random subspace. to build a PM2.5 prediction model. We implement the above classification algorithm through WEKA. Preprocessing and visualization work, we use the tool R Studio to complete. Through the research. We found that using Logistic Model Tree LMT). The base tree classification algorithm as Random Subspace algorithm has the best result: the prediction accuracy is 70.922. at the same time. We also found that humidity is the biggest factor affecting the concentration of PM2.5. By realizing the prediction of PM2.5, the study can inform the public and take appropriate measures to protect their health in a timely manner.
【学位授予单位】：天津工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：X513;TP181

【相似文献】