基于机器学习在空气质量指数中的应用

Application of Machine Learning to Air Quality Index

利用机器学习和多元线性回归模型对西安市近一年的空气质量指数进行了研究,首先利用随机森林思想对数据进行了补齐,然后运用交叉验证对神经网络模型选取最优的隐层节点数和训练周期数,最后,通过比较两种模型的拟合效果发现,神经网络模型在对空气质量指数的预测效果明显好于多元线性回归模型。

This paper uses machine learning and multivariate linear regression model to study air quality index of Xian City in nearly one year, firstly uses random forest philosophy to complete the data, then uses cross- validation on the neural network model to select the optimal number of hidden layer nodes and iterations, and finally by comparing the fitting effect of the two models, finds that neural network model is significantly better than multivariate linear regression model for the prediction effect of air quality index.