GBRT-组合优化预测模型——基于重庆市空气质量数据

GBRT-Combined Optimization Forecasting Model—Based on Air Quality Data of Chongqing

摘要: 在处理多因素数据预测问题时，采用变量选择进行显著因素的筛选，更利于因变量的预测。本次研究的目的是提高重庆市AQI预测的精度，AQI是定量描述空气质量状况的指数，其数值越大说明空气污染状况越严重。对参与空气质量预测的主要6个指标，进行GBRT变量选择余下细颗粒物(PM_2.5)、可吸入颗粒物(PM₁₀)、二氧化氮(NO₂)、臭氧(O₃)这4个指标，采用遗传算法改进的最小二乘支持向量机进行组合预测，其预测误差从1.5329164降到0.1993641，表明该组合预测模型于空气质量预测有很好的应用前景。

Abstract: AQI is an index that quantitatively describes the air quality. The larger the AQI is, the more serious the air pollution is. GBRT was used to select the variables of six main indexes involved in air quality prediction, and the remaining four indexes were fine particulate matter (PM_2.5), inhalable particulate matter (PM₁₀), nitrogen dioxide (NO₂) and ozone (O₃). On this basis, the least square support vector machine improved by genetic algorithm is used for combination prediction, and the prediction error is reduced from 1.5329164 to 0.1993641, which indicates that the combination prediction model has a good application prospect in air quality prediction.