運用機器學習方法校正空氣品質數據

蔡智鈞 ； 尹邦嚴 ； 戴榮賦

《NCS 2017 全國計算機會議》 2017 (2018/01) Pp. 778-782

https://doi.org/10.29428/9789860544169.201801.0146

摘要

在環保意識抬頭的環境下，政府和民間團隊積極投入改善空氣品質的工作，由於政府佈署的定點式空氣品質監測站，無法提供高密度的空氣品質監控，所以許多研究團隊提出佈署微型監控設備的概念來加強空氣品質的監控能力。本研究使用資料修補的技巧來解決監測數據常見的資料遺失問題，同時實驗了三個回歸模型來校正監測數據，實驗結果顯示extreme gradient boosting(XGB)回歸模型能有效的校正監測數據。我們相信本研究提供更精確的監測數據，能幫助民眾更了解環境品質變化，也更願意為自己的健康做行為改變，例如減少焚香及燃燒農業廢棄物等。

關鍵字

機器學習 ； 資料修補 ； 資料校正 ； 空氣品質 ； PM_(2.5)

並列摘要

Under the high awareness of environment protection, governmental and public community research teams progressively devoted to improving air quality. However, due to the fixed location of air quality monitoring sites cannot provide air quality information with high spatial resolution, several research teams deployed micro monitoring sites to enhance the coverage of air quality monitoring area. This paper employs data imputation technique to solve the data missing problem, and compares three regression models on the data calibration problem. The experimental result shows that the extreme gradient boosting (XGB) regression model performs better than the other compared models. We believe that providing more precise air quality information can help people realize the environment change and its impact on human's health. Therefore people are more willing to alter their behaviors such as reducing the burning of incent sticks and agricultural disposals.

並列關鍵字

Machine learning ； imputation ； calibration ； air quality ； PM_(2.5)