應用資料探勘技術預測空污指數對空氣清淨機銷量影響之研究

Application of Data Mining Techniques to Predict Air Pollution Index Study on the Effect on the Sales of Air Purifiers

吳文雄(Kevin Wu)

指導教授 ： 陳民枝 教授

共同指導教授 ： 黃馨逸 副教授(Hsin-I Huang)

國立臺中科技大學/資訊與流通學院/資訊工程系碩士班/碩士(2017年)

摘要

台灣政府於2014年5月正式將PM2.5納入環保署空氣品質監測的項目之一。媒體的報導提醒著民眾越來越重視空氣污染問題。而且隨著PM2.5相關的許多研究發表，一些空氣污染所引發的疾病受到重視。本研究應用支援向量迴歸技術（SVR）來分析空污指數對於空氣清淨機銷量的預測分析。研究資料之一來自於行政院環境保護署公開的十三種污染源監測數據，選取全國66個監測站2015年一整年的PM2.5監測資料。再藉由全國性某連鎖企業的門市位址配對挑選出與66個監測站最接近的66個門市。而研究資料之二來自於這66個門市2015年一整年的空氣清淨機銷售數據來執行實驗。鑒於SVM內的迴歸分析（SVR）在分析預測的領域頗有成效，所以本研究運用R語言程式中的SVR套件來執行訓練與測試分析。實驗結果顯示SVR可以預測出當PM2.5指數升高時空氣清淨機的銷量也顯著增加。誠如本文的雲嘉南區迴歸分析，平原及盆地的地形再加上旁邊有三千公尺的中央山脈佇立所以懸浮粒子不易散去，空氣清淨機的銷量比其他監測區域門市更高。實驗結果能協助空氣清淨機產品相關的業者在相對應的區域中，當空污指數升高時廠商可以設計呼應PM2.5的行銷活動來吸引消費者的購買需求。而且門市應主動增加進貨調高庫存來掌握銷售的黃金時期，畢竟需求預測是經營者每天都必須探討的事。本研究可以提醒企業在PM2.5指數低時減少空氣清淨機的庫存，而相對的PM2.5指數高時增加空氣清淨機的庫存。如此一來對於企業在庫存成本控制及增加營收利益上有很大的貢獻。

關鍵字

支援向量迴歸 ； 空污指數 ； PM2.5 ； 空氣清淨機

並列摘要

On May 2014, the Taiwan government formally incorporated PM2.5 into one of the Environmental Protection Department Air Quality monitoring projects. Media reports reminded the populous to attach importance to air pollution problems. Many related PM2.5 research indicates that some diseases are caused by certain air pollution. In this study, support vector regression (SVR) is used to analyze air pollution index to forecast air purifier sales. First part of the research data is open source from the Executive Yuan’s Environmental Protection Department from monitoring 13 different pollutants. The PM2.5 pollution index is chosen for 66 monitoring stations throughout the country from the year 2015. For the experimental tests, a total 66 test stores, each closest to the pollution test sites, were selected from a national chain store. Second part of the research data is from the year 2015 sales of air purifiers from the 66 test stores. The support vector regression (SVR) analysis of SVM has effective analysis and prediction. Therefore, this study made use of SVR in R language to perform training and test analysis. Experimental results showed that SVR can predict increase sales of air purifier when the PM2.5 index rises. Regression analysis of the Yunjia Southern area showed that sales are the highest amongst all the other test sites due to the fact that the suspended particles (PM) cannot disperse easily in the plains and basin topography together being in the 3000 meters central mountain range. The experimental results may be used by air purifier products related businesses as a reference to plan sales promotions to attract consumer to purchase the related products. Stores can take initiative to stock up air purifiers and to increase inventory in order to take advantage of the golden sales period. This study can be used as a reminder to businesses to reduce inventory of air purifier when the index is low and vice versa when the index is high. This could be beneficiary for inventory control leading to increasing revenue.

並列關鍵字

Support vector regression ； Air pollution index ； PM2.5 ； Air purifier

參考文獻

[17] E. Lavigne, M, Belair, M.T. Do, D.M. Stieb, P. Hystad, A. Donkelaar, R.V. Martin, D.L. Crouse, E. Crighton, H. Chen, J.R. Brook, R.T. Burnett, S. Weichenthal, P.J. Villeneuve, T. To, S. Cakmak, M. Johnson, A.S. Yasseen III, K.C. Johnson, M. Ofner, L. Xie, M. Walker, “Maternal exposure to ambient air pollution and risk of early childhood cancers : A population-based study in Ontario, Canada,” Environment International, Volume 100, Pages 139–147, March 2017.

[6] W. Beckroge, M. Bruse, “ISIS-Traffic, a monitoring model for car induced air pollution in build-up areas,” Science of The Total Environment, Volume 235, Issues 1–3, Pages 371–374, 1 September 1999.

[8] R. Brugha, J. Grigg, “Urban Air Pollution and Respiratory Infections,” Paediatric Respiratory Reviews, Volume 15, Issue 2, Pages 194–199, June 2014.

[9] M. Carminati, L. Pedala, E. Bianchi, F. Nason, G. Dubini, L. Cortelezzi, G. Ferrari, M. Sampietro, Capacitive detection of micrometric airborne particulate matter for solid-state personal air quality monitors, Sensors and Actuators A: Physical, Volume 219, Pages 80–87, 1 November 2014.

[11] R. Chen, B. Hu, Y. Liu, J. Xu, G. Yang, D. Xu, C. Chen, “Beyond PM2.5 : The role of ultrafine particles on adverse health effects of air pollution,” Biochimica et Biophysica Acta (BBA) - General Subjects, Volume 1860, Issue 12, Pages 2844–2855, December 2016.