本研究應用倒傳遞類神經網路(Back-Propagation Neural Network, BNN)與多元線性迴歸(Multiple Linear Regression, MLR)建立台中地區空氣品質之預測模式,BNN及MLR變數因子分別為以懸浮微粒(PM10)、懸浮微粒(PM2.5)、管線開挖工程、建築(房屋)工程(RC)、其他營建工、建築(房屋)工程(SRC)等條件,以2008年1月~11月為網路輸入參數建立最適化網路,對2008年12月之PM2.5作預測,並與MLR作比較。本研究利用BNN及MLR模擬預測臺中市PM2.5之研究結果顯示,BNN及MLR模型對於PM2.5之濃度及變動趨勢皆可掌握,MLR之測試及預測效果較佳,其中相關係數最低以忠明(6V1及4V1、2V1)為0.3,西屯為(2V1)0.36,BNN之預測效果較MLR差,其中相關係數最低以忠明(3V1)為0.27,西屯(6V1)為0.23。整体而言,BNN預測結果優於MLR。
This study employed Back-Propagation Neural Network(BNN) method and Multiple Linear Regression (MLR) method to establish an air quality prediction model of Taichung city. We used PM10 as the variable factor and PM2.5, building constructions using reinforced concrete(RC), building constructions using Steel Reinforced Concerte(SRC), Building constructions as input parameters in BNN method for an optimizing network. We inputted data from January to November 2008 as parameters, then used the model to predicted the air quality of December, 2008.This research uses Back-propagation Neural Network (BNN) and Multiple Linear Regression (MLR) to have produced a simulated forecast for PM2.5 in Taichung City. The results suggested that, both BNN and MLR forecasting models have showed capabilities in capturing the changes and trend of the PM2.5 concentration level.MLR outperformed BNN in forecasting results. The forecasted relative factor islowest at Chung Ming (6V1,4V1,2V1) at 0.3, and Xitun (2V1) at 0.36.BNN shows less than MLR in forecasting results in Chung Ming (3V1) at 0.27, Xitun (6V1) at 0.23. Overall speaking BNN’s forecast is better than MLR.