台灣溫泉主管機關逐年監測與紀錄溫泉水位資料，但因溫泉泉溫與泉性特殊，常導致儀器記錄資料不完整，本研究利用GM(1,1)進行溫泉水位之資料補遺，以便於後續之溫泉水位資料相關分析。本研究利用溫泉水位資料完整部份，重複訓練礁溪溫泉水位之灰預測最佳模式，最佳推估模式建置完成後，針對觀測水位資?遺漏部分進?補遺，以建置完整?續之溫泉水位資?。本研究以礁溪溫泉區2010年水位資料，進行最佳模式推估，分別針對溫泉公園站、國小站及太子站之遺?溫泉水位資料進行補遺。由本研究推估結果顯示，相異多筆資料若均為預測值，預測偏差將隨著時間增加而加大，原因為後項預測資料包含前項資料誤差，其所產生之偏差，為灰預測系統具有資料遞增或遞減趨勢所致，因此需搭配相異間距法以減少溫泉水位預測誤差。本研究利用各溫泉監測站之最佳模式，分別完成新小(深)站、新小(淺)站、遊客中心站、福崇寺站及大忠路站，歷?遺?溫泉水位資?，補遺成果對礁溪溫泉水位缺漏現象可趨於完整。 The administration of hot spring in Taiwan has collected and checked the data of hot spring level these years. Unique temperature and quality of the hot spring usually cause the device recorded any missing data. The data analysis used the gray theory in order to supplement the data of hot spring level. This study based on Lih Huey Kang’s observation of using whole data of hot spring level, tested and predicted the best model with gray prediction. The best model helped the fragmented part of data fixed, which arranged to produce the complete set of data. In the data of hot spring level in Jiaosi, the data of Hot Spring Park, Guo-Xiao and Tai-Tzu station fixed the four batches of the fragmented data using the best model estimation.The result show the predicted deviation appeared in these conflicting data, the error of the predicted deviation would increase with time. The reason that caused these errors of posterior is the anterior, and the trend of the deviation increased or decreased in the system of gray prediction. To prevent these errors, divergence distance method was needed. This study used the best model in each administrations of hot spring. The action made the fragmented data of hot spring level, which included Shin Xiao Shen Station、Shin Xiao Chien Station、Tourist Center Station、Fu Chong Station and Da Zhong Rd. Station during 2011~2014 fixed. The result of the supplementation brought down the gap between the actual hot spring levels in Jiaosi.
