本研究的主要目的在於建立與評估資料採礦分析技術（鑑別分析、羅吉斯迴歸、人工類神經網路）應用於臺灣地區民營羽球館之顧客流失模型，資料來源由作者於2003年調查臺灣地區民營羽球館之顧客資料時所提供，共計2,006筆。而為了驗證模型的適用性，本研究之研究架構分為模型訓練階段與模型測試階段進行，並以80:20的比例隨機抽出訓練模型的樣本及測試模型的樣本個數；在第一階段之模型訓練階段，主要在建立顧客流失模型；而第二階段之模型測試階段，主要是利用先前建立好之模型，來預測未來可能流失顧客之正確率。結果發現，利用人工類神經網路所建構之顧客流失模型，其整體正確判別率為最高(88.78%)，而流失顧客之重要特徵為男性、年齡介於23-35歲、未婚、高中（職）學歷、職業為公教人員、月收入為20,001-40,000元之問、居住在北部區域、球齡為l-5年、每週參與l次、每次持續l小時以下、參與時段為下午、參與夥伴為朋友、到達時問為60分鐘以上、滿意度（行政管理、服務態度、器材設施、場地環境）為中下程度的顧客。

The purpose of this study was to establish and apply data mining technology (discriminant analysis, logistic regression analysis, artificial neural networks) to evaluate the churn model of Taiwanese private badminton courts. The database, totally 2,006 records, was provided by Fang-yang Lu. In order to verify the applicability to the churn model, the research construction of this study was broken down into two stages: the model training stage model and model testing stage. The samples of the training model and testing model were extracted randomly in the proportion of 80:20. The customer churn model was built in the model training stage. In the model testing stage, the correct rate of future churn among customers was predicted by the model. The result found that whole correct classification rate reached the peak of highest of 88.78% in this churn model using artificial neural networks. The significant characteristics of churn customer were male, single, 23-25 years old, high school degree, government agent, salary of 20-40k per month, living in the north area of Taiwan, having participated for 1-5 years, spending less than an hour every time per week, playing sports in the afternoon, joining a partner or friends, arriving time above 60 minutes, customers’ satisfactions (administration, management, service attitude, equipment, and environment) of below average.