隨著台灣電信用戶市場趨於飽和化態勢以及競爭激烈所致，各家電信業者莫不努力推動優惠專案以追求提高用戶人數，以及提供優惠續約專案來避免舊用戶攜碼轉移流失的情況，與用戶建立良好互關係以確保用戶忠誠度。本研究目的在於建立電信用戶流失預測模型，研究對象為金門地區民眾使用行動通訊網路服務之用戶，希望能探討出影響電信用戶流失之影響因素，藉由研究模型分析了解電信用戶流失率、手機品牌忠誠度的因果關係。本研究將以類神經網路為架構，建構電信用戶流失與手機品牌忠誠度預測模型，實驗分析運用多層感知機方法進行分析比較，以探討金門地區民眾對電信用戶流失與手機品牌忠誠度的預測情況。經由多層感知機方法分析研究結果發現：一、金門地區電信用戶流失比率8.2%，流失預測模型準確率達到92.1%；二、金門地區用戶手機品牌忠誠度比率10.46%，手機品牌忠誠度預測模型達89.4；三、流失與手機品牌忠誠度兩個變量合併分析模型準確率達到90.8%；本研究的預測結果經實驗證明，本研究之預測模型準確率在九成以上，適用於電信行業用戶流失與手機品牌忠誠度預測。

As Taiwan’s telecommunications user market is becoming saturated and the competition is fierce, various telecommunications companies have worked hard to promote preferential programs in pursuit of increasing the number of users, and provide preferential renewal programs to avoid the loss of old users with code transfer. Users establish a good relationship to ensure user loyalty. The purpose of this research is to establish a telecommunication user churn prediction model. The research object is users in Kinmen who use mobile communication network services. It is hoped to explore the influencing factors of telecommunication user churn, and to analyze and understand the telecommunication user churn rate and mobile phone brand through the research model. The causality of loyalty. This study will build a prediction model of telecommunications user churn and mobile phone brand loyalty based on a neural network. The experimental analysis uses a multi-layer perceptron method to analyze and compare, to explore the prediction of the churn of telecommunications users and mobile phone brand loyalty among people in Kinmen. Happening. Through the analysis of the multi-layer perceptron method, the results found that: 1. The churn rate of telecommunications users in Kinmen area was 8.2%, and the accuracy rate of the churn prediction model reached 92.1%; 2. The mobile phone brand loyalty ratio of users in Kinmen area was 10.46%, and the mobile phone brand loyalty prediction model reached 89.4; 3. The accuracy rate of the analysis model combining the two variables of churn and mobile phone brand loyalty reached 90.8%; the prediction results of this study have been experimentally proved that the accuracy rate of the prediction model of this study is more than 90%, which is suitable for user churn and mobile phone brand loyalty. Mobile phone brand loyalty prediction.