了解遊客在觀光景點間的移動行為與景點的拜訪人次，對於改善觀光地區公共運輸服務與觀光管理當局之資源配置具有極大助益。隨著科技發展的日新月異，以及人們使用手機行動上網的普及率提升，在人口移動預測上，手機信令資料有著樣本數量大、涵蓋範圍廣且蒐集成本相對低廉的優點。手機信令資料分析可以有效建構使用者的時空軌跡，進而得到使用者在景點間潛在的移動型態。本研究建立一套有系統的景點拜訪人數預測方法，首先透過分析使用者的信令資料建立旅次鏈與旅次起迄矩陣，作為景點間轉移矩陣推估的基礎，再利用馬可夫鏈預測每小時觀光景點拜訪人次。本研究以花蓮縣作為研究場域，以交通部觀光局推薦的59個主要觀光景點為對象，透過由電信公司取得觀光客的信令資料，評估所提出預測方法的績效。結果顯示預測之平均絕對百分比誤差約為20%，屬於實務上可接受的範圍，表示本研究之預測方法具有不錯的成效。

Understanding the movement of tourists between attractions and the number of visitors at attractions is helpful for the authorities to improve the public transportation services in scenic areas and to allocate resources. With the rapid development of technology and the increasing popularity of people using mobile phones to access the Internet, cellular-based vehicle probe (CVP) data has the advantages of larger sample size, broader coverage and lower collection cost in human mobility prediction. Users’ spatial-temporal trajectories can be effectively constructed by analyzing CVP data. The potential movement pattern of users can be extracted from those trajectories. This study develops a systematic approach to predict the number of visitors at attractions. Firstly, trip chains and origin-destination (OD) matrix are constructed by analyzing users’ CVP data. The OD matrix is used as the basis for estimating the transition matrix between attractions. Then, a Markov Chain model is established to predict the number of tourists at each attraction in each hour. The proposed method is applied to predict the numbers of tourists at 59 major attractions, recommended by the Tourism Bureau, in Hualien county. The CVP data is provided by a major telecom in Taiwan. The mean absolute percentage error of the prediction results is about 20%, which is practically acceptable. The evaluation results indicate that the proposed method has a good prediction performance.