多元化計程車機場接送服務動態派車接機時間預測模型

A Dynamic Forecasting Model for Diversification Taxi Pickup Passengers in Taoyuan Airport

隨著政府開放計程車多元化營業的政策，多元化計程車經營策略的妥善規劃，將有助於計程車業者的獲利，而多元化計程車的經營策略，以無障礙計程車、觀光計程車及定點特殊接送計程車的經營最受業者重視，而機場接送服務則是定點特殊接送計程車業者重要的營運市場。然而就目前機場入境乘車空間的限制，僅能提供接送車輛暫停上下客的服務，不許可長時間的停滯等候，因此當旅客選擇多元化計程車做為機場接送的服務運具時，如何能準時的到達機場接送旅客，便成為重要的研究課題。本研究的進行，將建立多元化計程車機場接送服務動態派車接機時間預測模型，利用倒傳遞類神經網路，預測航班抵達至旅客搭乘接駁運具的時間。本研究共建立單隱藏層與雙隱藏層神經網路模型並進行預測經測試雙層隱藏層之倒傳遞神經網路模型預測步行時間誤差在 5 分鐘內之比例均在 60%以上。此外並與 MATLAB所附的類神經網路進行數值測試，由結果發現，本研究發展之模型較為準確與穩定，藉由本模型的發展，預期將可提供多元化計程車業者進行機場接送時，能準確的預測旅客到達入境旅客候車處的時間，以提高業者的服務水準，並增加產業的競爭力。

With the opening diversification policy for taxi, the diversification of taxi business strategy will create the benefit of Taxi operator. The business is focused on barrier-free taxi, tourism taxi, and fixed-point special shuttle taxi. The airport passengers pick-up service is one of the most important business for diversification taxi service. Due to the waiting area space restrictions for pick-up vehicle on airport, when passengers choose a diversified taxi as an airport shuttle service, how the airport can avoid the airport congestion problem is the critical issue. In this paper, the back-propagation neural network is used to construct the prediction model. This model effectively predicts the time between the flights arrives and passengers arrive take the feeder, expecting to reduce the vehicle's stay at the airport and avoid crowding. In this study, the double hidden layer neural network model were established and predicted. The results show the error of prediction in 5 minutes of the proportion of is more than 60%.The model of this research is more accurate and stable than MATLAB. With the development of this model, it could enhance the service performance of Taoyuan International Airport and increase the competitiveness of the national tourism industry.