停車問題長久以來一直為社會大眾所關切，也是政府機關感到棘手的問題之一，所以解決停車問題，刻不容緩，近年來由於智慧型車路系統的發達，以及用路人資訊日益受到重視，因此對於停車場即時的資訊也顯得格外重要。由於進入停車場的車輛數，具有不同時間與空間而變化的動態特性，故對此現象作有效的管理與控制，必須有效預測未來進入停車場車輛數，以便適時擬定控制策略進行及時管理與控制。 本研究以台北市信義計畫區內之停車場為範圍，先將停車場依旅次目的做分類，最後以凱悅、信義A21、國際會議中心及信義A2四個停車場，作為研究之對象。首先回顧各種交通車流預測模式之特性，以提升模式預測能力。然後在分析不同旅次目的下的進入停車場車輛數變化之情形，試圖引用類神經網路，採用三層、完全連結及前向式的網路架構，配合倒傳遞演算法，建立不同旅次目形態預測模式，利用親自調查所得之停車需求量及連接路段流量資料，作為類神經網路訓練與測試基礎。 經由不斷的校估與測試，由結果得知，本研究所構建之停車需求量預測模式，預測效果良好，誤差小，相關係數高。因此在即時停車需求量預測方面，本研究可提供未來相關單位，預測停車需求量參考之雛形。 Parking problem is one of the main concerns of public for a long time, and the thorny problems of the government. Therefore, to solve parking problem is the most urgent. Owing to the development of Intelligent Transportation Systems (ITS) and user’s information have attracted much interest, real-time information of parking lots are getting more and more important. As to vehicles entering parking lots are changeable with different time and space, it’s necessary to make real-time controlling policies to forecast parking flows efficiently. This study is ranged over parking lots of Hsin-Yi Planning Zone in Taipei City. By travelling goals, it is divided into Hyatt, Hsin-Yi A21, international convention center and Hsin-Yi A2 parking lots to be the objects of study. In order to promote the forecasting ability of model, it reviews kinds of traffic flow forecasting modeling and analyzes entering flows in different purposes, guiding in Artificial Neural Network with three layers, fully connected and feed—forward, and Backpropagation Algorithm to build forecasting models. By investigating of parking demand and connecting glows, it would be the base of training and testing of Artificial Neural Network. After repeatedly correcting and test, this model would forecast effectively with small error and high correlation. As to the result, this thesis can be provided to forecast parking demand in real-time parking demand estimation.
