在航空運輸業中，速度就是成本。準時交貨更是航空貨運營運被視為服務優勢的重要關鍵要素，貨物若無法在承諾的交貨時間內抵達，可能會衍生延誤成本。本研究旨在提供航空貨運延誤實時預測，資料來源為國內指標性航空貨運承載資料與全球起降航點歷史天氣資料，以貨物相關特徵，包含件數、體積、重量、特殊處理註記等，結合班機起飛、降落航點天氣狀況，藉由探究可能導致無法準時交貨的因素，進而建立一個結合貨物與天氣特徵之全新貨運準時度預測模型。從實驗發現，類神經網路因具有自我調適與處理非線性問題的特性，對於異質性大且無一定規則可循的航空貨運，預測準確率確實優於傳統分類演算法。本研究期望藉由貨運準時度預測，預告貨物可能面臨延遲抵達的情況，進而調整更有利的運送決策，將有助於提高航空運輸管理質量，也加深航空運輸業者與客戶之間的黏著度，經由長期信任而建立客戶滿意度，為航空運輸業者帶來競爭優勢。;In the air transportation industry, speed is cost. In air cargo operations, on-time delivery is regarded as an important key element of service advantages. If the delivery fails to arrive within the promised delivery time, delay costs may be incurred. The purpose of this research is to establish a predictive delivery model by exploring the factors that may lead to on-time delivery by exploring the relevant characteristics of the cargo and the weather conditions at the flight′s take-off/landing. The data sources are the domestic index air cargo carrying database and the historical weather data of global take-offs and landings. After selecting data mining features, a neural network prediction model is built through deep learning of large amounts of data. From experiments, it is found that, due to the characteristics of self-adjustment, the prediction accuracy of neural networks is indeed better than traditional classification algorithms for air cargo with large heterogeneity and no certain rules to follow. The punctuality predicted by the models is used to inform shippers in advance they may face the problem of delayed arrival of goods and then adjusting more favorable shipping decisions With the function of early warning, the closeness between airlines and customers can be improved. Moreover, airlines can gain a competitive advantage through the long-trusted customer relationships.
