학술논문
Spatial-Temporal Long Short-Term Memory (ST-LSTM) Modelling for Flight Departure Delay Prediction: Preliminary study
Document Type
Conference
Source
한국정보통신학회 종합학술대회 논문집. 2019-10 23(2):265-268
Subject
Language
Korean
ISSN
2092-8017
2671-8022
2671-8022
Abstract
항공기 지연은 운송 산업에서 가장 영향력 있는 성과 결정 요인 중 하나이다. 그러나 기존의 기계학습 (Machine Learning) 방법은 시간적/공간적 특징을 가지는 항공 지연을 예측하는 데 적합하지 않다. 본 연구는 미국 항공편 데이터에서 추출한 시공간 특징을 LSTM (Long Short-Term Memory)과 결합하여 데이터의 숨겨진 패턴을 알아내고, 항공 출발 지연을 예측하는 새로운 모델을 제안한다. LSTM 기반 아키텍처는 높은 수준의 시공간 구조를 갖는 시퀀스 및 데이터를 모델링 할 수 있다.
Flight delay is one of the most influential performance determinants in the transport Industry. Traditional Machine Learning methods are inadequate for the task of predicting flight delays. In this paper, we present a new predictive model using Spatial-temporal features extracted from the United States flights data, combined with Long Short-Term Memory (LSTM) to unveil hidden patterns in the data and predict flight departure delays. LSTM architectures are remarkably capable of modeling sequential data that has a high-level spatial-temporal structure.
Flight delay is one of the most influential performance determinants in the transport Industry. Traditional Machine Learning methods are inadequate for the task of predicting flight delays. In this paper, we present a new predictive model using Spatial-temporal features extracted from the United States flights data, combined with Long Short-Term Memory (LSTM) to unveil hidden patterns in the data and predict flight departure delays. LSTM architectures are remarkably capable of modeling sequential data that has a high-level spatial-temporal structure.