当前位置: X-MOL 学术IISE Trans. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Copula-based multi-event modeling and prediction using fleet service records
IISE Transactions ( IF 2.6 ) Pub Date : 2020-09-21 , DOI: 10.1080/24725854.2020.1802792
Akash Deep 1 , Shiyu Zhou 1 , Dharmaraj Veeramani 1
Affiliation  

Abstract

Recent advances in information and communication technology are enabling availability of event sequence data from equipment fleets comprising potentially a large number of similar units. The data from a specific unit may be related to multiple types of events, such as occurrence of different types of failures, and are recorded as part of the unit’s service history. In this article, we present a novel method for modeling and prediction of such event sequences using fleet service records. The proposed method uses copula to approximate the joint distribution of time-to-event variables corresponding to each type of event. The marginal distributions of the time-to-event variables that are needed for the copula function are obtained through Cox Proportional Hazard (PH) regression models. Our method is flexible and efficient in modeling the relationships among multiple events, and overcomes limitations of traditional approaches, such as Cox PH. With simulations and a real-world case study, we demonstrate that the proposed method outperforms the base regression model in prediction accuracy of future event occurrences.



中文翻译:

使用车队服务记录的基于 Copula 的多事件建模和预测

摘要

信息和通信技术的最新进展使得能够从包括潜在大量类似单元的设备组中获得事件序列数据。来自特定单元的数据可能与多种类型的事件有关,例如不同类型故障的发生,并记录为单元服务历史的一部分。在本文中,我们提出了一种使用车队服务记录对此类事件序列进行建模和预测的新方法。所提出的方法使用 copula 来近似对应于每种类型的事件的时间到事件变量的联合分布。通过 Cox Proportional Hazard (PH) 回归模型获得 copula 函数所需的时间到事件变量的边际分布。我们的方法在建模多个事件之间的关系方面灵活高效,并克服了传统方法(例如 Cox PH)的局限性。通过模拟和真实案例研究,我们证明所提出的方法在未来事件发生的预测准确性方面优于基本回归模型。

更新日期:2020-09-21
down
wechat
bug