Divide and recombine approach for warranty database: estimating the reliability of an automobile component  

作　　者：Md Rezaul Karim 

机构地区：[1]Department of Statistics,University of Rajshahi,Rajshahi,6205,Bangladesh

出　　处：《Data Science and Management》2024年第2期119-128,共10页数据科学与管理（英文）

摘　　要：The continuously updated database of failures and censored data of numerous products has become large, and on some covariates, information regarding the failure times is missing in the database. As the dataset is large and has missing information, the analysis tasks become complicated and a long time is required to execute the programming codes. In such situations, the divide and recombine (D&R) approach, which has a practical computational performance for big data analysis, can be applied. In this study, the D&R approach was applied to analyze the real field data of an automobile component with incomplete information on covariates using the Weibull regression model. Model parameters were estimated using the expectation maximization algorithm. The results of the data analysis and simulation demonstrated that the D&R approach is applicable for analyzing such datasets. Further, the percentiles and reliability functions of the distribution under different covariate conditions were estimated to evaluate the component performance of these covariates. The findings of this study have managerial implications regarding design decisions, safety, and reliability of automobile components.

关 键 词：Weibull regression model Warranty database RELIABILITY EM algorithm Divide and recombine approach Managerial implications 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]