Monotonic Change Point Estimation in Multistage Profiles

Please enable javascript in your browser.

Sepasi, Shabnam | 2021

420 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 54134 (01)
University: Sharif University of Technology
Department: Industrial Engineering
Advisor(s): Khedmati, Majid
Abstract:
In this thesis, a hybrid method is proposed to estimate the change point in the parameters of simple linear profiles in multistage processes under monotonic changes. In monotonic changes, the type of change is not known a priori, and the only assumption is the changes are of non-decreasing (isotonic) or non-increasing (monotonic) type. In the proposed method, at first, the stages and the parameters experiencing the change are identified and then, the changes occurred in these stages and parameters are identified and examined based on the moving window approach and support vector machine (SVM) algorithm. Finally, the maximum likelihood estimator of the change point is proposed. The performance of the proposed method is evaluated through simulation experiments. The results indicate the high-accuracy performance of the proposed support vector machine algorithm in recognizing control chart patterns and the maximum likelihood method in estimating single and multiple change points
Keywords:
Cascade Property ; Support Vector Machine (SVM) ; Multistage Process ; Moving Window ; Monotonic Change

No TOC