|
Publication
|
Research Title |
AN ALTERNATIVE METHOD TO DETECT OUTLIERS IN MULTIVARIATE DATA |
Date of Distribution |
29 November 2022 |
Conference |
Title of the Conference |
48th INTERNATIONAL CONGRESS ON SCIENCE TECHNOLOGY AND TECHNOLOGY-BASED INNOVATION |
Organiser |
สมาคมวิทยาศาสตร์แห่งประเทศไทยในพระบรมราชูปถัมภ์, Walailak University |
Conference Place |
Walailak University |
Province/State |
นครศรีธรรมราช |
Conference Date |
29 November 2022 |
To |
1 December 2022 |
Proceeding Paper |
Volume |
48 |
Issue |
1 |
Page |
299-307 |
Editors/edition/publisher |
สมาคมวิทยาศาสตร์แห่งประเทศไทยในพระบรมราชูปถัมภ์ |
Abstract |
This study proposes techniques for finding outliers in multivariate data. It is based on the Mahalanobis distance and multiple linear regression. The Mahalanobis distance is used to filter the data across all variables to break the data set into two groups, i.e., normal data and data that might be outliers. After that, a multiple linear regression model is built using the normal data to provide a reliable estimate for the cut-off point. For figuring out how well the proposed method works, a simulation study is done with multivariate normal data with and without contaminated data at different levels: 0, 0.01, 0.05, and 0.10. The performance of the proposed method is compared with the earlier methods; Mahalanobis distance and Mahalanobis distance with the robust estimators using the minimum volume ellipsoid method, the minimum covariance determinant method, and the minimum vector variance method. The findings demonstrated that the proposed method effectively identifies the highest accuracy of outliers detected at all levels of contamination regardless of sample size or the number of variables. When the proposed method is used on real data, it also shows that it can find outlier values that are consistent with the real data. |
Author |
|
Peer Review Status |
มีผู้ประเมินอิสระ |
Level of Conference |
นานาชาติ |
Type of Proceeding |
Full paper |
Type of Presentation |
Oral |
Part of thesis |
true |
Presentation awarding |
false |
Attach file |
|
Citation |
0
|
|
|
|
|
|
|