Please use this identifier to cite or link to this item: http://repository.ipb.ac.id/handle/123456789/157927
Title: Penggunaan Algoritma Detect Deviating Cells (DDC) untuk Mengatasi Pencilan Sel pada Regresi Partial Least Squares
Other Titles: Application of Detect Deviating Cells (DDC) Algorithm to Handle Cellwise Outliers in Partial Least Squares Regression
Authors: Ardana, Ngakan Komang Kutha
Sumarno, Hadi
Prastiwi, Tania Chandra
Issue Date: 2024
Publisher: IPB University
Abstract: Data multivariat seringkali mengandung pencilan yang dapat mengganggu hasil analisis. Algoritma Detect Deviating Cells (DDC) digunakan untuk mengidentifikasi dan menangani pencilan pada tingkatan sel pada gugus data, menggantinya dengan nilai imputasi (X_imp) sehingga menciptakan gugus data yang disempurnakan untuk regresi Partial Least Squares (PLS) berikutnya. Akurasi model dievaluasi menggunakan Root Mean Square Error (RMSE) dan metrik R-squared adjusted (R^2_adj). Hasil penelitian menunjukkan bahwa algoritma DDC pada regresi PLS menghasilkan model yang lebih akurat dalam menduga nilai sebenarnya, terutama pada data yang mengandung pencilan dan multikolinearitas. Model ini juga mampu menjelaskan variabilitas data yang lebih baik dibandingkan dengan metode regresi Ordinary Least Squares (OLS).
Multivariate data often contains outliers that can interfere the analysis results. The Detect Deviating Cells (DDC) algorithm was used to identify and handle cell-level outliers in the data clusters, replacing them with imputed values (X_imp) thus creating enhanced data clusters for subsequent Partial Least Squares (PLS) regression. Model accuracy was evaluated using Root Mean Square Error (RMSE) and adjusted R-squared metric (R^2_adj). The results show that the DDC algorithm in PLS regression produces a more accurate model in predicting the true value, especially in data containing outliers and multicollinearity. This model is also able to explain data variability better than the Ordinary Least Squares (OLS) regression method.
URI: http://repository.ipb.ac.id/handle/123456789/157927
Appears in Collections:UT - Mathematics

Files in This Item:
File Description SizeFormat 
cover_G5401201003_0cc0cbeb2fd646a19edcbf35d4e1bd0c.pdfCover808.56 kBAdobe PDFView/Open
fulltext_G5401201003_ca8bc751d6be45479fa466c701081b03.pdf
  Restricted Access
Fulltext2.71 MBAdobe PDFView/Open
lampiran_G5401201003_d41b0c4d76134bdb99c4dbf331162b9b.pdf
  Restricted Access
Lampiran936.5 kBAdobe PDFView/Open


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.