#TITLE_ALTERNATIVE#

• Main Author: RANI WIDYANING PANGESTUTI (NIM 15107071); Pembimbing: Dr. Andri Hernandi, ST, MSP dan Dr. Ir. B, DIAH
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/14117
• Institution: Institut Teknologi Bandung

The regression model is often applied to various kinds of modelling. However, the problems are often encountered in regression model is the correlation between observed variables that called multicollinearity. In statistic theory, regression model is said to be good if it satisfies the classical assumptions which one of them is no multicollinearity. Most researchers overcome the multicollinearity problem by eliminating one or more observed variables. However, this method is considered unfavorable because it could be omitted variables proved to have a major contribution to the creation of the model. In this study, multicollinearity problems solved by using Principal Component Analysis (PCA). This technique was chosen because it has the advantages which one of them is does not have to reduce the number of observed variables. With the PCA technique, initial variables are reduced into one or more new variables called principal component. The initial land value model that still contains multicollinearity is stated in equation HT = 9844386,383 + 686,834 LT – 28145,881 LD + 430099,058 LJ – 592,624 JAT – 1815,500 JKT – 5228,834 JLP – 1180,781 JAB – 714875,031 ZPK – 1705557,537 AK + 757307,047 DR. After using PCA techniques, a new land value model that has been free from multicollinearity is stated in equation HT = 3243740,394 – 1143906,503 F1 + 718696,229 F2 + 359146,609 F3 + 287429,307 F4. Based on some model fit tests is done, the new model can be considered good because it meets the requirements for a good model.