COMPARATIVE ANALYSIS OF VARIOUS MULTIPLE LINEAR REGRESSION METHODS AND HIERARCHICAL CLUSTERING ON DISAGGREGATE INDICATOR OF LIVABILITY INDEX (LI) DATA IN BANDUNG CITY

• Main Author: Khoiriyah Muwahidah, Imroatul
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/47749
• Institution: Institut Teknologi Bandung

Bandung is one of the cities in Indonesia which has a high density of population. Therefore, various problems began to arise from various sectors with different characteristic in Bandung. In 2017 the Bandung City Government conducted a survey to determine the Livability Index (LI). However, there are so many indicators that affect LI that it must take a long time to collect data about these indicators. So, in this final project a simpler LI model was made. A comparison between various methods of multiple linear regression includes the forward selection, backward elimination, and stepwise regression methods is used. Finally, the best regression model with a 92.3% model fit was generated. From this model, 6 indicators out of 27 indicators that most influence LI were obtained. The 6 indicators are the average time to go to the place of activity, the percentage of people connected to water, the percentage of people who have proper sanitation, availability of facilities, water quality, and affordability of health costs. In this study, a hierarchical cluster analysis was also conducted to classify regions in Bandung based on the 6 previous indicators. This hierarchical cluster analysis is done by comparing several distance analysis methods including single linkage, average linkage, and complete linkage. Finally, the best results are obtained by using the average linkage which divides sub-districts in Bandung into four clusters. That four clusters have different characteristics which will be used as recommendation to improve Livability Index (LI).