Transportation choice modeling on commuters in Jabodetabek using Bayesian network and polytomous logistic regression

摘要

Urban sprawl phenomenon that occurred in DKI Jakarta - Indonesia led to the expansion of the surrounding areas. Urban sprawl is a process of physical exposure to the outer city physical appearance caused by the rapid economic development and population growth. The “Bodetabek” area has turned into a densely populated residence and even an industrial estate. The compromise makes citizens of Bodetabek and DKI Jakarta leave their house to carry out their activities outside the administrative area of their residence and back on that day. People who do this kind of migration are called commuters. Their routine movement is the main cause of traffic jam. The transition of commuter transport modes from private vehicles to public transport is a solution to reduce congestion. However, not all commuters are interested in using public transportation. The results from the 2014 Commuter's Survey show that there is only 27 percent of Jabodetabek commuters use public transportation to go to their destination. Bayesian Network (BN) and logistic regression are proposed to model this kind of transportation choice in Jabodetabek. Logistic regression is widely used in classification modeling. While BN, including Naïve Bayes (NB) and Hierarchical Naïve Bayes (HNB), is a capable model of explaining the structure of relationships between complex random variables into diagrammatic forms based on conditional probability theory. The results show that Polytomous Logistic Regression has the highest Correct Classification Rate (CCR) and Area Under (a ROC) Curve (AUC). The Polytomous Logistic Regression, however, need more computational time consuming than NB and HNB.