DEVELOPMENT OF METHOD TO ANALYZE HUMAN MOBILITY PATTERN BASED ON GPS DATA FOR URBAN PLANNING

• Main Author: (NIM: 23516085), STEPHEN
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/31092
• Institution: Institut Teknologi Bandung

<p align="justify">Urban and city planning is one of the efforts that government need to do in order to ensure the prosperity of its people. One of the thing that the government can do so they can make a better urban and city planning is by studying, learning, and analyzing human mobility in their respective area. Human mobility can describe lots of things e.g. an uneven growth of economic, or uneven number of education institution between rural area and center of the cit. From that background, we can conclude that human mobility is very important for urban and city planning so there is a necessity for government to analyze it, and one of the way to analyze it is by predicting the mobility pattern that is going to happen. Three main aspects that we need to know from a mobility is from where it will begin, to where, and finally by using which trajectory. In this research, the best method to do those process will be found and a simple visualization will be built to show the prediction result. For predicting the area that will be the beginning of a trip, neural network regression will be used to predict which areas that will have the most people that are going to start their trip because it is suspected to be better in learning non-linearity, then for predicting the destination areas, Naïve Bayes classification will be used, and lastly for predicting trajectory, the transfer probability model will be used in this research with an additional time feature since it is ignored in the previous research. For the start area prediction, the R2 value the we managed to obtain was 0.66 which is slightly better compared to previous research using the same data. For predicting the destination area, a quite high accuracy was obtained which was 0.82 because of low number of labels that were being used, and finally for predicting the trajectory, the additional time feature can give a better information about the route because it can show that 2 trips that consist of same departure and destination area doesn't always use the same trajectory. A simple visualization is made to show the results of those 3 predictions and can be used by the government to analyze human mobility pattern which can later be used to make a better urban planning.<p align="justify">