APPLICATION OF PREDICTIVE ANALYTICS FOR PASSENGER LOAD FACTOR ESTIMATION IN CITY BUS SERVICES

• Main Author: Astari, Pradnya
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/43687
• Institution: Institut Teknologi Bandung

Users demand for city bus service is increasing in various cities and countries. With the increasing number of city bus service users, this can be a consideration in increasing the quality of service. One of the aspects to measure the quality of public transportation services is vehicle occupancy or load factor. This research conducts the application of predictive analytics for estimating passenger load factors in city bus services by analyzing bus user transaction data and bus availability data. The methodology used for modeling is CRISP-DM, until the Evaluation phase. Clustering used in this research is k-means clustering. The predictive modeling of the number of passengers is done by using Support Vector Regression algorithm. Modeling uses historical data on the number of passenger trips for 3 months for training data and the month data for test data. The results of the passenger load factor estimation modeling are dashboards to determine the load per month for each segment, the value of the passenger load factor, and the recommended number of fleets needed per hour. Testing the passenger load prediction model is done by using the RMSE and MAPE values. The kernel that produces the best test error value is Linear. The highest estimated passenger load factor is found in peak hour. The simulation results of the calculation of passenger load factor estimates indicate the need for additional fleets at peak hour, and at off-peak is appropriate because it has approached number 1. The simulation results for the recommended number of fleets is that there is a constant increase in demand at peak hours for each segment, and at off -peak hour increase needs only in certain segments.