Development of Bus Routes Reorganization Support Software Using the Naïve Bayes Classification Method

• Published in: Sustainability 2022-04, Vol.14 (8), p.4400
• Main Authors: Suh, Min-ho, Jeong, Minjoong
• Format: Article
• Language: English
• Subjects: Algorithms; Artificial intelligence; Bayesian analysis; Buses (vehicles); Business metrics; Classification; Decision making; Efficiency; Genetic algorithms; Learning algorithms; Local government; Machine learning; Methods; Passengers; Planning; Prediction models; Route optimization; Software; Support vector machines; Travel demand
• ISSN: 2071-1050; 2071-1050
• DOI: 10.3390/su14084400

Reorganizing city bus routes is generally accomplished by designing bus supply methods to meet passenger demand. The bus supply method involves establishing bus routes and planning their schedules. The actual bus route reorganization decisions are not determined simply by balancing passenger demand and bus supply, but are based on other complex interests, such as bus routes that must exist for welfare but where profits are low. Machine learned prediction models could be helpful when considering such factors in the decision-making process. Here, the Naïve Bayes algorithm was applied to develop the classifier model because of its applicability, even with a limited amount of training data. As the input characteristics for the Naïve Bayes algorithm, data for each individual bus route were featured and cleansed with the actual route improvement decisions. A number of classification models were created by changing training sets and then compared in terms of classification performance such as accuracy, precision, and recall. Modeling and tests were conducted to show how Naïve Bayes classifiers learned in the form of supervised learning can help the route reorganization work. Results from a local governments’ actual route reorganization study were used to train and test the proposed machine learning classification model. As the main contribution of this study, a prediction model was developed to support shortening decision-making for each route, using machine learning algorithms and actual route reorganization research case data. Results verified that such an automatic classifier, or initial route decision proposal software, can provide intuitive support in actual route reorganization research.