A Fuzzy Decision Tree for Bus Network Management

The planning and managing of bus networks as a urban public transport system is a relevant problem. Several proposals focus the planning of new bus networks. In this paper, we focus the speci c task of reallocating buses from di erent lines of a previously planned bus network in case of restrictions, such as vehicle breakdowns and drivers absences. Our proposal uses a fuzzy decision tree, which combines the graphical representation and high interpretability of decision trees with the use of linguistic variables of fuzzy systems. The fuzzy decision tree is intended to support human experts that have to make decisions regarding reallocating buses from prede ned lines in order to optimize the average time that users have to wait for a bus service in any bus stop. The model was induced by the FUZZYDT algorithm using real data collected from a public bus system with 26 bus lines. 16 variables related to characteristics of the bus lines and users were considered as inputs to the system. The output variable of the system presents an evaluation of bus lines when changes in the lines occur. The induced fuzzy decision tree includes only 7 key-variables forming 15 rules. The nal model is simple, highly interpretable, and has received positive feedback from experts.