Um ensemble baseado em árvores de decisão para predizer a ocorrência de aglomerados de ônibus.

Abstract

Travel delays and bus overcrowding are some of the daily dissatisfactions of public transportation users. These problems may be caused by bus bunching, an event that occurs when two or more buses are running the same route together, i.e., arriving at the same time at the bus stops. Due to the stochastic nature of the traffic, a static schedule is not effective to avoid the occurrence of these events; thus, preventive actions are necessary to improve the reliabil ity of the public transportation system. The works already proposed in the predictive context of bus bunching still have limitations related to the frequency or privacy of the data used, in addition to the effectiveness limited to specific contexts. Therefore, we propose a decision tree-based ensemble model to predict bus bunching. We use an ensemble of Random Forest, XGBoost and CatBoost models with buses geolocation, scheduled, weather and traffic situation data. In addition, an incremental learning technique was incorporated into the proposed model to continuously update it according new data arrives in real-time. The efficacy of the proposed model has been demonstrated using real data sets of two brazilian cities and has been compared with four competitors: Linear Regression, Logistic Regression, Support Vector Machine and Relevance Vector Machine. According to the results, the proposed model can achieve an efficacy between 73% − 80%, higher than the evaluated competitors models, and can be used to predict bus bunching in real-time up to ten stops before their occurrence.