Modelo de Predição de Conforto de Usuários do Transporte Coletivo
The small and medium-size cities are facing problems related to mobilitythat could be avoided by adopting the public transportationsystem, as buses and trains. However, in many Brazilian cities theuse of public transportation is neglected because it is considereduncomfortable, expensive and insecure. To attract passengers forsuch kind of transportation there are several possible approaches,the promotion of comfort perception is one of those. Several studieshave already approached this problem, however, few of themaddressed the perception of comfort felt by the passengers usingtelemetry data collected from the vehicle. Among the works thatuse such data, none of them applied data mining techniques toabstract a general model of comfort perception. Therefore, thiswork aims to apply mining techniques over telemetry data collectedfrom vehicles to build a comprehensible model to classifythe level of comfort of public transportation passengers. To achievethis objective machine learning techniques were used, centered ondecision trees. Due to the complexity of abstracting the model therewere constructed three models, one for each acceleration axis thatwere merged using a meta-classifier responsible to point out thepassenger general comfort. The results have reached an accuracyof 85,2%, which can be considered a promising result regardingthe difficulties of separating the data source in sets that can betteridentify the bus drivers behaviour.