Generating Artificial Data to Evaluate Machine Learning Predictive Algorithms for Bus Travel Time
This paper proposes a simulator capable of quickly generating a large amount of data that may be used to train bus travel time predictive algorithms in an urban transport network. To validate the proposal, a case study was carried out on a bus line in the city of Brasília/DF, Brazil. In the case study, the Simulator generated data for several scenarios that differ in distinct levels of variability and these data were used to evaluate the performance of machine learning predictors in each of the scenarios. Moreover, we provided experimental evaluation of several machine learning algorithms and we compared their performance to our proposed method based in Linear Combination of predictors. The mean absolute error was adopted in these experiments to evaluate the quality of predictors results, and our proposed linear combiner approach was able to improve the performance of the prediction in both real and artificial data.