Avaliação de técnicas de aumento de dados para trajetórias

Abstract

Currently, there is a large amount of data being generated every day, however, this data is not obtained uniformly, causing an imbalance problem. The same occurs with vehicle trajectory data, in which traffic sensors do not capture the passage of objects in motion with the same frequency, generating unbalanced data. In addition to the fact that the sensors are positioned at fixed locations, which does not allow the complete tracking of moving objects to be captured, generating sparse trajectories. These problems can hinder the performance of the prediction models of next location. With this, the present work proposes data augmentation techniques for trajectories in order to reduce the problem of data sparsity and imbalance with the use of resampling techniques. And yet, create several predictive models to assess whether there is a positive association between the application of these techniques and the performance of machine learning algorithms.