Adaptive weighting with SMOTE for learning from imbalanced datasets: A case study for traffic offence prediction
This paper proposes to augment the prediction capability of a classifier or of an ensemble of classifiers for an imbalanced set using a combination of informed sampling based on SMOTE (Synthetic Minority Oversampling Technique) and a post-classification adaptive weighting that takes into account a priori knowledge about a dataset. As a case study, the paper analyzes the relationship between traffic tickets (provincial offence notices), their types and the trends in attributes such as vehicle type, offence type, location, ticket status for the city of Ottawa, Canada with the purpose of enabling a proactive traffic enforcement.
|Keywords||Data imbalance, Ensemble models, Machine learning, Prediction, SMOTE, Traffic offences|
|Conference||23rd Annual IEEE International Conference on Computational Intelligence and Virtual Environments for Measurement Systems and Applications, CIVEMSA 2018|
Bobbili, N.P. (Naga Prasanthi), & Cretu, A.M. (2018). Adaptive weighting with SMOTE for learning from imbalanced datasets: A case study for traffic offence prediction. In CIVEMSA 2018 - 2018 IEEE International Conference on Computational Intelligence and Virtual Environments for Measurement Systems and Applications, Proceedings. doi:10.1109/CIVEMSA.2018.8439957