Multiclass Classification with Imbalanced Datasets for Car Ownership Demand Model – Cost-Sensitive Learning

In terms of the travel demand prediction from household car ownership model, if imbalanced data were used to support transportation policy via a machine learning it would negatively affect algorithm training process. The on obtained study project for expressway preparation in Khon Kaen Province (2015) was an unbalanced dataset. other words, number members minority class is lower than rest answer classes. result bias classification. Consequently, this research suggested balancing datasets with cost-sensitive methods, including decision trees, k-nearest neighbors (kNN), and naive Bayes algorithms. Before creating 3-class k-folds cross-validation method applied classify define true positive rate (TPR) model’s performance validation. outcome indicated that kNN demonstrated best compared It provides TPR rural suburban area types, which are region types very different imbalance ratios, before 46.9% 46.4%. After (MCN1), values 84.4% 81.4%, respectively.