Abstract
This study tests eleven distance metrics with four multilabel k-nearest-neighbour classifiers on weekly profiles of incoming calls, outgoing calls and internet activity from 241 manually labelled cells in Trentino. It uses four other classifiers as baselines. Canberra-family distances perform well in several classifier and activity combinations. For incoming calls, ML-localkNN reaches an average precision of 0.905 and a ranking loss of 0.125. The results also show that complexity correction can help or hurt depending on the classifier and activity data.
Citation
Mendoza-Hurtado, M., Cuevas-Muñoz, J. M., & Ortiz-Boyer, D. (2026). A Comparative Analysis of Distance Metrics for Multilabel Land-Use Classification Using Mobile Phone Data in Trento. Applied Sciences, 16(21), Article 11040. https://doi.org/10.3390/app162111040