SAMPLID: A new supervised approach for meaningful place identification using call detail records as an alternative to classical unsupervised clustering techniques

Abstract

SAMPLID introduces a supervised approach for meaningful place identification based on a problem-specific knowledge base. Using call detail records from Milan, the method outperforms classical unsupervised clustering and performs best when spatially related neighbouring cells are represented in the training data. The approach provides a direct path from mobile-network activity to interpretable home and work classifications and establishes the basis for later multi-label extensions.

Citation

Mendoza-Hurtado, M., Romero-del-Castillo, J. A., & Ortiz-Boyer, D. (2024). SAMPLID: A new supervised approach for meaningful place identification using call detail records as an alternative to classical unsupervised clustering techniques. ISPRS International Journal of Geo-Information, 13(8), Article 289. https://doi.org/10.3390/ijgi13080289