Receptor models, such as EPA’s CMB 8.2 (Chemical Mass Balance) and PMF 5.0 (Positive Matrix Factorization), have been extensively used in the last two decades to apportion sources based on the measurement of atmospheric pollutant time series. Although useful, these techniques seem to be gradually abandoned in favour of the more versatile Eulerian Chemistry-Transport Models, which consider both the chemical reactions and the spatial component of transport and can be used to investigate the impact of emission abatement measures. Nevertheless, with an increased availability of high temporal resolution data set and the development of new un/supervised machine learning techniques, opportunities arise to renew these data-driven models and their applications. In this study, we try to answer this question: Can a general-purpose clustering algorithm deal adequately with source apportionment receptor data? The considered techniques are compared with an application of PMF to a data set of PM2.5 compounds collected every two days in Engis (Belgium), a suburban industrial site, well known for its 1930 air pollution episode.