The Microbial Source Tracking problem (MST) has to do with the determination of the fecal pollution origin in waters by the use of microbial and chemical indicators. This document introduces a methodology for solving MST problem from the machine learning point of view reporting both the arising specifi c problems and challenges and how they have been addressed. The simplest instance of the MST problem has already been solved to satisfaction using machine learning techniques on recently and heavily polluted waters, however, our methodology accepts examples showing di fferent concentration levels and using indicators (variables) with diff erent environmental persistence. The theoretical methodology is supported by a software which implements it and has been validated using two real datasets with real data from diff erent geographical and climatic areas.