The progressive increaseand seemingly unstoppable accumulation of marine macro-litter on the bottom of the Mediterranean Sea is an urgent problem that needs accurate information and guidance to identify those areas most at risk of accumulation. In the absence of dedicated monitoring programs, an important source of opportunistic data are fishery-independent monitoring campaigns of demersal resources. These data have long been used but not yet extensively. In this paper, MEDiterranean International Trawl Survey (MEDITS) data was supplemented with 18 layers of information related to major environmental (e.g. depth, sea water and wind velocity, sea waves) and anthropogenic (e.g. river inputsimpact of river, shipping lanes, urban areas and ports, fishing effort) forcings that influence seafloor macro-litter distribution. The Random Forest (RF), a machine learning approach, was applied to: i) model the distribution of several litter categories at a high spatial resolution (i.e. 1 km2); ii) identify major accumulation hot spots and their temporal trends. Results indicate that RF is a very effective approach to model the distribution of marine macro-litter and provides a consistent picture of the heterogeneous distribution of different macro-litter categories. The most critical situation in the study area was observed in the north-eastern part of the western basin. In addition, the combined analysis of weight and density data identified a tendency for lighter items to accumulate in areas (such as the northern part of the Tyrrhenian Sea) with more stagnant currents. This approach, based on georeferenced information widely available in public databases, seems a natural candidate to be applied in other basins as a support and complement tool to field monitoring activities and strategies for protection and remediation of the most impacted areas.