Data integration aggregation have allowed to provide uniform interface for multiple heterogonous sources, metadata and MicroMetadata (MM); this issue has attracted a large amount of attention from different areas. Hence, the problem of finding which digital resources may belong to a specific interest demands specific research. We proposed a model named LBAM: The Learning & Boosting Architecture Model. This process makes emphasis on matching user evolutive interests and MM. It combines of context, geolocation, utility, group, content-venue, and user persona aware-approaches. It is a hybrid Machine Learning Model (MLM) and Boosting Models (MLBM): content-based MLM for events semantic MM extraction and collaborative filtering MLM. It uses Machine Learning Models to improve the identification of the User Interests according to different media types. Using simulation study and prototypes, we show that LBAM may propose many personal channels representing slightly the User Interests in a context of aware- approaches. We put in place a first prototype. This paper is the third part of LB project using LBAM.
