Extraction of user profiles by discovering preferences through machine learning

The recent evolution of e-commerce has emphasized the need for services to be suitable to the needs of individual users: as a consequence, personalization has become an important strategy to improve access to relevant products. This work presents a personalization process based on a text categorization method, which exploits the textual descriptions of the products in online catalogues, in order to discriminate between interesting and uninteresting items for the customer. Experimental results encourage the integration of the method in the personalization component we developed in the COGITO project(1).