Millora dels processos a través de recomanació d’algorismes d'experience-based

Abstract

With the recent increase in Internet use (among other data sources) there is a corresponding growth of the total information available to users. An important problem associated to this overload problem is that it is becoming more and more difficult for users to locate specific pieces of information. Therefore it is necessary to have some kind of help to guide users in finding exactly what they seek among this large volume of information. Recommender systems manage large volumes of data and information to support users in the search for relevant and specific search outcomes. There are many types of recommender systems, which can be broadly classified into two main categories, each used for a different purpose. These are content-based and collaborative filtering recommender systems. This project focuses on the first group of systems, those based on content. Specifically, we focus on conversational recommender systems concentrating on user critique-based feedback. The main objectives of the project are to develop a study of the different recommender systems; to implement, on the library’s recommendation research group Volume Visualization and Artificial Intelligence (WAI) developed in Java, experinced-based algorithms such as History-Aware Critiquing (HAC), Experience-based Critiquing (EBC) and Graph-based algorithms; and to perform a comparative analysis of every approach implemented in order to evaluate the improvement over the standard critiquing algorithms (IC and Std). The results obtained form this comparison show that experience-based algorithms performance is the best within current recommendation algorithms and they have remarkably improved upon the standard critiquing-based approaches.