Experimental performance evaluation of using multiset metrics in information retrieval problems

Abstract

The problem of the comparative experimental performance evaluation using multiset metrics in information retrieval problems is considered in the paper. The main purpose of the studies is to prove on the basis of actual experimental data, the feasibility of using multisets, as a fundamentally new mathematical tool, in information retrieval problems. As a result of the preliminary statistical analysis of the available data set ‘’Anonymouswebdatafromwww.microsoft.com’’, specific features and problems of information search in the Internet space are identified. As the main performance indicator of information retrieval the “half-life of usefulness”, which predicts the user the usefulness of a list of recommended facilities, taking into account the user’s partial view of a list of recommendations, is used. The calculated values of the performance indicator of information retrieval are compared with the results of similar studies on the effectiveness of traditional, alternative methods of information retrieval. It is shown that the use of metric multisets in information retrieval problems improves the quality of information retrieval