Enhancing the Recommender System Algorithm's Quality Through the Utilization of Affinity Analysis

The demand for auxiliary systems that utilize methods and technologies of intellectual data analysis and machine learning is increasing in the CRM system development field. These systems are capable of generating valuable insights from vast amounts of data collected in CRM. The article shows the results of the development and study of the CRM system using the methods of affinity analysis. Earlier, the authors developed and implemented the base version of the recommender algorithm based on the use of cluster data analysis and collaborative filtration [1],[2]. The new version of the algorithm additionally uses affinity analysis methods to form recommendations for the selection of products, which made it possible to increase the accuracy of the recommendation system (service) by metrics F2 on average from 67.98% to 81.24% with an increase in the time for issuing recommendations (On average by 2.47 ms). The study and comparison of the basic and modified versions of the algorithm were executed using data sourced from insurance companies, furnished by "FB Consult".