ERGONOMIC ANALYSIS OF TEMPORAL CHARACTERISTICS OF RECOMMENDATION FORMATION IN INFORMATION PROCESSING SYSTEMS

Goryachkin B. S., Andreev V. A., Semkin P. S.

Problem statement. In the modern information space, the key parameter of the quality of "human-machine" interaction is the system response time. Delays in content generation in recommender systems directly affect user satisfaction and cognitive load. Goal. To conduct an ergonomic analysis of the temporal characteristics of the recommendation formation process and identify the dependence of execution time on the architectural features of algorithms and the hardware platform. Results. Mathematical models of recommendation formation time have been developed for three approaches: User-based Collaborative Filtering, Item-based Collaborative Filtering, and Matrix Factorization (SVD). The physical meaning of coefficients taking into account equipment characteristics is defined. The advantage of the matrix factorization method has been experimentally established. Practical significance. The proposed calculation methodology and the findings allow predicting the performance of recommender systems and selecting the optimal algorithm to ensure ergonomic interaction standards in real time.

Keywords:Recommender systems, response time, ergonomics, User-based CF, Item-based CF, SVD, mathematical modeling, performance