Ponizovkin Denis Mikhailovch (Programmer, IT-Aces)
| |
The subject of the article is recommender systems. The main functionality of these systems is to help their users to do quick search for actual and useful information on the base of recommendations provided by a system. This task is difficult but actual in modern conditions, when a huge amount of information is available on the Internet and different devices. There are different techniques applicable in recommender systems in order to implement the target functionality. This work considers one of the most popular and successful techniques – collaborative filtration. This technique is filtration based on connections, and could be made either by users or system objects. We will consider filtration by objects. With the help of this technique the task of definition of a subset of objects, similar in characteristics to user preferences, with cardinality of N is solving. The task solution is qualitatively found if the defined by the system subset consists of N objects similar to the user. In order to implement filtration, recommender systems calculate function values named similarity measures. If the similarity measure value of items is greater than some threshold value, then there is similarity relation between these items and these items do not filtered. We propose sufficient condition of efficient solving of the topN task for collaborative systems. This condition is the transitivity property of the similarity relation. Accomplishment of the sufficient condition depends on a choice of the similarity measure and threshold value. We propose the method of the modeling of collaborative systems for which sufficient condition is accomplished.
Keywords:collaborative filtreing, recommender systems, topN task, similarity measure, transitivity of similarity relation
|
|
| |
|
Read the full article …
|
Citation link:
Ponizovkin D. M. Accuracy improvement of task for collaborative filtering recommender systems // Современная наука: актуальные проблемы теории и практики. Серия: Естественные и Технические Науки. -2017. -№07-08. -С. 67-70 |
|
|
LEGAL
INFORMATION:
