Yu.E. Gapanyuk − Ph.D. (Computer Sciences), Associate Professor,

Computer Science and Control Systems Department, Bauman Moscow State Technical University

E-mail: gapyu@bmstu.ru

A.S. Zenger − Student,

Computer Science and Control Systems Department, Bauman Moscow State Technical University

E-mail: azenger98@gmail.com

A.K. Cvetkova − Student,

Computer Science and Control Systems Department, Bauman Moscow State Technical University

E-mail: alenkin98@yandex.ru

S.A. Kochkin − Student,

Computer Science and Control Systems Department, Bauman Moscow State Technical University

E-mail: sergei.kochkinn@gmail.com

V.V. Cherkov − Student,

Computer Science and Control Systems Department, Bauman Moscow State Technical University

E-mail: vv-ch@bk.ru

Recommender systems are referred to as “information filtering systems” whose main task is to remove unwanted, irrelevant or redundant information from the user's area of visibility.

There are two main strategies for creating recommendation systems: content-based filtering and collaborative filtering, as well as various ways to hybridize these two options. For content-based filtering, the user's area of interest is determined by the domain objects that the user created, selected, or requested.

Two main approaches to the implementation of collaborative filtering can be distinguished: “user-based” and “item-based.” In the case of the “user-based” approach, the users who are most similar in their estimations to the current user (for whom the recommendation is generated) are identified. According to the estimations of the found users, estimations are predicted for the current user. The “item-based” approach is similar to the “user-based” approach, but in this case, the estimation is generated for the object rather than for the user. The objects that are most similar in their estimates to the current object (the estimate of which we want to predict) are determined. According to the estimates of the found objects, the estimate for the current object is predicted.

Currently, the following versions of Hybrid Recommender Systems (HRS) are used: “weighted” HRS, “branched” HRS, “cascade” HRS or “pipeline,” “multi-level” HRS. These approaches are mainly reminiscent of ensemble models of machine learning.

The approach based on Hybrid Intelligent Information Systems (HIIS) allows you to systematize various options for building recommendation systems. In most cases, the recommendation system based on the concept of HIIS is a holonic structure with multiple nesting options for the modules of consciousness and subconsciousness. Since recommendation systems are more likely to use machine learning methods than rule-based methods, in most cases, the recommendation system modules belong to the subconscious of the HIIS.

In the experimental part, based on the considered approaches, a recommender system of culinary recipes is implemented. The best result was shown by a hybrid version of the recommender system.

1. Ricci F., Rokach L., Shapira B. Recommender Systems Handbook (2nd. ed.). Springer. 2015.
2. Aggarwal C.C. Recommender Systems: The Textbook. Springer. 2016.
3. Lampropoulos A.S., Tsihrintzis G.A. Machine Learning Paradigms: Applications in Recommender Systems. Intelligent Systems Reference Library. V. 92. Springer. 2015.
4. Kolesnikov A.V. Gibridnye intellektual'nye sistemy. Teorija i tehnologija razrabotki. SPb: SPbGTU. 2001. 137 s. (In Russian).
5. Kolesnikov A.V., Kirikov I.A., Listopad S.V. Gibridnye intellektual'nye sistemy s samoorganizaciej: koordinacija, soglasovannost', spor. M.: IPI RAN. 2014. 189 s. (In Russian).
6. Chernen'kij V.M., Terehov V.I., Gapanjuk Ju.E. Struktura gibridnoj intellektual'noj informacionnoj sistemy na osnove metagrafov. Nejrokomp'jutery: razrabotka, primenenie. 2016. № 9. S. 3-14 (In Russian).
7. Chernen'kij V.M., Gapanjuk Ju.E., Revunkov G.I., Terehov V.I., Kaganov Ju.T. Metagrafovyj podhod dlja opisanija gibridnyh intellektual'nyh informacionnyh system. Prikladnaja informatika. 2017. T. 12. № 3(69). S. 57-79 (In Russian).
8. Prikladnye intellektual'nye sistemy, osnovannye na mjagkih vychislenijah. Pod red. N.G. Jarushkinoj. Ul'janovsk: UlGTU. 2004. 139 s. (In Russian).
9. Tarasov V.B. Ot mnogoagentnyh sistem k intellektual'nym organizacijam: filosofija, psihologija, informatika. M.: Jeditorial URSS. 2002. 352 s. (In Russian).
10. Epicurious - Recipes with Rating and Nutrition. URL: https://www.kaggle.com/hugodarwood/epirecipes (data obrashhenija: 29.05.2020).
11. Geron A. Hands-on machine learning with Scikit-Learn and TensorFlow: concepts, tools, and techniques to build intelligent systems. O'Reilly Media. 2017.