DOI: https://doi.org/10.32515/2409-9392.2018.31.181-186

The Mathematical Model of the Recommendation System, Taking Into Account the Emotional Coloring of the Comments as a Context

Dmitry Shyngalov, Yelyzaveta Meleshko, Roman Mynaylenko, Vitaliy Reznichenko

About the Authors

Dmitry Shyngalov, postgraduate, Central Ukranian National Technical University, Kropyvnytskyi, Ukraine, E-mail:dimashingalov@gmail.com

Yelyzaveta Meleshko, Associate Professor, PhD in Technics (Candidate of Technics Sciences), Central Ukranian National Technical University, Kropyvnytskyi, Ukraine, E-mail:elismeleshko@gmail.com

Roman Mynaylenko, Associate Professor, PhD in Technics (Candidate of Technics Sciences), Central Ukranian National Technical University, Kropyvnytskyi, Ukraine

Vitaliy Reznichenko, lecture, Central Ukranian National Technical University, Kropyvnytskyi, Ukraine

Abstract

The article proposes the mathematical model of a recommendation system, in which a sentiment-analysis of comments related to a objects of recommendations is used as the context. An attempt is made to draw the mathematical model based on matrix calculations for the further creation of a context-oriented advisory system. In addition, an analysis of the feasibility of using the factorization model method to contextual filtration is used. The paper investigates the structure of the recommendation system, in which the analysis of the emotional color of the comments to the objects of the recommendation is taken into account as a context. In the absence of explicit feedback, context analysis greatly improves the accuracy of the recommendations and the quality of prediction of user benefits. An attempt is also to deduce a mathematical model describing the work of such a recommendation system. Row down algorithms and the use of factoring machines for data processing are considered in data conditions without explicit response. Further practical application of hybrid joint filtration is considered as a solution to the practical implementation of the system of recommendations. Rating models have huge potential for solving common filtration tasks, as well as for other subject areas in which there are interactions between objects of different types. Also, the proposed method allows you to change the structure of the model by changing the spaces of the space. The advantage of adding attributes can be applied to real data when modeling user preferences based on the analysis of the comment text. Applying an analysis of the emotional color of the text of comments by social network users on recommended objects can greatly improve the quality of prediction recommendations, and the transition to affine transformations and matrix calculations simplifies the machine complexity of calculations and increases their speed.

Keywords

recommendation systems, sentiment-analysis, сollaborative filtration, machine learning, intelligent systems

Full Text:

PDF

References

Bal´azs, H. & Domonkos, T. (2012). Fast als-based tensor factorization for context-aware recommendation from implicit feedback. 67–82.

Ricci, Editors F., Rokach, L., Shapira, B. & Kantor, P. (2011). Recommender systems handbook . Berlin: Springer.

Kolda, T. & Bader, B. (2009). Tensor decompositions and applications. SIAM review, №51(3), 455-500.

Zhouxiao, B. & Haiying, X. (2012). Movie Rating Estimation and Recommendation.

Cantador, I., Bellogn, A. & Vallet, D. (2010). Content-based recommendation in social tagging systems. ACM RecSys, 10, 237–240.

Burke, R. (2002). Hybrid recommender systems: survey and experiments. User Modelling and UserAdapted Interaction, vol. 12, 4, 331–370.

Stuart, D. (2010). What are Libraries Doing on Twitter? Online 34, 1, 45–47.

Mitchell, T. (1997). Machine Learning. New York: McGraw-Hill.

Narayanan, V., Arora, I. & Bhatia, A. (2013). Fast and accurate sentiment classification using an enhanced naive bayes model. Intelligent Data Engineering and Automated Learning IDEAL, Vol. 8206, 194–201.

Norden, A. P. (1976). Prostranstva affinnoj svjaznosti [Affine Connected Spaces]. Moscow: Nauka.

GOST Style Citations

  1. Bal´azs H. Fast als-based tensor factorization for context-aware recommendation from implicit feedback [Text] / H. Bal´azs, T. Domonkos // Machine Learning and Knowledge Discovery in Databases – 2012. – pp. 67–82.
  2. Recommender systems handbook [Text] / Editors F.Ricci, L. Rokach, B. Shapira, P. Kantor. – Berlin: Springer, 2011. – 842 pages.
  3. Kolda T. Tensor decompositions and applications [Text] / T. Kolda, B. Bader. // SIAM review. – 2009. – №51(3). – pp. 455–500.
  4. Zhouxiao B. Movie Rating Estimation and Recommendation [Text] / B. Zhouxiao, X. Haiying. – 2012. – pp. 1–4.
  5. Cantador I. Content-based recommendation in social tagging systems. [Text] / I. Cantador, A. Bellogn, D. Vallet. // ACM RecSys. – 2010. – №10. – pp. 237–240.
  6. Burke R. Hybrid recommender systems: survey and experiments [Text] / R. Burke. // User Modelling and UserAdapted Interaction, vol. 12, no. 4. – 2002. – pp. 331–370.
  7. Stuart D. What are Libraries Doing on Twitter? [Text] / D. Stuart. // Online 34, no. 1. – 2010. – pp. 45–47.
  8. Mitchell T. Machine Learning [Text] / T. Mitchell. – New York: McGraw-Hill, 1997. – 414 p.
  9. Narayanan V. Fast and accurate sentiment classification using an enhanced naive bayes model. [Text] / V. Narayanan, I. Arora, A. Bhatia // Intelligent Data Engineering and Automated Learning IDEAL – Berlin: Springer, 2013. – (volume 8206 of Lecture Notes in Computer Science). – pp. 194–201.
  10. Норден А. П. Пространства аффинной связности [Tекст] / А. П. Норден. – М.: Наука, 1976. – 432 стр.
Copyright (c) 2018 Dmitry Shyngalov, Yelyzaveta Meleshko, Roman Mynaylenko, Vitaliy Reznichenko