The Machine Learning Algorithm for Solving the Problem of Generating Recommendations for Goods and Services

The article proposes an unsupervised machine learning algorithm for assessing the most possible relationship between two elements of a set of customers and goods / services in order to build a recommendation system. Methods based on collaborative filtering and content-based filtering are considered. A combined algorithm for identifying relationships on sets has been developed, which combines the advantages of the analyzed approaches. The complexity of the algorithm is estimated. Recommendations are given on the efficient implementation of the algorithm in order to reduce the amount of memory used. Using the book recommendation problem as an example, the application of this combined algorithm is shown. This algorithm can be used for a “cold start” of a recommender system, when there are no labeled quality samples of training more complex models.

1. Melville P., Mooney R., Nagarajan R. Content-Boosted Collaborative Filtering for Improved Recommendations. University of Texas, USA. Proceeding of AAAI-02, Austin, TX, USA, 2002. – 2002. – pp. 187–192.
2. Jannach D., Zanker M., Felfering A., Friedrich G., Recommender Systems: An Introduction. Cambridge University Press, 2010.
3. Ricci F., Rokach L., Shapira B., Kantor P. Recommender Systems: Handbook. Springer, 2011.
4. Linden G., Smith B., York J., Amazon.com recommendations: item-toitem collaborative filtering. Internet Computing – IEEE 7 2003 – pp. 76–80.
5. Melville P., Mooney R.J., Nagarajan R. Content-boosted collaborative filtering for improved recommendations. Proceedings of the National Conference on Artificial Intelligence – 2002 – pp. 187–192.
6. Belova K.M., Sudakov V.A. Issledovanie effektivnosti metodov ocenki relevantnosti tekstov [Research of the effectiveness of methods for assessing the relevance of texts]. Preprinty IPM im. M.V. Keldysha = Keldysh Institute preprints, 2020. No 68. 16 p. http://doi.org/10.20948/ prepr-2020–68. (In Russ.).