ДОСЛІДЖЕННЯ ВІДКРИТИХ НАБОРІВ ДАНИХ ВЕБ-РЕСУРСІВ У КОНТЕКСТІ ЗАСТОСУВАННЯ ЇХ ДЛЯ ТЕСТУВАННЯ РЕКОМЕНДАЦІЙНИХ СИСТЕМ
Д. В. Шингалов, Ye. Meleshko, V. Bosko · Системи управління навігації та зв’язку Збірник наукових праць · 2019
The subject matter of the article is the process of testing methods of building recommender systems based on open data sets from the Internet. The goal is to research open data sets of web-resources in the context of using them to test various methods of building recommender systems. The tasks to be solved are: to explore modern web-platforms with open data sets and to research the possibility of using their data to test the work quality of various recommender systems. The following results were obtained: The most popular web-platforms with open sets of various network data were considered. The comparative analysis of these platforms was carried out in terms of the availability of free access to downloads of data, their functionality and geographical location, data format and convenience for future use for machine learning, as well as the possibility of using for testing recommender systems. Also, an assessment of the relevance of data stored in repositories with free access and their availability over time was made. Conclusions. Web-platforms containing open data sets that can be used to test recommender systems were explored. The main advantages of most platforms are the support of modern data formats and conditionally free or free access. Among the shortcomings of the considered platforms, it is worth noting the lack of structuredness of some data sets, in particular, text data, which significantly limits their use for testing content-based filtering methods. In addition, one of the factors that limit the use of open data sets is their relevance, since some of the sets stored on the platforms are outdated and not updated. All considered data sets can be applied for research purposes and for testing the work of recommender systems.
Cited by 0 works. doi:10.26906/sunz.2019.4.110 Open-access PDF
Authors
References
- Amazon.com recommendations: item-to-item collaborative filtering
- Collaborative filtering with temporal dynamics
- A survey of active learning in collaborative filtering recommender systems
- Recommender Systems: The Textbook
- Dealing with the new user cold-start problem in recommender systems: A comparative review
- Addressing the Item Cold-Start Problem by Attribute-Driven Active Learning
- Towards the Next Generation of Recommender Systems
- Recommending Hotels based on Multi-Dimensional Customer Ratings
- Implementation of search mechanism for implicit dependences in process mining
- THE METHOD OF CONSTRUCTING RECOMMENDATIONS ONLINE ON THE TEMPORAL DYNAMICS OF USER INTERESTS USING MULTILAYER GRAPH