Kshour M, Ebrahimi M, Goliaee S, Tawil R
Faculty of New Sciences and Technologies, University of Tehran, Tehran, Iran.
Faculty of Sciences, Lebanese University, Beirut, Lebanon.
Heliyon. 2021 Jun 27;7(7):e07397. doi: 10.1016/j.heliyon.2021.e07397. eCollection 2021 Jul.
Currently, due to the increasing importance of recommender systems (RSs), especially in the fields of social networking and e-commerce, these systems represent one of the most interesting subjects in computer programming. Although many research reports have previously been published in this subject area, because of lack of clarity regarding their algorithms or limited comparisons with the literature, most of them are difficult to extend for similar applications in the future. Therefore, in the present study, we have attempted to improve two novel RS evaluation measures (variety and newness) developed from previous evaluator rules (namely, diversity and novelty) based on human behavior so as to be more reliable and compatible with various developments in RSs. The new rules provide higher weighting for suggestions and respect for users' behavior and can be used in place of diversity and novelty rules with better precision and centralization, by 22.54% for variety and by 14.84% for newness. In addition, we aim to use the developed measures to improve new RSs and support better comparative analyses in this field in the future. This contribution is expected to facilitate better RS research and competition, especially in the social networking domain.
目前，由于推荐系统（RSs）的重要性日益增加，尤其是在社交网络和电子商务领域，这些系统已成为计算机编程中最有趣的主题之一。尽管此前已有许多关于该主题领域的研究报告发表，但由于其算法不够清晰或与文献的比较有限，其中大多数报告在未来难以扩展用于类似应用。因此，在本研究中，我们试图改进基于人类行为从先前评估规则（即多样性和新颖性）发展而来的两种新颖的RS评估度量（多样性和新颖性），使其更可靠并与RSs的各种发展相兼容。新规则为建议提供了更高的权重，并尊重用户行为，并且可以以更高的精度和集中度取代多样性和新颖性规则，多样性提高了22.54%，新颖性提高了14.84%。此外，我们旨在利用所开发的度量来改进新的RSs，并在未来支持该领域更好的比较分析。这一贡献有望促进更好的RS研究和竞争，尤其是在社交网络领域。
