Abstract
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.
Highlights
250 million tweets and close to 15 billion messages are written every day [1]
The observed variations in the results suggest that these effects are reasonably stable, indicating the importance of controlling the suggestions presented by an recommender systems (RSs) proportionately to a user's behaviors in regard to diversity or novelty based on the new evaluation rules
Even with highly unstable results when the original RS suggestions are replaced with random suggestions, a pattern of change in the results from one randomness percentage to another can still be identified, especially when the average results are calculated for every level of randomness (Figures 50 and 51)
Summary
250 million tweets and close to 15 billion messages are written every day [1]. Many surveys have been performed in this domain in the last four years [4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19], and they have summarized most recommender system (RS)-related fields and provided good starting points to discover additional advantages and disadvantages. Additional research works are needed to enable further development in the RS field and facilitate the scientific utilization of these databases, for instance, to produce important and useful information. The means of producing such information in the RS field can be summarized by three words: detection, prediction, and suggestion
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
