چکیده:
One of the important consequences of the expansion of the internet in the present era is the emergence of e-commerce websites; however, the diversity of items offered can turn the selection of suitable products into a complex process for customers. Companies recommend using recommender systems to solve this problem. Due to the high percentage of errors in refining and presenting suggestions, several basic methods are usually used in these systems to suggest items of interest to the user. Collaborative filtering is one of the most successful methods used in these systems; but using this method with an increasing number of users and products faces problems such as inability to respond to the needs of new users and data sparsity. In order to solve the mentioned problem, a new method is introduced in this research that, by integrating user time-rate with Pearson similarity metric and also integrating semantic technologies and social network, provides a solution to reduce the problem of new users and data sparsity. The results of algorithm implementation show that the presented approach has better performance and higher accuracy and its predictions are more consistent with user preferences.
خلاصه ماشینی:
In order to solve the aforementioned problem, a new method is introduced in this research that, by integrating user rating time with Pearson similarity measure and also integrating semantic technologies and social networks, provides a solution to reduce the problem of new users and data sparsity.
The user-based method proposed in this research uses the effect of rating time on user interest over time to improve dealing with the problem of data sparsity and trust relationships between users in social networks, in order to better diagnose user interests and consequently improve recommendations.
The innovation of this research is that by integrating user rating time with the Pearson similarity criterion to reduce data sparsity and integrating semantic technologies and social networks, the problems of “new users” and in general “data sparsity” in recommender systems are solved.
One of the findings of these researches is that temporal information can improve the accuracy of recommenders in collaborative filtering-based recommender systems for features described in the mobile e-commerce environment; But in the reviewed articles, the unit of time and its effect on similarity criteria and the effect of semantic relationships in social networks have not been used, which this article focuses on.
In the first method, the time index with the Pearson correlation coefficient, which is based on the memory of collaborative filtering recommender systems (in this section, the times of user ratings to items are also used), are integrated.
(
, 
, 
)
