Solving Cold-Start Problem by Combining Personality Traits and Demographic Attributes in a User Based Recommender System

Abstract

Abstract-Several approaches have been suggested for providing users with recommendations using their rating history, most of these approaches suffer from new user problem (cold-start) which is the initial lack of items rating. Most hybrid approaches use the combination of the rating based similarity measure with demographic filtering to solve the cold-start problem. In this paper we combined the rating based similarity calculation with the personality traits based similarity and demographic filtering. We integrated personality traits with the demographic characteristics because users with the same demographic attributes may have different personality or behavior so personality information is a necessary filtering condition for users to find better reference users to be recommendation guides. Even though users don't have a rating history the proposed enhanced similarity measure will find users with the same demographic attributes like age, gender, location…etc and the same personality traits. This helps to find the k-nearest neighbors for new users and solve cold-start problem which also can improve the performance of the recommender system.

Clustering-Based Frequent Pattern Mining Framework for Solving Cold-Start Problem in Recommender Systems

“…Specifically, it concerns the issue that the recommender system cannot draw any inferences for new users, who signed up recently to the system, or new items, which added recently to the system, since it has not yet gathered sufficient information to generate reliable recommendations. Therefore, the RS (and especially the collaborative methods) suffers from both user cold-start and item cold-start problems due to deficient information about new entities [13], [14].…”

“…Although CF-based methods are very efficient and give good results, they still suffer from the cold-start problem which occurs whenever a recommender system tries to generate recommendations for either a new user who signed up recently to the system without having any rating records available yet or when a new item is added to the system without any rating given to that item so far. In fact, most state-of-the-art recommendation algorithms generate unreliable recommendations for such cases since they cannot learn the preference embedding of these new users/items [13], [55].…”

Enhanced Collaborative Filtering Algorithm for Movies Recommendation using Big Five Personality Traits

“…In this section, several collaborative filtering systems were reviewed by highlighting the contributions, strength and weaknesses of each research work as follows: [1] Proposes Solving Cold Start Problem by Combining Personality Traits and Demographic Attributes in a User Based Recommender System. The system fails to examine that people living in same demographic area tend to have different inferences, interest and personality therefore, their choices on item is quite different.…”

“…With Cold Start problem, there will be no accurate recommendation for new user. This prompted many research to mitigate the cold start problem [1], [2], [6]. For example [1], combined personality traits and demographic attributes in a user based recommender system to solve cold start but people living in same demographic area have different inferences, interest and personality therefore, their choices on item is quite different.…”

“…This prompted many research to mitigate the cold start problem [1], [2], [6]. For example [1], combined personality traits and demographic attributes in a user based recommender system to solve cold start but people living in same demographic area have different inferences, interest and personality therefore, their choices on item is quite different. [2], proposes an approach that apply the relationship of user feature scores derived from user item interaction via ratings to optimize the prediction algorithms.…”