Unit:
Κατεύθυνση Διαχείριση Δεδομένων, Πληροφορίας και ΓνώσηςΠληροφορική
Author:
Koliopoulou Maria-Myrto
Supervisors info:
Σταματόπουλος Παναγιώτης, Επίκουρος Καθηγητής στο Τμήμα Πληροφορικής και Τηλεπικοινωνιών του Πανεπιστημίου Αθηνών
Original Title:
Clustering in Recommendation Systems Using Swarm Intelligence
Translated title:
Clustering in Recommendation Systems Using Swarm Intelligence
Summary:
A recommender system (RS) is an application that exploits information to help users in decision making by suggesting items they might like. A collaborative recommender system generates recommendations to users based on their similar neighbor’s preferences. However, this type of recommender system faces the data sparsity and scalability problems making the neighborhood selection a challenging task. This thesis proposes three hybrid collaborative recommender systems that each one combines the k-means algorithm with a different bio-inspired technique to enhance the clustering task, and therefore to improve the recommendation quality. The used bio-inspired techniques are artificial bee colony (ABC), cuckoo search optimization (CSO), and grey-wolf optimizer (GWO). The proposed approaches were evaluated over a MovieLens dataset. The evaluation shows that the proposed recommender systems perform better compared to already existing techniques in terms of mean absolute error (MAE), precision, sum of squared errors (SSE), and recall. Moreover, the experimental results indicate that the hybrid recommender system that uses the ABC method performs slightly better than the other two proposed hybrid algorithms.
Main subject category:
Technology - Computer science
Keywords:
clustering, swarm intelligence, collaborative filtering, k-means, artificial bee colony, recommender systems
