Main Page
Deanship
The Dean
Dean's Word
Curriculum Vitae
Contact the Dean
Vision and Mission
Organizational Structure
Vice- Deanship
Vice- Dean
KAU Graduate Studies
Research Services & Courses
Research Services Unit
Important Research for Society
Deanship's Services
FAQs
Research
Staff Directory
Files
Favorite Websites
Deanship Access Map
Graduate Studies Awards
Deanship's Staff
Staff Directory
Files
Researches
Contact us
عربي
English
About
Admission
Academic
Research and Innovations
University Life
E-Services
Search
Deanship of Graduate Studies
Document Details
Document Type
:
Thesis
Document Title
:
APPLYING BEHAVIOURAL TARGETING TO SOLVE COLD START PROBLEM IN RECOMMENDER SYSTEMS
تطبيق استهداف السلوك لحل مشكلة البداية الباردة للمستخدم في نظام التوصية
Subject
:
Faculty of Computing and Information Technology
Document Language
:
Arabic
Abstract
:
Recommender systems are widely used in e-commerce to save users’ time and help them make more satisfying decisions. E-commerce can provide their users with better-personalized suggestions that may suit their preferences. However, the cold start problem is an inherent challenge in recommender systems that faces e-commerce when users are new to the system, and they have limited interaction experience with the system. To address the cold start problem and provide different solutions, analyzing different users’ behavior during sessions spent on the business website needs more investigations. We solve the user cold start problem by applying the behavioral targeting technique to study the users’ pattern of clicks on items and categories without the knowing explicit feedback such as ratings or written reviews. Behavioral targeting is a technique used to match suitable advertisements to the right user on a commercial website. In this research, a Collaborative Filtering technique is used to identify similar users in the context of their clicks’ behaviors. We have applied the collaborative Filtering technique using the K-nearest neighbor model to classify users based on their clicks. To conduct the study, we used a real dataset from an online bookstore containing users’ behavior such as clicks on category and purchases. The K-nearest neighbor were predicted with K=5 and for evaluation the predictions result we used the classification report metrics: accuracy, recall, precision, and the F1-score along with the confusion metrics. Results showed that studying user clicks can provide meaningful information about user interest, and can apply to user collaborative filtering approach to find user recommendations. The model achieved an accuracy of 34%, which is consider low accuracy rate for a recommender system. However, for user cold start problem, it is acceptable to suggest recommendations that are 34% accurate to new user who have
Supervisor
:
Dr. Saleh Binti Abdullah
Thesis Type
:
Master Thesis
Publishing Year
:
1442 AH
2020 AD
Added Date
:
Tuesday, January 5, 2021
Researchers
Researcher Name (Arabic)
Researcher Name (English)
Researcher Type
Dr Grade
Email
نورة خلف المطيري
Almutairi, Nora Khalaf
Researcher
Master
Files
File Name
Type
Description
46834.pdf
pdf
Back To Researches Page