GR Semicolon EN

Show simple item record

dc.contributor.author
Paraskevopoulos, Athanasios
en
dc.date.accessioned
2018-05-05T11:51:06Z
dc.date.available
2018-05-06T00:00:17Z
dc.date.issued
2018-05-05
dc.identifier.uri
https://repository.ihu.edu.gr//xmlui/handle/11544/29088
dc.rights
Default License
dc.subject
Recommender systems
en
dc.title
Product Recommendation System
en
heal.type
masterThesis
en_US
heal.classification
Recommender systems
en
heal.classification
Cold start
en
heal.classification
Collaborative filtering
en
heal.classification
Machine Learning
en
heal.keywordURI.LCSH
Recommender systems (Information filtering)
heal.keywordURI.LCSH
Personal communication service systems
heal.keywordURI.LCSH
Machine learning
heal.language
en
en_US
heal.access
free
en_US
heal.license
http://creativecommons.org/licenses/by-nc/4.0
en_US
heal.recordProvider
School of Science and Technology, MSc in Mobile and Web Computing
en_US
heal.publicationDate
2018-05-05
heal.abstract
This dissertation was written as a part of the MSc in Mobile and Web Computing at the International Hellenic University. The aim of this work is to investigate if user’s demographics data and ratings entropy0 scores can have an impact on addressing the cold start problem in collaborative filtering. We propose four collaborative movie recommender systems that use ask - to - rate techniques by displaying movies to users for rating. The implementation of the aforementioned systems was done in Python 3.6 programming language, developing four independent scripts that display movies for rating using different ask - to - rate techniques: random choice of movies, demographic based, entropy0 based , mix of demographic and entropy0 based. In the evaluation we have taken into consideration both the accuracy of the predictions but also the user effort. The results have shown that there is (almost) a tie for the first place between demographic - based and entropy0 - based systems both in terms of user preference score but also in terms of user’s effort (entropy0 based system is only margin ally better) . Furthermore, we can also see that the system with the combination of demographics and entropy0, is slightly better (in terms of user preference score ) than the basic (random selection), even if the user effort is much higher. Finally, for a future work we can use a movie - set with newer movies or a completely different dataset with another type of products like electronic devices, books etc. Moreover, a mobile implementation can make recommender systems even more useful and also valuable.
en
heal.advisorName
Tjortjis, Christos
el
heal.committeeMemberName
Tjortjis, Christos
en
heal.committeeMemberName
Berberidis, Christos
en
heal.committeeMemberName
Gatzianas, Marios
en
heal.academicPublisher
IHU
en
heal.academicPublisherID
ihu
en_US
heal.numberOfPages
108
en_US


This item appears in the following Collection(s)

Show simple item record

Related Items