Please use this identifier to cite or link to this item: https://dl.ucsc.cmb.ac.lk/jspui/handle/123456789/3688
Title: Content Based Cross-Domain Recommendation Using Linked Open Data
Authors: Munasinghe, P.P.
Issue Date: 8-Sep-2016
Abstract: Recommender systems can be leveraged to ease the comfort level of people as they presumably lter and provide information that is relevant to individual users and can be classi ed in to two approaches; Content based recommender systems that recommend items based on the similarity the new item's features re ect when compared to user's past preferred items' features, and collaborative ltering systems that recommend the user a set of items that are preferred by similar users. In this thesis we present a Content Based Cross-Domain Recommendation System using Linked Open Data to address the issue of cold-start problem from which a majority of recommender systems su er. Not being able to predict items to a new user due to not having access to his previous preferences, and not being able to recommend a new item to users due to not having any prior ratings on the individual item are the two cold-start problems. Even though content based recommender systems are immune to item cold-start problem, they are comparatively less used due to lack of up-to-date data sources that provide item features and also due to high amount of pre-processing required when using existing data sources for retrieving meta-data. We used Linked Open Data sources to retrieve item features and used a Vector Space Model based approach to calculate the similarities amongst the source domain items with the target domain items. We have shown that the Cross-Domain recommendation can be used to solve the cold-start problem when the source and target domain attributes overlap. We observed that the pre-processing e ort of recommender systems is reduced due to leveraging Linked Open Data which provides a universally structured data set. Reasonable performance in the proposed recommender system was found while generating the recommendation list. The obtained results imply that the prevailing issue of content based recommender systems which forced them to take the backseat will no longer be applicable when Linked Open Data is used and also the cold-start situation in recommender systems can actually be tackled by leveraging a Content based Cross-Domain Recommendation approach.
URI: http://hdl.handle.net/123456789/3688
Appears in Collections:SCS Individual Project - Final Thesis (2015)

Files in This Item:
File Description SizeFormat 
2011CS098.pdf
  Restricted Access
1.24 MBAdobe PDFView/Open Request a copy


Items in UCSC Digital Library are protected by copyright, with all rights reserved, unless otherwise indicated.