Please use this identifier to cite or link to this item:
Title: Efficient approach for Social Recommendations using Graphs on Neo4j
Authors: Kaur, Ashamdeep
Supervisor: Rani, Rinkle
Keywords: Cold start, Sparsity, Trust, Social Recommendation
Issue Date: 24-Jul-2018
Abstract: Social networks are developing in number and size, with a huge number of client accounts and enormous amount of data. Explosive growth of data and digital information on internet has created a potential challenge for visitors to get coherent information which further hinders the user to access proper items of his interest on internet timely. . Recommendation Systems check the client's inclinations for proposing components to purchase or browse. They have turned out to be an essential applications in internet business and access to data that gives proposals that successfully diminish extensive data to the things that best address user’s issues and inclinations. Still sometimes we face a cold start problem and we do not get accurate recommendations. Another problem which is faced by basic recommender is sparsity problem where there is not enough data to extract recommendations. We have proposed an extra favourable position of these systems in which collaborative filtering is used plus clients can encode more data about their relations than essentially say who they trust. Trust is the feature in which one user can explicitly state his trust on any user whose choice he likes. Graphs are used to apply recommendations in the proposed approach. The graph database is an effective tool for handling relationships between entities of data model. Neo4j is used as graph database to implement the proposed algorithm as with the help of Neo4j we can manipulate graphs accordingly. Along with trust, also proposed different feature is transitivity in social networks through which we can get more accurate recommendations. Manipulated graph is called influenced graph in the whole research work. We have achieved sufficient results which prove this approach of transitivity between nodes is helpful for better recommendations.
Appears in Collections:Masters Theses@CSED

Files in This Item:
File Description SizeFormat 
Ashamdeep04.pdf1.85 MBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.