Murdoch University Research Repository

Welcome to the Murdoch University Research Repository

The Murdoch University Research Repository is an open access digital collection of research
created by Murdoch University staff, researchers and postgraduate students.

Learn more

An integrated mobile content recommendation system

Paireekreng, Worapat (2012) An integrated mobile content recommendation system. PhD thesis, Murdoch University.

PDF - Front Pages
Download (409kB)
PDF - Whole Thesis
Download (2MB)


Many features have been added to mobile devices to assist the user's information consumption. However, there are limitations due to information overload on the devices, hardware usability and capacity. As a result, content filtering in a mobile recommendation system plays a vital role in the solution to this problem. A system that utilises content filtering can recommend content which matches a user's needs based on user preferences with a higher accuracy rate.

However, mobile content recommendation systems have problems and limitations related to cold start and sparsity. The problems can be viewed as first time connection and first content rating for non-interactive recommendation systems where information is insufficient to predict mobile content which will match with a user's needs. In addition, how to find relevant items for the content recommendation system which are related to a user's profile is also a concern.

An integrated model that combines the user group identification and mobile content filtering for mobile content recommendation was proposed in this study in order to address the current limitations of the mobile content recommendation system. The model enhances the system by finding the relevant content items that match with a user's needs based on the user's profile. A prototype of the client-side user profile modelling is also developed to demonstrate the concept.

The integrated model applies clustering techniques to determine groups of users. The content filtering implemented classification techniques to predict the top content items. After that, an adaptive association rules technique was performed to find relevant content items. These approaches can help to build the integrated model.

Experimental results have demonstrated that the proposed integrated model performs better than the comparable techniques such as association rules and collaborative filtering. These techniques have been used in several recommendation systems. The integrated model performed better in terms of finding relevant content items which obtained higher accuracy rate of content prediction and predicted successful recommended relevant content measured by recommendation metrics. The model also performed better in terms of rules generation and content recommendation generation. Verification of the proposed model was based on real world practical data. A prototype mobile content recommendation system with client-side user profile has been developed to handle the revisiting user issue. In addition, context information, such as time-of-day and time-of-week, could also be used to enhance the system by recommending the related content to users during different time periods.

Finally, it was shown that the proposed method implemented fewer rules to generate recommendation for mobile content users and it took less processing time. This seems to overcome the problems of first time connection and first content rating for non-interactive recommendation systems.

Item Type: Thesis (PhD)
Murdoch Affiliation(s): School of Information Technology
Supervisor(s): Wong, Kevin and Fung, Lance
Item Control Page Item Control Page


Downloads per month over past year