Improving Context Aware Recommendation Performance by Using Social Networks

Improving Context Aware Recommendation Performance by Using Social Networks

Golshan Assadat Afzali Boroujeni (Department of Computer Engineering and IT, Amirkabir University of Technology, Tehran, Iran) and Seyed Alireza Hashemi Golpayegani (Department of Computer Engineering and IT, Amirkabir University of Technology, Tehran, Iran)
Copyright: © 2014 |Pages: 14
DOI: 10.4018/jitr.2014070101
OnDemand PDF Download:
$30.00
List Price: $37.50

Abstract

Ecommerce systems employ recommender systems to enhance the customer loyalty and hence increasing the cross-selling of products. In collaborative filtering—as the most popular method in recommender systems—an implicit network is formed among all the people. In any network, there are some individuals who have some inspirational power over the others leading them to influence their decisions and behaviours. But it seems that these methods do not support context awareness in mobile commerce environments. Furthermore, they lack high accuracy and also require high volume of computations due to not distinguish between neighbours as a friend or a stranger. This paper proposes a new model for recommender systems which are based on mobile data. This model uses these data to extract current users' context and also to identify individuals with the highest influence. Then, the system uses the information of these identified impressive users in the current context existed in the social networks for making recommendations. Beside of achieving higher accuracy, the proposed model has resolved cold start problem in collaborative filtering systems.
Article Preview

Introduction

In a mobile commerce environment, most of the users are equipped with mobile devices such as cell phone, smart phone, and handhelds. Since everybody has her devices with her almost everywhere, these devices can be used to obtain more detailed and reliable tracking of user’s location and activities. This information are called “context” (Su and Univ, 2010). With these devices, the user can also interact with her surrounding information resources and also receive latest news and information (Chiu, 2010). Due to the pure volume of information exists in these resources, it is possible for user to confront with “information overload” problem as though (Alexander, Blanzieri, & Giorgini, 2005). Thereby, information is customized accordingly to the customer’s needs and preferences where often various algorithms are used for this customization. Often, recommender systems using diversified algorithms are applied.

Recommender systems are techniques and intelligent applications to assist users in the decision making process (Su & Khoshgoftaar, 2009). Among the various methods used by different recommender systems, collaborative filtering (CF) has become the most popular one (Herlocker, Konstan, Terveen, & Riedl, 2004). CF algorithm makes recommendation based on users’ similar rates to the existing items (Zheng, Wilkinson, & Provost, 2008). In this manner an implicit network is formed among people who are the nodes and each link between two users represents the similarity between them. This similarity implies the similarity between their rates to the existing items.

Besides all the advantages of CF, one drawback of collaborative filtering algorithm is that it considers all similar users for making recommendation and does not distinguish users with higher influence. High influence users are those who have higher impact on the other users` decision making process. However, this method does not provide high accuracy and also require high volume of computations. Cold start is another weakness of this algorithm. Cold start pops up when the algorithm is not able to make recommendations for new users who have no rating and activity in the system - because the similarity between them and other users could not be computed (Su & Khoshgoftaar, 2009; Huang, Chen, Zeng, 2004).

To overcome these shortcomings in CF algorithm, many works have been done. Some of the proposed methods use social networks to increase the performance of recommender systems. Social networks are good resources of similarity data for recommender systems. This data usually is provided in the form of explicit user-generated connections linking pairs of users together (Zheng, Wilkinson, & Provost, 2008). In doing this, since links between the users illustrate their similarity, the computational step of identifying users with the similar rates to the rates of the target user is removed, and the complexity of CF algorithm is reduced from O(N2 + NM) to O(1) - for a system of N users and M items (Golder, Wilkinson, & Huberman, 2007).

In this study, a recommender system in the m-commerce platform is modelled, which is based on developed collaborative filtering algorithm. This system uses mobile data with user consent for identifying individuals with the highest influence in order to make recommendation and extract current user’s context, too. Then, the system uses the information of these impressive users in the current context existed in the social networks for making recommendations. The proposed system avoids the mentioned weaknesses of traditional collaborative filtering. In the other word, in this paper a recommender system is modelled which represent context aware recommendations to any user doing any activity, in any location and time; this proposed system obviates the weaknesses of traditional collaborative filtering and has a higher accuracy than it.

There are some assumptions for designing this system. The first assumption is that the user’s mobile data are available and can be used with her consent. These data are the ID of people in the contact list of mobile, the relationship between the subject and those contacts, and the call and SMS log in a specific period.

Complete Article List

Search this Journal:
Reset
Open Access Articles: Forthcoming
Volume 10: 4 Issues (2017)
Volume 9: 4 Issues (2016)
Volume 8: 4 Issues (2015)
Volume 7: 4 Issues (2014)
Volume 6: 4 Issues (2013)
Volume 5: 4 Issues (2012)
Volume 4: 4 Issues (2011)
Volume 3: 4 Issues (2010)
Volume 2: 4 Issues (2009)
Volume 1: 4 Issues (2008)
View Complete Journal Contents Listing