Adaptive Mobile News Personalization Using Social Networks
Posted: 2 Jul 2014
Date Written: May 1, 2014
We present an adaptive personalization system for personalizing news feeds on mobile devices. It learns from an individual’s reading history, automatically discovers new material as a result of shared interests in the user's social network and adapts the news feeds shown to the user. We develop two text analysis algorithms, a Modified Naïve Bayes algorithm, and an approach based on Bayesian Logistic Regression. We demonstrate in a field study that the Modified Naïve Bayes algorithm outperforms the Bayesian Logistic regression and other benchmark approaches. In addition, using article choices from an individual’s social network improves the quality of personalization, leading to more readership of the news articles provided. Both homophily and induction influence readership, but the induction effect is stronger. Finally we illustrate in a second field study that the news personalization system based on the Modified Naïve Bayes algorithm, which uses only closed form calculations and runs in real time, can be effectively implemented on mobile devices. This research suggests that utilizing social networks may be a promising avenue for improving personalization of services.
Keywords: Personalization, Social Networks, News, Bayes Classifier, Recommendation Systems, Mobile Commerce, Smart Phones, Service Marketing
Suggested Citation: Suggested Citation