Online prediction algorithm of the news' popularity for wireless cellular pushing

There is an obviously converging property of online news browsed by mobile users, and a small number of popular contents play an important role in data traffic of wireless cellular network. Therefore, if we are able to estimate the grade of news' popularity from early observation of users' clicks, the transmitting resources would be saved significantly by broadcasting the top-N popular news to mobile devices of potential users. In this paper, we present an improved linear prediction scheme on a logarithmic scale, which uses an optimal observation threshold to make better performance of prediction. And the proposed prediction scheme is compared with classic rank algorithms using machine learning techniques. Based on data set collected from Chinese commercial cellular network, the simulation results indicate that this scheme provides a much better prediction accuracy as evaluated by Normal Discounted Cumulative Gain (NDCG) metric.

[1]  Urs Niesen,et al.  Fundamental Limits of Caching , 2014, IEEE Trans. Inf. Theory.

[2]  Hang Li,et al.  AdaRank: a boosting algorithm for information retrieval , 2007, SIGIR.

[3]  Serge Fdida,et al.  From popularity prediction to ranking online news , 2014, Social Network Analysis and Mining.

[4]  Jaana Kekäläinen,et al.  IR evaluation methods for retrieving highly relevant documents , 2000, SIGIR '00.

[5]  Xin Chen,et al.  A Popularity-Based Prediction Model for Web Prefetching , 2003, Computer.

[6]  Maarten de Rijke,et al.  News Comments: Exploring, Modeling, and Online Prediction , 2010, ECIR.

[7]  Jaana Kekäläinen,et al.  IR evaluation methods for retrieving highly relevant documents , 2000, SIGIR Forum.

[8]  Kavé Salamatian,et al.  Modeling and predicting the popularity of online contents with Cox proportional hazard regression model , 2012, Neurocomputing.

[9]  Jing Wang,et al.  Content Aware Soft Real Time Media Broadcast (CASoRT) , 2008, 2008 Third International Conference on Communications and Networking in China.

[10]  Vicenç Gómez,et al.  Description and Prediction of Slashdot Activity , 2007, 2007 Latin American Web Conference (LA-WEB 2007).

[11]  Serge Fdida,et al.  Predicting the popularity of online articles based on user comments , 2011, WIMS '11.

[12]  Bernardo A. Huberman,et al.  Predicting the popularity of online content , 2008, Commun. ACM.

[13]  Shai Shalev-Shwartz,et al.  Stochastic dual coordinate ascent methods for regularized loss , 2012, J. Mach. Learn. Res..

[14]  Urs Niesen,et al.  Fundamental limits of caching , 2012, 2013 IEEE International Symposium on Information Theory.

[15]  J. Friedman.  Greedy function approximation: A gradient boosting machine. , 2001 .

[16]  Jian Sun,et al.  Recommendation scheme based on converging properties for contents broadcasting , 2015, 2015 International Conference on Computing, Networking and Communications (ICNC).