Content and location based big data analytics for mobile notification delivery

Abstract

The news media these days use mobile applications to reach out to their users. Users are ubiquitously equipped with mobile devices, and pushing notifications to them can improve the user experience. In this thesis, mobile notification opening data of 1.8 million users amounting to 240 million log-records in a real mobile news application, Apple Daily, is collected for 2 months and studied. A framework is designed for effective mobile notification delivery to the end users to improve their average notification opening rate and response time. It is the first work to propose a model to quantify notification effectiveness in terms of the number of notifications received in a given time. A method of grouping similar locations based on their social nature or function is then described, and compared to existing feature-based location grouping methods. Correlations among the location groups are identified based on user preferences for news categories, and clusters of location groups are observed based on news categories read. These are then used in a novel convex optimization scheme to increase users overall notification opening rate and reduce their notification response times. This work is also the first to address the problem of notification recommendation as an optimization problem. The method shows an improvement of the average user notification opening rate by up to 62.4%, and that of the average user response time by up to 17.2%, depending on the location grouping algorithm used.

Date of Award	2018
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology