Nokia Mobile Data Challenge 2012

A team from ODU Computer Science successfully completed the Nokia Mobile Data Challenge.  Results will be presented at a workshop organized in connection with Pervasive 2012 in Newcastle, UK, June 18-19.  Team members were PhD. student Shahram Mohrehkesh and faculty members Dr. Shuiwang Ji, Dr. Tamer Nadeem and Dr. Michele Weigle.

The challenge participants competed in solving data mining problems of big data in the mobile computing domain. There were 108 submissions in two tracks: 59 to the Dedicated Track and 49 to the Open Track. In the open challenge, participants were trying to find new knowledge and insight from the dataset. In the Dedicated Track, three tasks were proposed for participation: finding semantic location (home, office, etc.) of user presence, prediction of next location, and prediction of users' demographic information.  The ODU team participated in the Dedicated Track, where only 16% of submissions were accepted for oral presentation. The ODU team's task was to predict users' demographic information from their mobile phone usage. Teams with the highest prediction accuracy were selected to present their results.  See the team's poster for more details.

The Challenge dataset was released in early January 2012, and teams had four months to analyze this big data. For the dedicated task, the dataset contained a 10-month log of mobile usage for 80 users. For each user, acceleration, application, Bluetooth, calendar, call log, GSM, media, media play, system, and WirelessLAN were logged. Details about the dataset and collection process can be found here. Pre-processing of the data and extraction of features was the main part of the team's activity; it took more than 70% of the ODU team's time. I, personally, learned much from this challenge. The main thing for me was the experience of mining on big data.