Date of Original Version

August 2013



Abstract or Description

In this paper we examine link prediction for two types of data sets with mobility data, namely call data records (from the MIT Reality Mining project) and location-based social networking data (from the companies Gowalla and Brightkite). These data sets contain location information, which we incorporate in the features used for prediction. We also examine different strategies for data cleaning, in particular thresholding based on the amount of social interaction. We investigate the machine learning algorithms Decision Tree, Naïve Bayes, Support Vector Machine, and Logistic Regression. Generally, we find that our feature selection and filtering of the data sets have a major impact on the accuracy of link prediction, both for Reality Mining and Gowalla. Experimentally, the Decision Tree and Logistic Regression classifiers performed best.


Published In

Eleventh Workshop on Mining and Learning with Graphs.


To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.