A study of age and gender seen through mobile phone usage patterns in Mexico

Abstract

Mobile phone usage provides a wealth of information, which can be used to better understand the demographic structure of a population. In this paper we focus on the population of Mexican mobile phone users. Our first contribution is an observational study of mobile phone usage according to gender and age groups. We were able to detect significant differences in phone usage among different subgroups of the population. Our second contribution is to provide a novel methodology to predict demographic features (namely age and gender) of unlabeled users by leveraging individual calling patterns, as well as the structure of the communication graph. We provide details of the methodology and show experimental results on a real world dataset that involves millions of users.

Improving individual predictions using social networks assortativity

“…Although optimal s jk values are quite independent of β, the performances are not, with highest accuracy gains in the range [70,85]%. This range covers the accuracies reached by stateof-the-art techniques aiming to predict gender using individuallevel features [4], [5], [26]. Likewise, for an assortativity coefficient similar to the one of G S (≈ 0.25), the accuracy gains on synthetic networks are significant when β ∈ [0.62, 0.92].…”

“…Our methodology is now tested on G S , while simulating individual prior predictions. The obtained performances are compared with the results of a baseline method, termed the reaction-diffusion algorithm [5].…”

“…Some methods are currently developed to exploit that assortativity [2], [16]. However, few studies take the global network structure into account [5], [17]. Also, to the best of our knowledge, no research quantifies how the performances are related to the assortativity strength.…”

“…This type of information is indeed a key input for the establishment of economic and social policies, health campaigns, market segmentation, etc. [3]- [5]. Nevertheless, especially (but not exclusively) in developing countries, such statistics are often scarce or even lacking, as local censuses are costly, rough, time-consuming and hence rarely up-to-date [6].…”

A comparative study of social network classifiers for predicting churn in the telecommunication industry

“…Homophily is used to describe peoples' tendency to associate with those that are similar to them [22]. Studies have shown that there is significant difference in the phone usage of different genders and age groups, and the usage patterns can be used to predict the demographic features [23], [24]. Additionally, homophily can be used to predict similarities between people who interact frequently or to predict interactions between people who behave in a similar way [13], [25].…”