A Study of WhatsApp Usage Patterns and Prediction Models without Message Content

TL;DR

This study analyzes WhatsApp usage patterns using activity data from over 6 million messages to develop demographic prediction models, revealing significant behavioral differences across gender and age groups without examining message content.

Contribution

The paper introduces novel demographic prediction models based solely on activity data, highlighting usage differences and behavioral patterns in WhatsApp without analyzing message content.

Findings

Successful prediction of user gender and age demographics.

Identification of distinct usage habits across different groups.

Models predicting group behavior attributes like file sharing and activity frequency.

Abstract

Internet social networks have become a ubiquitous application allowing people to easily share text, pictures, and audio and video files. Popular networks include WhatsApp, Facebook, Reddit and LinkedIn. We present an extensive study of the usage of the WhatsApp social network, an Internet messaging application that is quickly replacing SMS messaging. In order to better understand people's use of the network, we provide an analysis of over 6 million messages from over 100 users, with the objective of building demographic prediction models using activity data. We performed extensive statistical and numerical analysis of the data and found significant differences in WhatsApp usage across people of different genders and ages. We also inputted the data into the Weka data mining package and studied models created from decision tree and Bayesian network algorithms. We found that different genders and age demographics had significantly different usage habits in almost all message and group attributes. We also noted differences in users' group behavior and created prediction models, including the likelihood a given group would have relatively more file attachments, if a group would contain a larger number of participants, a higher frequency of activity, quicker response times and shorter messages. We were successful in quantifying and predicting a user's gender and age demographic. Similarly, we were able to predict different types of group usage. All models were built without analyzing message content. We present a detailed discussion about the specific attributes that were contained in all predictive models and suggest possible applications based on these results.