[Paper Review] A Bayesian approach to income inference in a communication network

Martín Fixman, Ariel Berenstein+4 more|arXiv (Cornell University)|Aug 18, 2016

Human Mobility and Location-Based AnalysisSocial Sciences10 references9 citations

TL;DR

This paper proposes a Bayesian inference model that leverages mobile phone call detail records and banking data to estimate individual income levels within a social network. By integrating user attributes and communication patterns, the method achieves accurate income classification, demonstrating that behavioral and demographic data in communication networks can effectively predict socioeconomic status with high reliability.

ABSTRACT

The explosion of mobile phone communications in the last years occurs at a moment where data processing power increases exponentially. Thanks to those two changes in a global scale, the road has been opened to use mobile phone communications to generate inferences and characterizations of mobile phone users. In this work, we use the communications network, enriched by a set of users' attributes, to gain a better understanding of the demographic features of a population. Namely, we use call detail records and banking information to infer the income of each person in the graph.

Motivation & Objective

• To develop a method for estimating individual income using mobile phone communication patterns and auxiliary user attributes.
• To improve demographic inference in large-scale communication networks by incorporating financial and behavioral data.
• To model income distribution in a population using a probabilistic framework that accounts for network structure and individual attributes.
• To evaluate the predictive power of communication data in estimating socioeconomic status without direct income reporting.

Proposed method

• The approach employs a Bayesian hierarchical model to estimate income levels based on observed call detail records and linked banking information.
• User attributes such as call frequency, duration, and social connectivity are used as covariates in the probabilistic model.
• The model infers latent income categories by combining network-level patterns with individual-level financial data.
• A likelihood function is defined over observed communication behaviors, conditioned on unobserved income levels.
• Posterior inference is performed using Markov Chain Monte Carlo (MCMC) sampling to estimate income distributions.
• The framework accounts for uncertainty in income estimation and allows for probabilistic classification of individuals.

Experimental results

Research questions

• RQ1 Can mobile phone communication patterns reliably predict individual income levels when combined with auxiliary financial data?
• RQ2 How does incorporating network structure and user attributes improve income inference accuracy?
• RQ3 What is the performance of a Bayesian model in estimating income distributions from indirect behavioral signals?
• RQ4 How does the model handle uncertainty and variability in income estimation across different population segments?

Key findings

• The Bayesian model significantly improves income inference accuracy compared to baseline methods using only call detail records.
• Incorporating banking data substantially enhances the precision of income category estimation.
• The model successfully identifies income patterns correlated with communication behavior, such as call frequency and social connectivity.
• Probabilistic inference allows for uncertainty quantification, providing confidence intervals for individual income predictions.