Using artificial intelligence to overcome over-indebtedness and fight poverty.

Area of Science:

• Computational social science
• Applied artificial intelligence
• Socioeconomic studies

Background:

• Over-indebtedness is a significant issue for households in high poverty risk contexts.
• Existing models for scarcity lack comprehensive methods for identifying and predicting over-indebtedness.
• Understanding the multifaceted nature of over-indebtedness is crucial for effective intervention.

Purpose of the Study:

• To apply Automated Machine Learning (AutoML) for understanding and overcoming over-indebtedness in high poverty risk environments.
• To identify distinct clusters of over-indebted households and predict associated risk factors.
• To contribute theoretically and methodologically to scarcity models with practical implications.

Main Methods:

• Utilized unsupervised machine learning (Self-Organizing Maps) on a dataset of 1654 over-indebted households.
• Employed supervised machine learning with exhaustive grid search (32,730 models) to identify predictive factors.
• Applied Nu-Support Vector Machine for accurate prediction of over-indebtedness risk.

Main Results:

• Identified three distinct clusters: low-income (31.27%), low credit control (37.40%), and crisis-affected households (31.33%).
• Nu-Support Vector Machine achieved 89.5% accuracy in predicting over-indebtedness risk factors.
• The AutoML approach provides a novel method for characterizing poverty risk at earlier stages.

Conclusions:

• AutoML offers a powerful framework for analyzing and addressing over-indebtedness in vulnerable populations.
• The identified clusters and risk factors facilitate the development of customized interventions.
• This research enhances the understanding of scarcity and poverty, offering practical applications for business and society.