Predicting financial default risks - A machine learning approach using smartphone data

Aug 5, 2026

This study leverages machine learning to predict financial default risks using smartphone-derived behavioral data. Data from 1,000 individuals who took personal loans were collected over six months before application, emphasizing app usage frequency, GPS mobility signals, and communication patterns. Logistic Regression, Decision Trees, and Random Forest were used to model correlations with default outcomes. Random Forest achieved the highest performance with 85% accuracy. Results indicate that frequent use of financial apps relates to lower risk, whereas irregular communication and erratic mobility signal higher default likelihood. Findings support enhanced credit scoring while highlighting privacy and data security considerations.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Introduction

Why machine learning?

What is considered machine learning?

Predicting financial default risks: A machine learning approach using smartphone data Shinta Palupi 1, Gunawan 2, Ririn Kusdyawati 3, Richki Hardi 4*, Rana Zabrina 5 1 Information System Department, Universitas Mulia, Indonesia 2,4 Informatics Department, Universitas Mulia, Indonesia 3 Office Administration Department, Universitas Mulia, Indonesia 5 Information Technology Department, Universitas Mulia, Indonesia *Corresponding Author: [richki@universitasmulia.ac.id](richki@universitasmulia.ac.id) Abstract: This study leverages machine learning (ML) techniques to predict financial default risks using smartphone data, providing a novel approach to financial risk assessment. Data were collected from 1,000 individuals who had taken personal loans, focusing on key behavioral parameters such as app usage frequency, GPS location data, and communication patterns over six months before loan application. The analysis employed Logistic Regression, Decision Trees, and Random Forest models to determine correlations between these parameters and default risks. The Random Forest model demonstrated superior performance, achieving 85% accuracy. Key findings show that high usage of financial apps was associated with lower default risks, while irregular communication patterns and erratic mobility were significant indicators of higher risk. These results suggest that smartphone-derived behavioral data can significantly enhance traditional credit scoring methods. The study not only contributes to predictive analytics in financial risk management but also raises ethical considerations around privacy and data security. Keywords: Financial Default, Machine Learning, Predictive Analytics, Risk Prediction, Smartphone Data History Article: Submitted 14 April 2024 | Revised 9 October 2024 | Accepted 11 November 2024 How to Cite: S. Palupi, Gunawan, R. Kusdyawati, R. Hardi, R. Zabrina,“Predicting financial default risks: A machine learning approach using smartphone data,” Matrix: Jurnal Manajemen Teknologi dan Informatika, vol. 14, no. 3, pp. 107-118, [2024. doi.org/10.31940/matrix.v14i3. 107-118](2024. doi.org/10.31940/matrix.v14i3. 107-118) Introduction The rise of smartphone technology has revolutionized numerous aspects of daily life, from communication and entertainment to shopping and finance. These devices collect vast amounts of data on user behavior, offering new opportunities for various applications, including financial risk assessment. Traditional credit scoring models, which rely heavily on historical financial data such as credit scores and income levels, often fail to capture the full spectrum of individual financial behaviors, particularly in regions where formal credit histories are scarce [1] . These traditional models also tend to be static, overlooking real-time behavioral data that could provide deeper insights into financial risk. A significant gap in the literature lies in the exploration of non-traditional data sources, such as smartphone-derived behavioral data, for financial risk prediction. Recent studies have begun to examine digital footprints, including app usage and communication patterns, as potential indicators of financial behavior [2] . However, the practical application of such data in financial models remains underexplored, particularly in the context of integrating advanced machine learning (ML) techniques [3] . This research aims to address this gap by investigating how ML can be employed to analyze smartphone data and predict financial default risks. Why machine learning? Machine learning was chosen for this study because of its ability to process large datasets and uncover complex, non-linear patterns that may not be evident using traditional statistical methods [4] . Unlike conventional models, machine learning algorithms can continuously learn and improve from new data, making them particularly suited for dynamic and behavioral data like smartphone usage patterns. The flexibility of ML

FAQs

What data source does the study use to predict financial default risk?

The study uses smartphone-derived behavioral data collected over six months before loan application, including app usage frequency, GPS location/mobility patterns, and communication patterns.

Which machine learning models are evaluated in the research?

Logistic Regression, Decision Trees, and Random Forest are employed to determine relationships between the behavioral parameters and default risk.

What factors are associated with higher or lower default risk?

High usage of financial apps is linked to lower default risks, while irregular communication patterns and erratic mobility are significant indicators of higher risk.

Aug 5, 2026

Upload by River Wang