IMPORTANT NOTICE: The date, time, and room assignment of YOUR presentation is SUBJECT TO CHANGE.
Proposal authors can use this tool to see where they have been placed in the program agenda for an Oral or Poster Session.
Scroll down to search by the Submitter or Author Name, by Date/Time, or by Keywords by expanding FILTERS.
Confirm your place in the schedule by following the instructions that were emailed to you. Each presentation must have a separate paid registration. Contact the ACCI office immedicately by email at admin@consumerinterests.org to report any conflict, all corrections to the details of the presentation (including author names and the order they are listed as this is how it will be in the final program), or if you have any questions. Please be sure to reference the session title(s), date(s), and time(s) when you contact us.
P111 Predicting default risk of disabled adults in South Korea using machine learning techniques
Short Description
This study applies machine learning (ML) techniques to predict default risk among disabled adults in South Korea, utilizing data from the Panel Survey of Employment for the Disabled (PSED). Disabled individuals face unique socioeconomic challenges, including limited employment opportunities, inadequate social support, and significant medical expenses, which contribute to financial instability and heightened default risk. Existing financial risk models typically overlook these multifaceted aspects of disability, resulting in a critical gap in accurately assessing default risk within this population. Our research aims to develop predictive models specifically tailored to the socioeconomic backgrounds of disabled adults, enhancing the precision of default risk assessments.
The methodology involves comprehensive preprocessing of PSED data, such as demographics, income levels, disability type and severity, and employment status. We employ various machine learning algorithms, including logistic regression, random forests, and extreme gradient boosting, validated through cross-validation and performance metrics. Preliminary results indicate that ensemble learning models, especially extreme gradient boosting, exhibit superior predictive performance. Key predictors of default risk include employment, disability severity, and dependence on social support.
The findings could inform financial institutions and policymakers by highlighting overlooked factors in risk assessments and promoting inclusive financial practices that support the economic well-being of disabled individuals.
Type of presentation
Accepted Poster Presentation