|
KMID : 1124020210370030071
|
|
Korean Social Security Studies
2021 Volume.37 No. 3 p.71 ~ p.90
|
|
|
Predicting the Quality of Household Debt using Big Data and Machine Learning
|
|
Park Jung-Min
Song Tae-Min
|
|
|
Abstract
|
|
|
|
This study aims to predict the quality of household debt using big data and machine learning approach. Data for this study include a total of 3,927,165 debt-related documents collected through 280 publicly available online channels in South Korea for the 5 year period between 2014 and 2018. Supervised machine learning techniques used in this study include naive Bayes classification, logistic regression, random forest, decision tree, artificial neural network, support vector machine algorithms. An unsupervised machine learning technique, association analysis, was also applied. The results show that machine learning algorithms were highly capable of predicting the quality of household debt based on a combination of an array of debt-related and sociodemographic characteristics without such information as income, asset, total amount of debt, amount of repayment. Practice and methodological implications of the findings were discussed.
|
|
|
KEYWORD
|
|
|
debt, household default, big data, machine learning, artificial intelligence
|
|
|
FullTexts / Linksout information
|
|
|
|
|
|
Listed journal information
|
|
|
|