Machine learning-based analysis of adolescent gambling factors.

Area of Science:

• Adolescent psychology
• Computational social science
• Public health

Background:

• Adolescent problem gambling is a growing concern due to increased online accessibility.
• Easy access to online gambling poses significant risks to adolescent well-being.

Purpose of the Study:

• To develop and evaluate a machine learning-based method for predicting the severity of problem gambling in adolescents.
• To identify key predictors of problem gambling among Korean youth.

Main Methods:

• Utilized data from the 2018 National Survey on Youth Gambling Problems (n=5,045 adolescents).
• Employed the Gambling Problem Severity Scale for outcome labeling.
• Trained and compared four machine learning models: Random Forest (RF), Support Vector Machine (SVM), Extra Trees (ETs), and Ridge Regression after feature selection.

Main Results:

• Online gambling behavior, winning experiences, and social gambling were identified as crucial predictors.
• All models achieved an Area Under the Curve (AUC) > 0.7.
• Extra Trees (ETs) model yielded the highest AUC (0.755), Random Forest (RF) showed the best accuracy (71.8%), and Support Vector Machine (SVM) achieved the highest F1 score (0.507).

Conclusions:

• Machine learning models effectively predict adolescent problem gambling severity.
• The developed method can assist in screening adolescents at risk.
• Future research with larger datasets can further enhance these machine learning approaches.