Improving Predictive Classification Models Using Generative Adversarial Networks in the Prediction of Suicide Attempts

Abstract

A number of machine learning methods can be employed in the prediction of suicide attempts. However, many models do not predict new cases well in cases with unbalanced data. The present study improved prediction of suicide attempts via the use of a generative adversarial network.

Keywords:

• Predictive classification
• generative adversarial networks
• GAN pre-training
• unbalanced data
• suicide prediction

Disclosure Statement

No potential conflict of interest was reported by the authors.

Additional information

Notes on contributors

Anthony A. Mangino

Anthony A. Mangino is a doctoral candidate in Educational Psychology at Ball State University. His research interests are in applying novel machine learning techniques to various data structures and contexts in the social sciences.

Kendall A. Smith

Kendall A. Smith is currently a Research Scientist in the Human Centered Analytics Lab at Notre Dame University. His current research includes Interpretability and Fairness in Machine Learning.

W. Holmes Finch

W. Holmes Finch teaches statistics and psychometrics in the Department of Educational Psychology at Ball State University.

Maria E. Hernández-Finch

Dr. Maria E. Hernández-Finch is an associate professor of educational psychology at Ball State University. Her primary research areas include the intersection of diversity, early learning, and assessment/identification with a focus on equity and traditionally understudied and disenfranchised populations on both ends of the human exceptionality continuum. Current studies and grants focus on early gifted identification and twice exceptionality, investigating the developmental trajectories of neurodiverse individuals with Autism Spectrum Disorder, efficacious suicide prevention training, and culturally sustaining, relevant and responsive intervention, consultation, and school support team collaboration.