Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Handling Extreme Class Imbalance: Using GANs in Data Augmentation for Suicide Prediction

Vaishnavi Visweswaraiah, Tanvi Banerjee, William Romine

TL;DR

The paper tackles extreme class imbalance in suicide prediction using GAN-based data augmentation to synthesize minority-class examples. It compares Logistic Regression, Support Vector Machines, and Random Forest trained on real versus GAN-generated data, evaluating performance on a real test set and analyzing how augmentation affects recall and precision for rare positive cases. Findings show that LR and SVM trained on real data can detect the sole positive case, while GAN-trained models can identify the positive with more false positives; among GAN-trained models, SVM_G offers the most balanced performance. The work demonstrates the potential of conditional GANs for augmenting tabular behavioral health data and informs model selection under extreme imbalance, highlighting the need for clinical validation.

Abstract

Suicide prediction is the key for prevention, but real data with sufficient positive samples is rare and causes extreme class imbalance. We utilized machine learning (ML) to build the model and deep learning (DL) techniques, like Generative Adversarial Networks (GAN), to generate synthetic data samples to enhance the dataset. The initial dataset contained 656 samples, with only four positive cases, prompting the need for data augmentation. A variety of machine learning models, ranging from interpretable data models to black box algorithmic models, were used. On real test data, Logistic Regression (LR) achieved a weighted precision of 0.99, a weighted recall of 0.85, and a weighted F1 score of 0.91; Random Forest (RF) showed 0.98, 0.99, and 0.99, respectively; and Support Vector Machine (SVM) achieved 0.99, 0.76, and 0.86. LR and SVM correctly identified one suicide attempt case (sensitivity:1.0) and misclassified LR(20) and SVM (31) non-attempts as attempts (specificity: 0.85 & 0.76, respectively). RF identified 0 suicide attempt cases (sensitivity: 0.0) with 0 false positives (specificity: 1.0). These results highlight the models' effectiveness, with GAN playing a key role in generating synthetic data to support suicide prevention modeling efforts.

Handling Extreme Class Imbalance: Using GANs in Data Augmentation for Suicide Prediction

TL;DR

The paper tackles extreme class imbalance in suicide prediction using GAN-based data augmentation to synthesize minority-class examples. It compares Logistic Regression, Support Vector Machines, and Random Forest trained on real versus GAN-generated data, evaluating performance on a real test set and analyzing how augmentation affects recall and precision for rare positive cases. Findings show that LR and SVM trained on real data can detect the sole positive case, while GAN-trained models can identify the positive with more false positives; among GAN-trained models, SVM_G offers the most balanced performance. The work demonstrates the potential of conditional GANs for augmenting tabular behavioral health data and informs model selection under extreme imbalance, highlighting the need for clinical validation.

Abstract

Suicide prediction is the key for prevention, but real data with sufficient positive samples is rare and causes extreme class imbalance. We utilized machine learning (ML) to build the model and deep learning (DL) techniques, like Generative Adversarial Networks (GAN), to generate synthetic data samples to enhance the dataset. The initial dataset contained 656 samples, with only four positive cases, prompting the need for data augmentation. A variety of machine learning models, ranging from interpretable data models to black box algorithmic models, were used. On real test data, Logistic Regression (LR) achieved a weighted precision of 0.99, a weighted recall of 0.85, and a weighted F1 score of 0.91; Random Forest (RF) showed 0.98, 0.99, and 0.99, respectively; and Support Vector Machine (SVM) achieved 0.99, 0.76, and 0.86. LR and SVM correctly identified one suicide attempt case (sensitivity:1.0) and misclassified LR(20) and SVM (31) non-attempts as attempts (specificity: 0.85 & 0.76, respectively). RF identified 0 suicide attempt cases (sensitivity: 0.0) with 0 false positives (specificity: 1.0). These results highlight the models' effectiveness, with GAN playing a key role in generating synthetic data to support suicide prevention modeling efforts.
Paper Structure (15 sections, 5 figures, 4 tables)

This paper contains 15 sections, 5 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Experiment
  4. Initial and GAN Data Exploration
  5. Real Data
  6. Synthetic GAN Data
  7. Model Training & Results
  8. Discussion
  9. Limitations and Future Work
  10. Overall Experiment Architecture
  11. Dataset
  12. Experiments
  13. Data Generation Using GAN
  14. Model Training
  15. Results

Figures (5)

  • Figure 1: real vs gan data distribution for all samples
  • Figure 2: Real Data distribution for all suicide attempt samples ( label-0 : no suicide attempt (Blue), label-1: suicide attempt (Red)
  • Figure 3: GAN Data distribution for all suicide attempt samples ( label-0 : no suicide attempt (Blue), label-1: suicide attempt (Red)
  • Figure 4: Overall Experiment Architecture
  • Figure 5: GAN loss highlighting the changes in generator loss versus discriminator loss over epochs