Reliable Multi-Class Mental Health Prediction Using a WiSARD Discriminator Model on Imbalanced Data

Area of Science:

• Computational psychiatry
• Machine learning for healthcare

Background:

• Machine learning (ML) is crucial for mental disorder prediction, aiding early screening and personalized care.
• Challenges include high dimensionality, class imbalance, and subtle psychological features in multi-class classification.

Purpose of the Study:

• To introduce and evaluate an interpretable, RAM-based WiSARD classifier for multi-disorder mental health prediction.
• To compare WiSARD's performance against established ML models on a public dataset.

Main Methods:

• A retrospective study used the Kaggle Mental Disorders Dataset (637 complete cases, 29 features).
• WiSARD was tested using 10-fold stratified cross-validation against Multilayer Perceptron, Naïve Bayes, DTNB, IB1, and A1DE.
• Performance metrics included precision, recall, F-measure, accuracy, MCC, MAE, and KS.

Main Results:

• WiSARD achieved superior performance with 98.27% accuracy, 0.983 F-measure, 0.982 MCC, and 0.981 KS.
• WiSARD demonstrated better tolerance for misclassifications in minority classes, addressing data imbalance.
• An ablation study confirmed WiSARD's reliability and interpretability via RAM-based pattern recognition.

Conclusions:

• WiSARD is a promising, interpretable model for multi-class mental disorder prediction, particularly in imbalanced datasets.
• Findings are limited to a single non-clinical dataset with self-reported data and require formal psychiatric validation.