[Paper Review] Machine Learning, Deep Learning and Data Preprocessing Techniques for Detection, Prediction, and Monitoring of Stress and Stress-related Mental Disorders: A Scoping Review

Moein Razavi, Samira Ziyadidegan+6 more|arXiv (Cornell University)|Aug 8, 2023

Mental Health Research Topics4 citations

TL;DR

This scoping review analyzes 98 peer-reviewed studies on machine learning (ML) and deep learning (DL) for detecting, predicting, and monitoring stress and stress-related mental disorders. It identifies SVM, neural networks, and random forest as top-performing models, with physiological signals like heart rate and skin response as key predictors, and highlights dimensionality reduction as a critical preprocessing step for improved model performance.

ABSTRACT

Background: Mental stress and its consequent mental disorders (MDs) are significant public health issues. With the advent of machine learning (ML), there's potential to harness computational techniques for better understanding and addressing these problems. This review seeks to elucidate the current ML methodologies employed in this domain to enhance the detection, prediction, and analysis of mental stress and MDs. Objective: This review aims to investigate the scope of ML methodologies used in the detection, prediction, and analysis of mental stress and MDs. Methods: Utilizing a rigorous scoping review process with PRISMA-ScR guidelines, this investigation delves into the latest ML algorithms, preprocessing techniques, and data types used in the context of stress and stress-related MDs. Results and Discussion: A total of 98 peer-reviewed publications were examined. The findings highlight that Support Vector Machine (SVM), Neural Network (NN), and Random Forest (RF) models consistently exhibit superior accuracy and robustness among ML algorithms. Physiological parameters such as heart rate measurements and skin response are prevalently used as stress predictors due to their rich explanatory information and ease of data acquisition. Dimensionality reduction techniques, including mappings, feature selection, filtering, and noise reduction, are frequently observed as crucial steps preceding the training of ML algorithms. Conclusion: This review identifies significant research gaps and outlines future directions for the field. These include model interpretability, model personalization, the incorporation of naturalistic settings, and real-time processing capabilities for the detection and prediction of stress and stress-related MDs. Keywords: Machine Learning; Deep Learning; Data Preprocessing; Stress Detection; Stress Prediction; Stress Monitoring; Mental Disorders

Motivation & Objective

• To map the current landscape of machine learning and deep learning applications in stress and stress-related mental disorder detection, prediction, and monitoring.
• To identify the most effective ML algorithms, data types, and preprocessing techniques used in this domain.
• To assess methodological trends, including data acquisition modalities and model performance metrics.
• To highlight research gaps in model interpretability, personalization, real-time processing, and naturalistic data collection.
• To guide future research by outlining key challenges and opportunities in deploying ML for mental health applications.

Proposed method

• Conducted a systematic scoping review using PRISMA-ScR guidelines to identify and analyze peer-reviewed publications on ML/DL for stress and mental disorders.
• Selected 98 relevant studies based on predefined inclusion and exclusion criteria from a comprehensive literature search.
• Categorized studies by ML algorithm type (e.g., SVM, Random Forest, Neural Networks), data modality (e.g., physiological, self-reported, behavioral), and preprocessing techniques.
• Evaluated preprocessing methods including feature selection, filtering, noise reduction, and dimensionality reduction (e.g., PCA, LDA) applied before model training.
• Synthesized findings on model performance, data sources, and methodological approaches across studies.
• Identified recurring patterns in model architecture, data collection settings (e.g., lab vs. real-world), and evaluation metrics.

Experimental results

Research questions

• RQ1 Which machine learning and deep learning models demonstrate the highest accuracy and robustness in detecting and predicting stress and stress-related mental disorders?
• RQ2 What types of data—especially physiological, self-reported, or behavioral—are most commonly used as inputs for these models?
• RQ3 What preprocessing techniques are most frequently applied, and how do they contribute to model performance?
• RQ4 How do current studies address challenges such as model interpretability, personalization, and real-time monitoring?
• RQ5 What are the key research gaps and future directions in applying ML to stress and mental health monitoring?

Key findings

• Support Vector Machines (SVM), Neural Networks (NN), and Random Forest (RF) models consistently demonstrated superior accuracy and robustness across multiple studies.
• Physiological signals such as heart rate variability and skin conductance response were the most prevalent and informative predictors due to their direct link to autonomic nervous system activity.
• Dimensionality reduction techniques—including feature selection, filtering, and noise reduction—were consistently applied before model training to improve performance and reduce overfitting.
• Despite high performance in controlled settings, few studies implemented real-time processing or deployed models in naturalistic, ambulatory environments.
• Model interpretability and personalization remain underdeveloped, with limited integration of individual-specific data or explainable AI techniques.
• A significant proportion of studies relied on small, lab-based datasets, highlighting a need for larger, diverse, and ecologically valid data collection in future work.