Comparative Analysis of Machine Learning Techniques for Human Stress Level Prediction: Performance Metrics and Insights

Biomedical Engineering

In this review study, several machine learning models that are used to predict human stress levels are thoroughly analyzed, with a particular emphasis on Random Forest (RF), Support Vector Machine (SVM), and Long Short-Term Memory (LSTM) networks. The design, benefits, limitations, and performance measures of each model are thoroughly examined to provide a comprehensive understanding of their capabilities. Known for its ensemble learning methodology, Random Forest constructs several decision trees to improve precision and robustness. Even though its astounding accuracy rates, which range from 72% to 95%, can be challenging to interpret and computationally demanding, especially in vital industries like healthcare. However, with accuracy rates ranging from 79% to 95%, SVM is acknowledged for its efficiency in managing high-dimensional spaces and complicated datasets. Nevertheless, it is sensitive to feature scaling and can yield difficult-to-interpret complex decision limits. On the other hand, LSTM networks are especially well-suited for stress prediction from time-series data since they are expressly made for sequence prediction tasks and are excellent at capturing temporal relationships in data. LSTM models have proven to be remarkably effective, with accuracy reaching 99.71%. However, their usefulness in real-time scenarios may be limited because to their requirement for substantial processing resources and a huge volume of labeled data for effective training. All in all, these models' performance measures highlight how successful they are at predicting stress, with RF, SVM, and LSTM each having particular advantages and disadvantages. This study offers insightful information for academics and practitioners in the subject of stress analysis, emphasizing the significance of choosing the right model depending on particular requirements.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.