Multivariate Time series classification for stress detection A case study in near real-time stress

Gathered physiological data including heart rate and respiration to serve as inputs for stress prediction models.
Explored various machine learning and deep learning models suitable for time series classification, considering the complexity of stress detection.
Identified and implemented preprocessing techniques tailored to each model to ensure optimal data quality and model performance.
Evaluated multiple machine learning and deep learning models, including Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and transformers, for their effectiveness in stress prediction.
Assessed the performance of the trained models using metrics such as accuracy and F1 score, indicating the model's ability to correctly classify stress levels.
Recognized the potential for improvement with additional sensor data, such as Electrodermal Activity (EDA) and skin temperature, and outlined steps for further research and development.

Collected comprehensive medical data comprising nearly 50,000 patient records.
Conducted in-depth Exploratory Data Analysis (EDA) to discern the demographic distribution based on age, gender, and pre-existing health conditions.
Mitigated the issue of an imbalanced dataset by employing the Synthetic Minority Over-sampling Technique (SMOTE) to generate synthetic data.
Identified and extracted pertinent features highly correlated with stroke occurrences.
Implemented a suite of advanced Machine Learning algorithms including linear regression, Random Forest, and Support Vector Machine (SVM) to detect strokes.
Achieved a remarkable accuracy rate of nearly 92% following model training.