Panic Attack Detection using Machine Learning and Deep Learning

A Panic Attack Detection System based on physiological signals such as heart rate and breathing speed. This project explores multiple machine learning models and deep learning techniques to classify whether a person is experiencing a panic attack.

Key Features

• Exploratory Data Analysis (EDA)

  • Data visualization and preprocessing using pandas, seaborn, and matplotlib.

• Data Preprocessing

  • Feature scaling using StandardScaler and RobustScaler.
  • Handling imbalanced data using SMOTE.

• Deep Learning (ANN) Model

  • Implemented using TensorFlow/Keras (Sequential model).
  • Optimized using:
    • Adam optimizer for efficient learning.
    • EarlyStopping to prevent overfitting.
    • ModelCheckpoint to save the best model.

• Model Evaluation

  • Performance metrics: Accuracy, Precision, Recall, F1-Score, Confusion Matrix.
  • Hyperparameter tuning using GridSearchCV and RandomizedSearchCV.

At the end of the project different models for this have been compared to see how their perfromances change from on another