COVID-19 Detection Model

A Python-based machine learning project that predicts COVID-19 test results using patient symptoms and demographic data. The model analyzes various health indicators to classify coronavirus test outcomes with high accuracy using multiple ML algorithms.

---

Features

• Analyzes COVID-19 patient data from 2020-2021

• Data preprocessing with Label Encoding for categorical variables

• Multiple machine learning algorithms comparison

• Achieves 91.53% accuracy with Random Forest Classifier

• Comprehensive data visualization and analysis

• Training vs Testing performance comparison

• Built using Jupyter Notebook for interactive experimentation.

---

How to Run

1. Clone the repository:

   git clone https://github.com/AdarshZolekar/COVID19-Detection-Model.git

2. Prepare the dataset:

• Download or add the covid_data_2020-2021.csv.zip file to your Google Drive.

• Update the file path in the script to match your location.

3. Run the script:

• Open a terminal or command prompt.

• Navigate to the project directory:

   cd COVID19-Detection-Model

• Run the Jupyter Notebook:

   jupyter notebook Corona-Detection-Model.ipynb

• Or run the Python script directly:

   python Corona-Detection-Model.py

• Execute the notebook cells step-by-step to train models and visualize results.

---

How It Works

1. Data Loading: Reads the COVID-19 dataset containing patient symptoms and test results from 2020-2021.

2. Data Preprocessing:

• Removes unnecessary columns (test_date)

• Checks for duplicates and null values

• Converts categorical variables to numerical using Label Encoding

3. Feature Engineering: Separates features (symptoms, age, gender) from target variable (corona_result).

4. Model Training & Comparison: Tests multiple algorithms to find the best performer:

   • Logistic Regression (90.84% accuracy)
   • Random Forest Classifier (91.53% accuracy) Best
   • Naive Bayes Classifier
   • XGBoost Classifier (91.39% accuracy)

5. Visualization: Generates plots showing predictions, error distribution and feature correlations.

---

Dependencies

• pandas – Data handling and manipulation

• numpy – Numerical computations

• scikit-learn – Machine learning models and preprocessing

• xgboost – Gradient boosting classifier

• matplotlib – Data visualization

• seaborn – Statistical data visualization

• Jupyter Notebook – Interactive environment

Install them manually if needed:

pip install pandas numpy scikit-learn xgboost matplotlib seaborn jupyter

---

Dataset

• File: covid_data_2020-2021.csv.zip

• Description: Contains COVID-19 patient records with symptoms and test results

• Features:

  • cough – Presence of cough symptom (Yes/No)

  • fever – Presence of fever (Yes/No)

  • sore_throat – Sore throat symptom (Yes/No)

  • shortness_of_breath – Breathing difficulty (Yes/No)

  • head_ache – Headache symptom (Yes/No)

  • age_60_and_above – Age category (Yes/No)

  • gender – Patient gender (Male/Female)

  • test_indication – Reason for testing

  • corona_result – COVID-19 test result (Positive/Negative) [Target Variable]

---

Results

Model Performance Comparison:

Algorithm	Training Accuracy	Testing Accuracy
Logistic Regression	90.84%	90.84%
Random Forest Classifier	~100%	91.53%
Naive Bayes	Lower	Lower
XGBoost	~100%	91.39%

Best Model: Random Forest Classifier with 91.53% test accuracy

Visualizations Include:

• Actual vs Predicted comparison plots
• Error distribution charts
• Feature histograms
• Correlation scatter plots.

---

Model Insights

The Random Forest Classifier provides the best balance between training and testing accuracy, indicating:

• Strong predictive capability for COVID-19 detection

• Good generalization to unseen data

• Effective use of symptom-based features

• Minimal overfitting compared to other models.

---

Future Improvements

• Implement deep learning models (Neural Networks) for better accuracy

• Add cross-validation for more robust evaluation

• Include feature importance analysis

• Expand dataset with more recent COVID-19 variants

• Build a web interface using Flask or Streamlit for real-time predictions

• Add SHAP or LIME for model interpretability

• Implement ensemble methods combining multiple models.

---

License

This project is open-source under the MIT License.

---

Contributions

Contributions are welcome!

• Open an issue for bugs or feature requests

• Submit a pull request for improvements.