Skip to content

Latest commit

 

History

History

COVID-19 Detection Using Machine Learning: A Comparative Study

README.MD

Covid-19 Detection Using Machine Learning

Authors

  • Bhavin Patel
  • Kunj Kansara

Project Overview

The project focuses on developing a machine learning model to detect COVID-19 using various classification algorithms. The goal is to evaluate and compare the performance of different algorithms to identify the most effective method for predicting COVID-19 based on clinical and symptom data.

Introduction

  • Context: With the COVID-19 pandemic posing significant challenges globally, there is a need for effective and scalable detection methods. Traditional diagnostic tools may not be sufficient for rapid and widespread testing.
  • Problem Statement: The rapid transmission of COVID-19 highlights the need for efficient detection systems. This project leverages machine learning to explore novel solutions for early detection.
  • Aim: To compare multiple machine learning algorithms, including K-Nearest Neighbors (KNN), Random Forest, and Naive Bayes, and identify the best model for predicting COVID-19.

Datasets

The project uses the following datasets:

  1. Covid Dataset.csv:

    • Description: The primary dataset used for training and testing the machine learning models. It contains various features related to symptoms and clinical markers relevant for predicting COVID-19.

  2. Raw-Data.csv:

    • Description: This file contains the initial, unprocessed data before any cleaning or preprocessing steps. It is used as the starting point for data preparation.

  3. Cleaned-Data.csv:

    • Description: The dataset after preprocessing steps have been applied, including handling missing values, feature selection, and encoding. This cleaned data is used for model training and evaluation.

Methodology

  • Data Collection and Preprocessing:

    • Dataset: The dataset is sourced from Kaggle, focusing on symptoms and clinical markers.
    • Preprocessing:
      • Handling missing values.
      • Feature selection for identifying relevant predictors.
      • Encoding categorical variables for model compatibility.

  • Algorithms Used:

    • Logistic Regression
    • K-Nearest Neighbors (KNN)
    • Random Forest
    • Naive Bayes
    • Support Vector Machine (SVM)

Implementation

  • Tools and Libraries:

    • Programming Language: Python
    • Libraries: Pandas, Seaborn, Matplotlib, Scikit-learn

  • Exploratory Data Analysis (EDA):

    • Visualizations such as count plots and pie charts to understand the distribution of symptoms and their relationship with COVID-19 status.

  • Model Evaluation:

    • Evaluated models based on accuracy, confusion matrix, ROC-AUC score, and classification reports.

Results

  • Performance: The Random Forest algorithm demonstrated the highest accuracy among the tested models.
  • Recommendation: Random Forest is recommended as the most effective model for COVID-19 detection in this dataset.

Conclusion

  • Contribution: Provides a machine learning-based approach for early COVID-19 detection, potentially enhancing rapid intervention and risk assessment.
  • Implications: The findings offer valuable insights into improving detection methods and supporting healthcare systems in managing the pandemic.

Strengths

  • Comprehensive Analysis: Thorough comparison of various algorithms for a well-rounded understanding of their performance.
  • Effective Visualizations: Use of EDA to visualize and interpret the data distribution and its impact on detection.

Areas for Improvement

  • Dataset Limitations: Acknowledges potential issues with dataset completeness and representativeness, which could affect generalizability.
  • Bias Discussion: Limited exploration of dataset biases and their impact on model robustness in diverse healthcare contexts.