Leveraging COVID-19 Predictive Models for Early Monkeypox Detection and Risk Assessment

Abstract

This research explores the adaptation of machine learning (ML) models successfully used in COVID-19 prediction to facilitate early detection and risk assessment of monkeypox. We hypothesize that the similarities in epidemiological characteristics and data availability between both diseases allow for the transfer learning of existing COVID-19 models. The study focuses on modifying and validating a pre-trained model using a newly curated dataset of monkeypox cases. Preliminary results indicate promising accuracy in monkeypox case prediction, offering a potentially faster and more efficient approach compared to traditional methods. The limitations of data availability and potential biases are acknowledged.

Introduction

Monkeypox, a zoonotic viral disease, experienced a significant global outbreak in 2022, highlighting the need for efficient surveillance and early detection strategies. The rapid spread and associated challenges presented by the COVID-19 pandemic demonstrated the crucial role of predictive analytics using ML. Given the similarities in transmission dynamics and data collection methods between COVID-19 and monkeypox (e.g., symptom reporting, geographical spread), adapting existing COVID-19 prediction models represents a potentially cost-effective and time-saving approach. Current monkeypox surveillance heavily relies on reactive measures, creating a critical gap in proactive prediction and risk assessment. This research aims to bridge this gap by leveraging the power of transfer learning applied to existing, well-established COVID-19 ML models.

Technical Details

• Utilizes transfer learning on pre-trained COVID-19 ML models
• Monkeypox dataset curated and preprocessed
• Implements classification and risk scoring for early detection
• Evaluates performance using accuracy, precision, recall, and F1 score