Predicting Dengue Fever Outbreaks in Sri Lanka

Challenge Background

Dengue fever is a mosquito-borne viral infection that poses a significant public health threat in Sri Lanka. The country has faced several severe outbreaks in recent years, with 2017 seeing over 186,000 reported cases and 320 deaths. The disease is endemic in Sri Lanka, with peak transmission typically occurring during the monsoon seasons (May-July and October-December). Key factors influencing dengue outbreaks include:

1. Climate conditions (temperature, rainfall, humidity)
2. Urbanization and population density
3. Mosquito breeding sites
4. Virus serotype circulation
5. Public health interventions Current early warning systems in Sri Lanka are limited in their predictive power and often rely on case reports, which can delay response times.

The Problem

The core problem this project addresses is the reactive nature of current dengue fever management in Sri Lanka. Despite being endemic, dengue outbreaks often catch local health systems off-guard, leading to overwhelmed hospitals, shortage of resources, and preventable deaths. The key issues are:

1. Late Detection: Current systems rely heavily on reported cases, which means response efforts begin only after an outbreak is already underway.
2. Resource Misallocation: Without accurate predictions, health authorities struggle to allocate resources efficiently, leading to shortages in high-risk areas and waste in low-risk areas.
3. Limited Preventive Action: The lack of reliable early warning systems hinders proactive measures like mosquito control and public awareness campaigns.
4. Data Silos: Relevant data (health, climate, demographic) are often scattered across different government departments, making holistic analysis difficult.
5. Complexity of Factors: The interplay of various factors influencing dengue outbreaks (climate, urbanization, virus serotypes) is complex and challenging to model using traditional statistical methods.

Goal of the Project

• Develop a machine learning model that can predict dengue outbreaks 2-3 months in advance with at least 80% accuracy.
• Create a user-friendly interface for health officials to access and interpret predictions.

Project Timeline

What you'll learn

1. Data Collection and Preprocessing:

• Gather historical dengue case data from the Sri Lankan Ministry of Health.
• Collect climate data (temperature, rainfall, humidity) from meteorological departments.
• Obtain population density and urbanization data from census reports.
• Clean and preprocess the data, handling missing values and outliers.

2. Feature Engineering:

• Create relevant features such as lagged climate variables, population mobility indices, and seasonality indicators.
• Explore and validate relationships between features and dengue incidence.

3. Model Development:

• Experiment with various machine learning algorithms (e.g., Random Forests, Gradient Boosting Machines, LSTM networks).
• Implement time series cross-validation to assess model performance.
• Fine-tune hyperparameters to optimize predictive accuracy.

4. Model Interpretation and Validation:

• Use techniques like SHAP (SHapley Additive exPlanations) values to interpret model predictions.
• Validate the model's performance on held-out test data.
• Conduct sensitivity analyses to understand the model's robustness.

5. Deployment and Visualization:

• Develop a web-based dashboard using tools like Dash or Streamlit.
• Implement real-time data ingestion pipelines for ongoing predictions.
• Design intuitive visualizations of outbreak risks and contributing factors.

6. Collaboration and Communication:

• Work closely with local health officials to understand their needs and incorporate feedback.
• Prepare clear documentation and conduct training sessions for end-users.
• Present findings to both technical and non-technical audiences..