MPI - SAP Submission

Inspiration

We wanted to create a flexible multidimensional index to analyze international trade and provide impactful policy recommendations using Graph Neural Networks (GNNs).

What it does

The project builds a multidimensional index to study global trade patterns, using GNNs for flexibility and Explainable AI (XAI) for transparent policy recommendations.

How we built it

We used PCA for dimensionality reduction, Random Forest for feature selection, and developed a GNN to analyze trade. XAI with SHAP values helps explain the model's predictions.

Challenges we ran into

• Cleaning and imputing missing data
  
• Feature selection and deciding the most important indicators
  
• Constructing appropriate edges for the GNN

Accomplishments we're proud of

• Creating a flexible index
  
• Building an adaptable GNN model
  
• Using XAI for clear, transparent policy insights

What we learned

• The value of data preprocessing and feature selection
  
• Integrating GNNs and XAI for explainable predictions
  
• How to create flexible models that scale with complex datasets

What's next for MPI - SAP Submission

We’ll refine the model, improve its accuracy, and scale it for larger datasets, while continuing to focus on explainable policy recommendations.

Built With

• numpy
• python
• pytorch
• scikit-learn