As part of a university project for the course Economic Demography, I developed an interactive educational website that simulates how populations change over time.
The core idea is to allow users to configure their own virtual country by setting specific parameters influencing fertility, mortality and migration. These inputs are processed internally using a cohort component model, resulting in a population development projection for up to 100 years. This project aims to make the mathematical foundations of demographic shifts easy to visualize and thus accessible to everyone.

In a team of four, I investigated the possibilities and limitations of Explainable AI (XAI) methods to make complex model decisions more transparent. The project focused on bridging the gap between the high predictive power of „black-box“ models and the need for human-friendly interpretability.
We developed a high-performance XGBoost model trained on a medical dataset to predict heart disease. To decode the model’s inner logic, we implemented and compared for major model-agnostic XAI methods using Python:

• LIME
• Anchors
• Shapley Values (SHAP)
• Counterfactual Explanations