Fabian Steinmetz

Fabian Steinmetz

CS & Math Undergraduate | Minerva University

I am a Computer Science and Mathematics junior at Minerva University. I'm curious and learning about the mathematical foundations of Bayesian statistics, Machine Learning, and Monte Carlo Methods — as well as their applications in epidemiology, climate systems, and big data. From June to August, I'll be working as a summer researcher at CERN!

Selected Academic Projects

Wildfire Risk Map

Wildfire Spread & Intervention Analysis

Python | NetworkX • Numpy

According to the European Forest Fire Information System (EFFIS), large parts of Western Germany carry high-risk wildfire classification codes. I implement a stochastic network simulation on NASA grid data to identify high-risk areas and empirically evaluate several intervention strategies.

Methods: Cellular Automaton • Monte Carlo • Network Analysis • Renormalization Analysis • Discrete-time simulation • Stochastic dynamics • Mean-field approximation
[ Read Report ] [ View Code ]
Berlin Traffic Simulation

Berlin Traffic Modeling

Python | OSMnx • SciPy • Networkx

Does open-source street data suffice in the modeling of congestion on real-world street networks? I simulate congestion flow dynamics in Berlin through a discrete time algorithm on Open-Street Maps data and validate mathematical relationships on another real out-of-sample road network.

Methods: Monte Carlo • Discrete-time simulation • Empirical Convergence • Real Network Analysis • Nagel-Schreckenberg Model • Out-of-sample validation
[ Read Report ] [ View Code ]
Predator Prey Patterns

Predator-Prey Ecosystem Sensitivity

Python

What tipping points can be identified in a simple ecological reaction-diffusion model? I implement a 2D cellular automaton to simulate predator-prey dynamics and analyze spatial structure facilitateing species coexistence through sensitivity analysis and mean-field approximation.

Methods: Cellular Automaton • Sensitivity Analysis • Monte Carlo Simulation • System Convergence • Mean-field approximation
[ Read Report ] [ View Code ]
Bayesian Football Model

Bayesian Modeling of Football Attendance

Python | PyMC • Arviz

How can we use Bayesian statistics to model and compare the effect of distinct processes on a dataset with incomplete data on the outcome variable? I use Bayesian hierarchical modeling, model comparison, MCMC sampling and imputation to analyze a real-world dataset on sports match attendance.

Methods: Bayesian hierarchical modeling • MCMC sampling (NUTS) • Model comparison (WAIC/LOO) • Data Imputation • PPC
[ Read Report ] [ View Code ]
Education Inequality Regression

Educational Inequality in Argentina

Python | PyMC • Arviz

How does Argentina's low Gini coefficient of 0.42 manifest in varying education outcomes? Using mathematics test score data, I compare multiple Bayesian models handling heavy-tailed data.

Methods: Bayesian robust regression • Likelihood Comparison • Reparameterization • NUTS Sampler • Hierarchical modeling • Posterior inference on SES effects
[ Read Report ] [ View Code ]
SQL FSM Diagram

SQL to CSV Streaming Converter

Python • XML | regex

Developed a memory-efficient streaming parser converting large (5GB+) SQL dumps (.sql suffix) into csv-files. Handled multi-line inserts and dynamic quoting while batch-processing for higher memory-efficiency.

Methods: Streaming architecture • SQL Dump • Regex-based pattern matching • Batch processing • Memory complexity optimization
[ View Code ]

Working Papers & Current Research

Does Going Slower Really Make Us Safer?

Minerva University R • Python | fixest Co-authors: K. Falsen and C. Dinuta

A difference-in-difference approach on the Effect of Speed Limit Reductions on the Collision Rate in Wales (2023). We are using a Two-Way Fixed Effects Model after establishing parallel trends. .

[ Read Draft ] [ View Code ]

Applications of Python ML & Decision Tree Reconstruction on NPS Open Data

National Taiwan University Python | scikit-learn Co-authors: Prof. Ma, R. Li

Developing ML pipelines (Random Forests) to generate sustainability indicators for smallholder farms using National Panel Survey data.

[ View Lab Presentation ]