Donia Besher | Data Science Portfolio

Data Science and Artificial Intelligence Portfolio

Welcome to my Portfolio!

I am Donia Besher, a Sorbonne Mathematics graduate specializing in Data Science for Artificial Intelligence. My journey began with late nights solving math puzzles just for fun, but it quickly grew into a passion for uncovering hidden signals in complex data. Along the way, I have forecasted exchange rates, explored climate policy uncertainty, and learned that sometimes the most valuable insights come from models that don’t perform as expected. One of my favorite moments was watching my ranking swing wildly with every submission in Kaggle competitions, and finally hitting the top spot after countless tweaks. It shows that persistence, a bit of luck, and maybe some extra caffeine can matter more than perfect conditions. Small victories like that are what make me love this field: data science is as much about experimenting and bouncing back as it is about generating insights and solving problems. Looking ahead, I am excited to explore spatiotemporal modelling and contribute to more informed decision-making in uncertain environments.

Projects Portfolio

Exchange Rate Forecasting

Developed the NARFIMA (Neural AutoRegressive Fractionally Integrated Moving Average) model to forecast exchange rates of BRIC economies. NARFIMA integrates neural networks with fractional differencing to capture nonlinear patterns and long-term dependencies in financial time series.

View the Exchange Rate Forecasting Project | Read the Exchange Rate Forecasting Paper

Climate Policy Uncertainty Forecasting

Investigated drivers of the US Climate Policy Uncertainty index using macroeconomic, financial cycle variables, and real-time public sentiment data to improve forecasting.

View the Climate Policy Forecasting Project | Read the Climate Policy Forecasting Paper

Will They Get the Job?

Winning Kaggle machine learning model to predict student job placement outcomes from a highly imbalanced dataset with limited samples, nonlinear separability, and multicollinearity.

Read the Machine Learning Project | Download Will They Get the Job? PPTX

Credit Card Debt Prediction

Winning Kaggle ensemble regression model for predicting credit card debt based on socio-economic factors dataset with outliers, nonlinearity, multicollinearity, and nonnormality.

Read the Regression Project | Download Credit Card Debt Prediction PPTX

Analysis of COVID-19 Mortality Data

Probabilistic analysis of COVID-19 mortality data to identify the underlying distribution and provide insights for pandemic management strategies.

Read the Analysis Project | Download Probabilistic Analysis of COVID-19 Mortality Data PPTX

Publications and Research Software

Publications

Chakraborty, T., Besher, D., Panja, M., Sengupta, S. (2025+) Neural ARFIMA Model for Forecasting BRIC Exchange Rates with Long Memory under Oil Shocks and Policy Uncertainties. Submitted. arXiv:2509.06697
Besher, D., Sengupta, A., Chakraborty, T. (2025+) Probabilistic Forecasting of Climate Policy Uncertainty: The Role of Macro-financial Variables and Google Search Data. Submitted. arXiv:2507.12276

Research Software

NARFIMA R Package: Developed an R package for time series forecasting using the NARFIMA model, which integrates neural networks with fractional differencing to capture nonlinear patterns and long-term dependencies. Supports seasonal adjustment, Box-Cox transformations, exogenous variables, simulated prediction intervals, and benchmarking against alternative models (NARIMA, NBSTS, NNaive).