Credit Risk Analysis
Improved classification model accuracy by 18%, increasing performance from 65% to 77%, and analyzed the trade-off between accuracy and F1 score for predicting the minority class.
Machine Learning
Health Insurance Cost Prediction
Developed a predictive model that improved insurance forecast accuracy by 33%, significantly outperforming previous forecasting methods and enabling more accurate insurance planning.
CO₂ Emissions Forecasting
Developed and evaluated time series forecasting models; one-step ahead model achieved a 95% lower Mean Squared Error (MSE = 0.09) compared to dynamic forecasts (MSE = 2.01).
Feature Selection for Wine Classification
Built several classification models for predicting wine classes using a subset of features obtained by applying various methods of feature selection.