It all began with my curiosity about how machine learning models work after joining a data science competition at my university. We were tasked with creating a fraud classifier model, and my team decided to use an ensemble model consisting of XGBoost (XGB) and LightGBM (LGB). We achieved around 28% accuracy, but unfortunately, my team didn’t make it to the finals. Later, I noticed announcements that similar models could be built with neural networks. I realized these models could be exported ...