Shay Samat

I just finished my Winter quarter at UC San Diego and was able to create 2 Machine Learning Final Projects. My first project was an Unsupervised Machine Learning project where our group looked into FIFA 24 players. We employed dimensionality reduction and clustering techniques to group players into positions based on their playing styles and attributes. By leveraging techniques like UMAP and Gaussian Mixture Models, we achieved promising results with an ARI score of 0.498 and a silhouette score of 0.62, indicating a fair amount of similarity between predicted and true player clusters and well-separated clusters conducive to effective modeling. These findings demonstrate the efficacy of using data-driven approaches to accurately categorize FIFA players into distinct positions, providing valuable insights for team management and strategic decision-making in virtual soccer environments. I also worked on a solo project I developed and trained various neural network architectures to predict player performance metrics such as points per game (PPG), assists per game (APG), and rebounds per game (RPG) for the 2021-2022 season and tested on the 2022-2023 season. By experimenting with different model complexities and optimization techniques, I achieved notable results, with the best-performing models demonstrating high accuracy and low mean squared error (MSE) values. These findings underscore the effectiveness of neural networks in forecasting NBA player performance, providing valuable insights for teams, analysts, and stakeholders in optimizing player selection and team strategies.