Contributors: Varun Jayathirtha, Aditya Bharadwaj
• Built a player transfer prediction model as part of the Automated Learning & Data Analysis course
• Analyzed a European Soccer Database with data across the top 11 leagues spanning over 8 years. Original data set can be found here: https://www.kaggle.com/hugomathien/soccer
• After pre-processing the data, we used multiple techniques for classifying players into their playing positions, namely t-SNE (for dimensionality reduction) with clustering algorithms like K-Means and DBSCAN applied on player skills and a Decision tree built over the players’ (X, Y) co-ordinates
• Uncovered the weakest region in the team and the corresponding weakest player using player skills/attributes giving us the likeliest outgoing player transfer
• Predicted suitable replacements for the outgoing transfer based on multiple distance metrics like Manhattan and Euclidean distance to give us likeliest incoming player transfer.
• Implemented using Python, R and DB Browser for SQLite