Level: Beginner
Vineela Datla, Student, University of Connecticut
Thore Koch, Student, University of Connecticut
Johana Rodriguez, Student, University of Connecticut
Anan Garg, Student, University of Connecticut
Assessing the estimation of a player can be helpful for the FIFA clubs while making contracts with players as they to work on a spending budget. The data scraped from sofifa.com has been sampled, explored, modified, modeled and analyzed using JMP. The data set has 88 different columns broadly classified as player information and performance statistics and each row belongs to individual FIFA players in 2019. In the sampling phase, the data is partitioned into 40-30-30 for training, validation, and testing. During the modification phase, 29 key performance attributes are featured in six primary attributes using principal component analysis. The data is feature engineered and analyzed using JMP formulas and distribution plots. The decision tree, K- nearest neighbors and neural network models are built to identify the best model to predict the value of a player. The models are then compared using the model comparison feature and the best model is identified based on its performance on test data. The model can best predict the values of low- to mid-range players given the performance attributes.