Data analysis of footballers to find similarity using KNN and Streamlet framework

Abstract

Predicting in sporting fields has kept its landmark in Machine Learning by analyzing the statistics and other indicators to predict a certain result as desired. The result usually shows the prediction as the user inputs the value to know about their similarity with the professional figures. We collected the dataset from different sources like FIFA21 and Kaggle to determine the best dataset for our prediction. It obtains the players’ names and other indicators to show their capabilities and position where they usually play. The previous works and hypotheses were analyzed and the implementation of KNN was done after getting it as the best algorithm among K-Means, SVM, ADASYN which is a custom model of Linear Regression, and KNN considering parameters like accuracy, precision, recall, and F1-score. We found 92%, 85%, and 87% accuracy in KNN, ADASYN, and SVM respectively in finding similarity with the professional players from the user inputs. K-Means didn't go well with this type of dataset and aim and failed to execute a result. With the best algorithm found here, KNN, we built a WebApp using the StreamLit framework and ensured four functions with the essential model. We can 1) find similar player, 2) find position, 3) know players’ information, and 4) compare your ratings with any player from the WebApp. We applied these functionalities in all four algorithms and found KNN giving the best value of each parameter. This WebApp will be beneficial to dream of being a footballer and also lead allto a better version of their selves along with storing data in the dataset for futureexperiments.