PREDICTION OF BASKETBALL ROOKIE LONGEVITY USING STACKED ENSEMBLE MODEL

Basketball players' longevity is influenced by various factors, including injury history and performance metrics, often leading to subjective and inconsistent assessments by scouts. This study addresses this challenge by developing and deploying a machine learning-based model that predicts player longevity using game data, providing an objective and data-driven approach to player evaluation. The primary goal is to enhance decision-making processes in player assessments and reduce reliance on subjective judgements. To maximize the predictive potential of the dataset, a hybrid feature selection strategy combining the Binary Bat Algorithm (BA) and Correlation Feature Selection (CFS) was implemented. This approach identified the most relevant features for predicting player longevity. Using the selected features, a Stack Ensemble Model was constructed, incorporating KNN, SVC, Random Forest, and Decision Tree as base learners, with Logistic Regression serving as the meta-learner. The system was developed using Python and integrated with Streamlit to create a user-friendly web application for practical deployment and real-time predictions. The model achieved a remarkable prediction accuracy of 98%, significantly outperforming the baseline accuracy of 80.2%, underscoring its robustness and effectiveness. This study highlights the critical role of injury history and performance metrics in predictive modelling and demonstrates the potential of machine learning techniques in sports analytics. By providing basketball management stakeholders with a reliable tool for assessing player potential, the study bridges the gap between traditional scouting methods and modern data-driven strategies, fostering more informed decision-making in player evaluations and talent management.