Predicting individual player performance, particularly scoring metrics like points per game (PPG), has become a significant area of research in sports analytics, driven by advances in data collection and machine learning. Previous studies primarily emphasized team performance or used aggregated data without in-depth feature optimization, leaving gaps in accurately forecasting individual player metrics. This paper addresses these limitations by introducing the Correlation-Optimized NBA Scoring Estimator (CONSE) model, designed to predict NBA player performance based on statistical data spanning ten NBA seasons (2013–2023). Our approach includes rigorous data preprocessing, extensive exploratory data analysis (EDA), and leverages correlation-based feature selection to enhance interpretability and predictive accuracy. Multiple regression models, including Multiple Linear Regression, K-Nearest Neighbors (KNN), Decision Tree, and Random Forest, were evaluated within the CONSE model framework using R-squared and Mean Absolute Error along with cross-validation. Random Forest Regression demonstrated superior performance due to its robustness against overfitting and outliers. Our findings provide valuable insights for coaches, analysts, and enthusiasts aiming to identify future NBA stars using a data-driven approach.