A Ranking Between Attributes Selection Models Using Data from NCAA Basketball Players to Determine Their Tendency to Reach the NBA

The present research work explores historic data from NCAA men’s basketball datasets with the aim of providing decision-makers with relevant information and improving their judgment when hiring. However, such datasets may contain redundant, noisy, and irrelevant features that could potentially have a negative influence on decision-making processes. In particular, we have addressed a feature selection question to identify player attributes that contribute the most to being chosen by an NBA team for a professional contract. In this context, this paper proposes a data mining approach using a feature selection method to identify relevant characteristics and rank features to assist stakeholders in decision-making processes. Thus, we have used a data mining approach in terms of a feature selection method by testing some models and performing a combination of genetic algorithm with decision tree and SVM. Feature selection is an important step in the data mining process that allows for obtaining better results using a lower number of features. To this end, we performed an experimental study to evaluate the performance of the proposed method with datasets from the NCAA repository, mainly comparing it with conventional feature selection algorithms from the categories Wrapper, Filter, and Embedding, using SVM and Decision Tree as classifiers. The results show that the proposed feature selection technique outperforms the other used methods on this feature selection problem. The results also show that in this specific model, the dataset was reduced from 65 features to only 6, and metrics such as accuracy, recall, precision, and F1-score showed the best values for both the proposed feature selection technique. Furthermore, we found that dbpm, treb, and ogbpm are the three best predictors to determine the tendency of a player to reach the NBA.