The influence of team feature vectors on NBA championship team prediction

Abstract

Predicate National Basketball Association championship has always been the focus of people's attention. This paper makes an in-depth analysis of the prediction of the 2021-2022 National Basketball Association championship team. Select 22 possible influencing factors as feature vectors, such as 3-pointers, rebounds, and steals. SVM, KNN, and three different grouping methods were used to predict performance, including accuracy, recall, and F1-score. The result shows that the grouping method of a solid team and a weak team has a remarkable effect. Then principal component analysis was used to visualize the grouping methods of the four categories in two and three dimensions. The results showed that the champion team did not have significant characteristics. However, it was found that the teams were divided into two clusters, so the hypothesis was proposed: whether the league ball style changed with the year. Further verification found that the two clusters were not significantly related to the year. The conclusion is that under this method, the teams entering the playoffs can be selected with high accuracy, that is, the strong and weak teams in the experiment, but the characteristics of the champion team are not evident and difficult to predict.