Article Preview
TopIntroduction
As a collective, comprehensive, and three-dimensional offensive and defensive game, basketball, originates from the process of human labor and survival and is a reflection of social and cultural progress (Choi et al., 2023). With the development of modern society, basketball has become an international mass competitive sport that integrates technology, culture, education, and skills (Montgomery et al., 2010). However, when analysts and coaches are evaluating the competitive performance of basketball players, traditional statistical data cannot fully reflect every aspect of their potential impact on future teams, and the simple addition of individual competitive abilities cannot accurately evaluate the overall competitive performance of basketball teams (Sansone et al., 2023). To address the aforementioned issues, we propose a video-based radial basis function (RBF) neural network competitive performance scoring method. By watching a large number of basketball videos of basketball players, analysts using this method can capture every subtle difference in their abilities and use discriminators to distinguish sports performance, achieving objective evaluation of players’ competitive performances, and thus adjusting training intensity accordingly. However, in traditional RBF networks, most of the adjustments to the prediction models of basketball players’ competitive behavior are based on subjective desires; these adjustments lack evaluation methods for the overall performance of basketball teams. To improve this issue, we propose three improvement methods based on the aforementioned challenges: the introduction of hidden layer neurons, network parameter adjustment, and hidden layer neuron deletion. Using these three methods will further improve the accuracy of basketball player competitive behavior prediction models. In this study, we used online learning models to predict the competitive behavior of basketball players and to verify the effectiveness of the improved RBF-EVA(Radial Basis Function Network - External Validation Approach) algorithm proposed in this article through comparative experiments with traditional prediction methods. The experimental results show that the improved RBF neural network that we designed for this paper has significantly better prediction performance than traditional convolutional neural networks. In summary, this study aims to improve the basketball player competitive behavior prediction model through a video-based RBF neural network competitive performance scoring method to provide more accurate and objective evaluation methods, and thus, provide strong support for the training and development of basketball players.
Combining the RBF neural network model with the EVA method, we established an RBF-EVA evaluation model to compensate for the shortcomings of traditional EVA methods in predicting player ability EVA and form a more scientific player ability evaluation method.
We applied the EVA method to the emerging industry of artificial intelligence, verified its applicability in the artificial intelligence industry, and further promoted and optimized the role of the EVA method in player ability assessment.