Research on competition score prediction based on GA-BP neural network model and RBP inverse neural network model

This paper presents a study on competition score prediction using a hybrid GA-BP neural network model and an RBP inverse neural network model. The GA-BP neural network model, which combines genetic algorithms with BP neural networks, is established with "score" as input and "prediction of scoring" as output, integrated with the calculation of "objective scores" to assess accuracy. Upon achieving the set standard, the RBP inverse neural network model is developed to determine which factors are most relevant to the volatility of the game. This study utilizes these relevant factors as input features and inflection points as output features, with the top three factors being the irresistible ball, serving factors, and process factors. Additionally, an index is introduced to summarize the comprehensive impact of environmental factors, technical proficiency, and psychological traits on game volatility. Experimental results demonstrate an improvement in both Precision and F1 values, indicating better accuracy for positive class samples with the new model. Thus, it is evident that the model possesses good adaptability and compatibility.