Research On Performance Prediction Of NBA Competition Results Based On BP Neural Network

NBA basketball,as one of the most influential sports events in the world,is loved by the majority of basketball fans,and is also closely concerned by the world.At the same time,the basketball gambling market has also been developing rapidly,which makes the prediction of the outcome of basketball games become a challenging new task.In recent years,with the rapid development of big data technology,data acquisition and collation are easier and faster than before.It has become one of the important research directions of sports events to obtain sufficient data through big data technology,and then predict the results of competitions through data mining algorithm.This paper reviews the research status of NBA basketball game results in related literature at home and abroad,and the application status of BP neural network algorithm in various fields.It is found that at present,not only big data technology is rarely used in the sports industry,but also for the existing NBA game prediction papers,there are generally fewer samples and low accuracy.Therefore,this paper takes the NBA basketball game as the research object,because there is a nonlinear complex relationship between the indexes that affect the results of the NBA game,and the traditional prediction method has been difficult to deal with the decline of prediction accuracy caused by this complex nonlinear relationship.In this paper,the BP neural network model with reliable statistical theory and better processing ability for nonlinear relationship is selected.This model is very suitable for NBA prediction.Therefore,based on the BP neural network to predict the outcome of NBA games,the specific research content includes the following aspects:(1)Download the regular season and playoff data of NBA data website from 2008-2009 season to 2018-2019 season through Python crawler technology,and preprocess the data and indicators by removing some features,feature structure,missing value processing and standardized methods.In this paper,taking 2018-2019 season as an example,the data set is divided into training data set and test data set,A three-layer BP neural network model is trained to predict the outcome of NBA games,and then the accuracy rate,accuracy rate,recall rate and F1 value in the season are analyzed according to the algorithm evaluation index.(2)This paper selects NBA game data to use statistical evaluation indexes in machine learning algorithm for comparative analysis.The algorithm includes logistic algorithm,naive Bayes algorithm,k-nearest neighbor algorithm,SVM algorithm and BP algorithm.The comparative results show that the accuracy of BP neural network algorithm in each season is more than 80%,which is better than that predicted by other algorithms It shows that BP neural network algorithm has good practicability in NBA games.(3)In this paper,the trained BP neural network model is applied to the dynamic prediction of NBA games,and May 12,2019 is selected According to the dynamic data in the ongoing NBA game of 76 vs.raptors,the game results at each time point are predicted according to the data changes in the process of the game,and the dynamic prediction ability of the algorithm is realized,which has more practical application value in the basketball gambling market.