A Method of Basketball Game Prediction Based on LSTM Neural Network

• Main Authors: Cheng-Yu Lee, 李承祐
• Other Authors: 黃德成
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/cgi-bin/gs32/gsweb.cgi/login?o=dnclcdr&s=id=%22107NCHU5394060%22.&searchmode=basic

碩士 === 國立中興大學 === 資訊科學與工程學系所 === 107 === Sport result prediction is a popular topic .The outcome of a contest depends on many complicated factors due to its uncertainty. It is an interesting funny topic of fan for contest prediction. In this paper, we applied LSTM Neural Networks to Build a predictive model and predict the result of basketball games in NBA league .We focus on 2017/2018 NBA season and select the key attributes which may affect the outcome of the contest. The selected features can be used for training, testing and result analysis. In terms of input data can be divided into two parts: the first part is to divide the data with three games, four games and five games as a number of test sets to predict the outcome of contest for the LSTM method. The second part is to conduct the set prediction with respect to all teams or a specific team. We can find the possibility with better prediction by testing several input modes and adjusting model parameters timely. The results show that the accuracy of specific team prediction is up to 71%.