- 學位論文
籃球分差預測與博弈分析模型
A Model for Basketball Game Point Difference Prediction and Odds Analysis
若您是本文的作者，可授權文章由華藝線上圖書館中協助推廣。
摘要
預測比賽結果可幫助賭徒賺取可觀的財富，亦可幫助莊家設定合理的賠率減少風險。因此，在本研究中，我們提出了一個新穎的研究框架叫作籃球賠率分析模型，並以此模型預測籃球比賽的分差並進行賠率分析。我們所提出的研究框架包含四個階段，首先，將球員與球員的互動關係表示成一個球員關係圖，並將球隊與球隊的互動關係表示成一個球隊圖，然後利用關係圖卷積網路聚集相關球員的資訊，並利用圖卷積網路聚集相關球隊的資訊；接著，運用協同注意力機制結合球員與球隊的資訊；再利用門控循環單元累積每個球員與每個球隊的績效；最後，使用多頭自注意力機制與多層感知器預測比賽的分差，並根據預測的結果進行賠率分析。實驗結果顯示，我們提出的研究框架在平均絕對誤差、均方根誤差、準確性和增益方面都優於最先進的方法。實驗結果也表明了我們模型引入的圖卷積網路、門控循環單元和注意力機制都有助於推進比賽結果預測和博弈分析領域的研究。另外，我們的研究框架也可以提供有價值的預測見解來幫助籃球隊、教練和分析師做出數據驅動的決策。
並列摘要
Predicting game outcomes can help gamblers make a lot of wealth from lottery and bookmakers set odds more robustly and reduce risk. Therefore, in this study, we propose a novel Basketball Odds Analysis Model, called BOAM, to predict the game point differences of basketball games and conduct odds analysis. The proposed framework contains four phases. First, we construct a player graph to represent the interactions among players and a team graph to represent the interactions among teams, where the interactions between players include competition and cooperative relationships. Then, we employ the relational graph convolutional network to aggregate the structural information in the player graph and the graph convolutional network to aggregate the structural information in the team graph. Second, we use the co-attention mechanism to fuse the player and team feature vectors together. Third, we use the gated recurrent units to accumulate the performance for each player and each team. Last, we exploit the multi-head self-attention mechanism and multi-layer perceptron to predict the game point difference for each game and conduct odds analysis based on the predicted results. The experimental results show that the proposed model outperforms the baseline models in terms of mean absolute error, root mean square error, accuracy, and earnings gain. It also shows that the graph convolutional network, gated recurrent units, and attention mechanism introduced by our model are all helpful to enhance the performance of game outcome prediction and odds analysis. Also, our model can provide valuable insights for decision makers in the basketball domain to implement their game strategy, player rotation, and tactical decisions.
參考文獻
Agrawal S, Singh SP, Sharma JK (2018) Predicting results of Indian premier league T 20 matches using machine learning. Proceedings of the International Conference on Communication Systems and Network Technologies. 67–71.
Aoki RYS, Assuncao RM, Vaz de Melo POS (2017) Luck is hard to beat: The difficulty of sports prediction. Proceedings of the International Conference on Knowledge Discovery and Data Mining. 1367–1376.
Basit A, Alvi MB, Jaskani FH, Alvi M, Memon KH, Shah RA (2020) ICC T20 cricket world cup 2020 winner prediction using machine learning techniques. Proceedings of the International Multitopic Conference. 1–6.
Breiman L (2001) Random forests. Machine Learning. 45(1):5–32.
Cunningham P, Delany SJ (2021) K-nearest neighbour classifiers - A tutorial. ACM Computing Surveys. 54(6):128:1-128:25.