機器學習自行車賽預測及人工智慧信任度應用

Machine learning in cycling race prediction and the application of trustable AI

劉皇慶(Huang-Ching Liu)

指導教授 ： 張瑞益

國立臺灣大學/工學院/工程科學及海洋工程學研究所/碩士(2022年)

https://doi.org/10.6342/NTU202201243

摘要

使用機器學習分析運動賽事相當普遍，自行車賽事便是其中一項。然而傳統機器學習只提供模型，並未考慮到模型每一次預測的信任度。隨著人工智慧信任度的議題逐漸崛起，找到一個指標或是標準來衡量模型的信任度變得十分重要。本文建立一個自行車賽事賭局系統，使用者能給定一個輸入金額，系統會依照機器學習的預測結果進行下注，最後得到一個獲利的輸出金額，此外我們更提出一個評估模型信任度的算法，並將其應用在優化獲利的輸出金額。資料選自2014年至2018年的車手資料，預測目標為環法比賽的前十名。在機器學習的模型實驗成果上，在預測2019年的環法自由車賽中，top10等於1的精確率能達到0.62，並達到獲利。在加入預測的信任度參考後，在獲利為正的時候，獲利最大能增長約40%。

關鍵字

自行車 ； 機器學習 ； 人工智慧信任度 ； 運動分析 ； 體育博彩

並列摘要

It is quite common to use machine learning to analyze sports events. Cycling races are one of them. However, conventional machine learning provides only the model without its trust in each prediction. As the issue of trustable AI gradually rises, it becomes very important find an indicator or standard to measure the trust of the model. This article establishes a cycling race gambling system. The user can give an input amount, the system will bet according to the prediction result of machine learning, and finally an output amount is given. In addition, we also propose an algorithm for evaluating the trust of the model, and its application is to optimize our output amount. The cyclist data is selected from 2014 to 2018, and the predicted target is the top ten in the Tour de France. In the experimental results of the machine learning model, the precision rate of top10 equal to 1 can reach 0.62 in predicting 2019 Tour de France. In the part of optimizing the output amount, after adding the reference of the trust score in each prediction, the maximum profit can increase by about 40% when the original profit is positive.

並列關鍵字

Cycling ； Machine Learning ； Trusted AI ； Sports Analytics ； Sports Betting

參考文獻

[1] M. C. Gabriele, The Golden Age of Bicycle Racing in New Jersey. The History Press, 2011.

[2] A. Soneja. (2019). Artificial Intelligence in Sports: A Smarter Path to Victory. Available: https://www.cio.com/article/3400877/artificial-intelligence-in-sports-a-smarter-path-to-victory.html

[3] S. DAS. (2019). Trust Issues: Is AI Black Box Creating A Black Future? Available: https://analyticsindiamag.com/trust-issues-is-ai-black-box-creating-a-black-future/

[4] S. Gilchrist. (2020). A Brief History Of Cycling and Cycling Betting. Available: https://kayokokishimoto.com/cycling-betting/

[5] F. Thabtah, L. Zhang, and N. J. A. o. D. S. Abdelhamid, "NBA game result prediction using feature analysis and machine learning," vol. 6, no. 1, pp. 103-116, 2019.