基于实时战略游戏重放记录数据编码和机器学习的游戏获胜者预测  

Winner prediction based on RTS game replay data encoding and machine learning

作　　者：王韦清 王全迪[1,2] 周杰[3] WANG Weiqing;WANG Quandi;ZHOU Jie(School of Mathematics,South China University of Technology,Guangzhou Guangdong 510641,China;School of Continuing Education,South China University of Technology,Guangzhou Guangdong 510641,China;Guangdong Key Lab of Communication and Computer Network(South China University of Technology),Guangzhou Guangdong 510006,China)

机构地区：[1]华南理工大学数学学院,广州510641 [2]华南理工大学继续教育学院,广州510641 [3]广东省计算机网络重点实验室(华南理工大学),广州510006

出　　处：《计算机应用》2021年第S01期87-92,共6页journal of Computer Applications

摘　　要：对实时战略(RTS)游戏仿真平台μRTS自带的RTS游戏AI机器人之间进行游戏比赛产生的重放记录数据进行采样,用独热编码对采样点数据中的游戏玩家在游戏中的状态和动作信息进行编码,利用卷积神经网络、支持向量机和K-近邻等机器学习算法对RTS游戏AI机器人在游戏比赛中的获胜者进行预测。实验结果表明,结合给出的编码方法和机器学习算法预测RTS游戏获胜者的准确率与已有方法相比有显著提高,预测结果ROC曲线的AUC值较高。The replay data generated by μRTS,a Real-Time strategy(RTS)game simulation platform,in a game match between build-in RTS game AI robots were sampled,the players’sampled state and action information were encoded by one-hot encoding method,and the winner of the RTS game AI robots in the game match was predicted by machine learning algorithms such as Convolutional Neural Network(CNN),Support Vector Machine(SVM)and K-Nearest Neighbor(KNN).Combined with the encoding method and machine learning algorithm,the experimental results show that the accuracy of prediction and AUC(Area Under the ROC(Receiver Operating Characteristic)curve)are significantly improved compared with existing RTS game AI robot’s winner prediction methods.

关 键 词：实时策略游戏 游戏AI机器人 重放记录数据编码 机器学习 游戏获胜者预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]