Abstract: Player modeling is an important research area in game design field. Foreign scholars have proposed the modeling method based on statistics, but this kind of method needed to build the mathematic model subjectively, which is lack of objectivity. This paper proposed a quantitative method, which was based on the evolution neural network, to build the nonlinear model between game data and the players emotional preferences. The experimental results show that the model constructed by this method is of high efficiency and accuracy.

Key words: Feature selection, preference learning, evolution neural network, player affective modeling

摘要： 玩家建模是游戏设计领域中一项重要的研究内容，国外学者已经提出了基于统计学的建模方式，但是这类方法需要主观构建数学模型，缺乏客观性。提出一种基于进化神经网络的定量建模方法，构建从游戏数据到玩家情感偏好之间的非线性模型。实验结果表明，该方法所建模型具有客观、高效的特点，并且具有较高准确度。

关键词: 特征选择, 偏好学习, 进化神经网络, 玩家情感建模