基于控制原理和情感计算的信息推荐

INFORMATION RECOMMENDATION BASED ON CONTROL THEORY AND AFFECTIVE COMPUTING

在分布式合作信息推荐系统中 ,系统的目标是使信息尽量流动到需要该信息的用户手中 ,促进宏观信息的有效流动 ,这是一种有目的性的行为 ,可以抽象为复杂系统的控制问题 .由于系统的复杂性和动态性 ,集中式控制难以奏效 ,研究表明 :通过用户的局部 PI控制 ,可以获得很好的宏观控制质量 ,简单的控制规则可以在分布式信息推荐中产生很好的效果 .信息推荐需要基于用户的需求模型 ,鉴于基于符号空间建立用户需求模型存在的不足 ,给出了基于情感空间建立用户需求模型的方法 ,用户通过“很喜欢 ,喜欢 ,一般 ,不喜欢”等情感词汇表达自己的偏好 .仿真实验表明 :情感空间模型可以很好地表达用户的偏好 ,模糊化的局部 PI控制能够使计算机逐渐适应用户的需要 .

In a collaborative information recommendation system, the aim is to take the right information to the right users and enhance the efficiency of the macro information flow. This can be viewed as a purposive control problem of complex system. Since the system is complicated and dynamic, the centralized method is not effective. Our study shows that we can obtain global optimization by local PI(proportional and integral) control, and the simple control rule can be used well in distributed information recommendation. The content-based method has many shortcomings, therefore we give an affective method based on trust relationship. The user's profile is represented by the trust-value vector. The user can easily express his interest by 'like, neutralism, or dislike'. Simulation experiments show that the affective method can represent the user's profile well and the computer can adapt to the user's need gradually.