INFORMATION RECOMMENDATION BASED ON CONTROL THEORY AND AFFECTIVE COMPUTING
基于控制原理和情感计算的信息推荐

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 that 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.