An Intelligent Decision-Making System Based On Multiple Classifiers Updated Using Confidence Rates And Stress Parameters
CONTROL AND INTELLIGENT SYSTEMS(2011)
摘要
We develop an intelligent multi-classifier decision-making system for multi-class classification tasks. The proposed system called I-MDS (Intelligent Multiple Decision System) uses a dynamic scheme to combine the information provided by the individual classifiers and make a classification decision. The individual classifiers in the system are interconnected and use a negotiation scheme to come up with a unified decision. During the interactive and reactive negotiation process, individual classifiers are allowed to revaluate their confidence in their individual decisions and to respond to a system-wise stress parameter that keeps increasing as long as the system does not reach a unified decision. If after a certain number of negotiation rounds the system can not reach a unified decision, the input pattern is rejected. The proposed systems were tested on multi-class classification problems from the UCI repository and were shown to produce better classification rates and fewer misclassifications than majority voting combination technique.
更多查看译文
关键词
Pattern recognition, multiple classifier systems, multi-agent design, negotiation, confidence rate, stress rate
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络