Abstract: Aiming at the problem of measurement level output, an information fusion algorithm based on multi-agent theory is presented. The concept of fancy judgment matrix is given, and an integration method of multiple classifiers based on adaptive weight adjusting is presented. Adaptive weight adjusting fusion method adaptively assigns weights to classifiers based on the sample. According to the experiments on standard database, this algorithm leads to less error than other methods and individual classifier. Experiments show that the algorithm is convergent.

Key words: classifier fusion, fancy judgment matrix, weight

摘要： 针对度量层输出的多分类器融合，该文提出一种基于Multi-agent思想的融合算法。该算法给出样本集在多分类器下的偏好判断矩阵概念，可以根据各个样本的具体情况自适应地为各分类器赋予权值。实验证明，该算法可得到比其他方法更低的分类错误率。

关键词: 分类器融合, 偏好判断矩阵, 权值

CLC Number: 

• TP312