摘要
食品卫生的HACCP自动分类要处理的数据集形状呈现多样性,对分类结果的准确性和专业性要求很高,已有的算法难以满足。该文基于经典BIRCH算法,结合多阈值思想和多代表点特征树思想,提出多阈值多代表点的BIRCH算法,增加了专业分类知识的指导,并对每一个代表点设立单独的阈值,使得该算法能适应各种形状的数据集,减少了聚类特征树重建次数,提高了算法的效率。
The HACCP data of food shows diversity shapes, its classification results on the accuracy and professionalism. The existed algorithms have been difficult to meet it. Based on the classic BIRCH algorithm, and the existed two algorithms multi-threshold and multi-representation points CF tree, a new multi-threshold and multi-representation BIRCH algorithm is designed, and the professional knowledge of the classification is added to guide set different variable thresholds to every representation points. Thus, the new algorithm can meet diversity data shapes, reduce the times of reconstruction of the CF tree, and improve the efficiency of the algorithm.
出处
《计算机工程》
CAS
CSCD
北大核心
2008年第23期59-61,共3页
Computer Engineering
基金
上海大学研究生创新基金资助项目"食品安全法规
标准文献信息平台建设--乳制品项目"(A.16-0108-07-001)
关键词
BIRCH算法
聚类特征树
多代表点
多阈值
BIRCH algorithm
cluster feature tree
multi-representation point
multi-threshold
