文献类型：专利

英文题名：Method for cleaning data classification training database based on mixed normalization, involves completing emptying of sample according to number of labeled samples, and obtaining final result

作者：YUAN Y;GU Y;RUAN T;ZHAO T;QIU W;PU D;GAO J;YIN Y

机构：[1]UNIV EAST CHINA SCI & TECHNOLOGY

申请号：CN106055613-A

申请日：2016-05-26

公开日：2016-10-26

语种：英文

收录：DERWENT

摘要：NOVELTY - The method involves pre-processing the input of a database, processing of missing value, and data collection pre-cutting of an abnormal value. The cutting is performed according to a rank and the category number of the data set by analyzing and removing the abnormal value. The mixed normalization is used to extract the representative samples from the pre-processed database. The best sample is selected, using an orthogonal technology for sample correction. The emptying of the sample is completed according to a number of labeled samples, and a final result is obtained. USE - Method for cleaning data classification training database based on mixed normalization. ADVANTAGE - The modeling time and memory space of data classification are significantly reduced, and the learning efficiency is improved. The cleaning efficiency index is given. The cleaning efficiency is made higher by carrying out a data classification test, and the cleaning efficiency reaches more than 140. The number of training sample and dimension is reduced in large-scale. DESCRIPTION OF DRAWING(S) - The drawing shows a flowchart illustrating the process of cleaning data classification training database based on mixed normalization. (Drawing includes non-English language text)