Correlation Between Taxonomy Data Distribution and Classification Performance

摘要

Web is providing tremendously increasing amount of information and mining it from a data on the Web has been recognized as critical issue in big data analytics. Classification is one of important technique for data mining, and performance of image or text data classification by machine learning has been investigated a lot. Our interest in this paper is ontological text data distribution to observe classification performance of the data. Classification performance on documents by ontological taxonomy is important to get information with intelligence that makes reasoning among identified objects. Analysis of data distribution of the ontological data was carried out by data clustering and Principle Component Analysis (PCA). We found that overlap distance of document clusters has a strong negative relationship to classification accuracy.