The social concern with communication between a consumer and maker has been growing for the last several years. A large number of unspecied users are investigating Consumer Generated Media's communication on the Internet. The claim has the tremendous impact on the purchasing behavior in Consumer Generated Media. This paper is intended as an investigation of signicant feature quantity for topic forecast. In this Paper, We investigate online media's topic forecast on Online Community article. We comprehensively extracted text information's feature quantity using existing research. Feature quantity is used, for 31 types and 2,071 dimensions in text information. Proposed Method consists of feature transformation and the base on evaluation using Nonnegative Matrix Factorization (NMF). The validity of the Proposed Method is veried by Support Vector Regression (SVR) and Classiers. It was found from the evaluative result that have an impact on view count. We report that research result.