A data-driven Takagi-Sugeno fuzzy model is developed for modeling a real plant situation with the dependent inputs and the nonlinear and time-varying input-output relation. The collinearity of inputs can be eliminated through the principal component analysis. The TS model split the operating regions into a collection of IF-THEN rules. For each rule, the premise is generated from clustering the compressed input data, and the consequence is represented as a linear model. A post-update algorithm for model parameters is also proposed to accommodate the time-varying nature. Effectiveness of the proposed model is demonstrated using real plant data from a polyethylene process.