In this paper an effective modeling approach for building heating dynamics is presented. The proposed model is a so-called fuzzy black-box model. This data-based model is able to reproduce the nonlinear dynamic behavior of a complex office building split into different zones. The fuzzy black-box model is based on local linear models (LLM), for which a 2-dimensional partition space is obtained. In the proposed work data pre-processing is one of the most important tasks and thus extensively explained. This includes the selection of suitable input and output data and selection of appropriate model orders. Furthermore, it is presented that even with a small amount of data good models can be achieved. The output of this LLM algorithm is an autoregressive model with exogenous input (ARX). The resulting model is specifically tailored for model-based control design. The uniqueness of the presented work lies not only in the low-order of the building model, but also in the way of retrieving a model in the most effective and time-saving way. (C) 2015 Elsevier B.V. All rights reserved.