Some measured data matrices in projects are subjected to (not necessarily small) deterministic perturbations. Different from former Robust Fuzzy Tree model (RFT) which studies structured uncertainties, we propose another RFT to deal with unstructured uncertainties. This RFT uses fuzzy tree model (FT) and unstructured robust least squares (RLS) solution to work on the nonlinear modeling problem with unstructured bounded data uncertainties. The RFT not only keeps the features that FT can deal with high dimensional problem, has less computation load and has high precise, but also decreases drastically the sensitivity of FT to bounded uncertainties.