This paper presents a soft methodology for predicting the 28-day compressive strength of Portland cement (CCS) by making use of the 1-day, 3-day and 7-day CCS values. Data taken from a cement plant in Turkey have been employed in the model construction and testing. For implementation, linguistic models were designed based on if-then fuzzy rules. In addition, predictions of these models were compared with results of the regression models. The performance evaluations showed that the linguistic-based fuzzy predictions are very satisfactory in estimating cement strength and the linguistic modeling performs better than regression modeling.