This paper reports the application of an adaptive neuro-fuzzy inference system (ANFIS) to model the experimental results of heat transfer in an air-cooled heat exchanger equipped with classic twisted tape inserts. The aim is to consider the effects of the twist ratio of classic inserts and Reynolds number variation on average heat transfer in the air cooler. The training data for optimizing the ANFIS structure are based on available experimental data. A hybrid learning algorithm consisting of the gradient descends method and the least-squares method is used for ANFIS training. The proposed ANFIS was developed using MATLAB functions. For the best ANFIS structure obtained in this study, the maximum errors of the training and test data were found to be 0.111% and 2.378%, respectively. Also, the mean relative errors of the training and test data were found to be 0.011% and 1.316%, respectively. The obtained results showed that ANFIS can be used to predict experimental results precisely.