INTRODUCTION: KNOWLEDGE OF WAVE CHARACTERISTICS IS NECESSARY FOR DESIGN OF THE MARINE STRUCTURES. SINCE THE WAVE MEASUREMENTS ARE RARE OR LIMITED, HINDCASTING OF THE WAVE PARAMETERS IS USUALLY USED FOR PROVIDING LONG TIME RECORDS. THERE ARE SEVERAL METHODS FOR WAVE HINDCASTING SUCH AS EMPIRICAL, NUMERICAL AND SOFT COMPUTING METHODS. EMPIRICAL METHODS SUCH AS CEM AND SPM ARE SIMPLE, BUT THEY HAVE BEEN DEVELOPED FOR SPECIFIC CONDITIONS AND ARE NOT ACCURATE ENOUGH. NUMERICAL MODELS SUCH AS WAM AND SWAN ARE COSTLY AND REQUIRE HIGH SPEED COMPUTERS. SOFT COMPUTING METHODS SUCH AS ARTIFICIAL NEURAL NETWORKS (ANNS) AND CLASSIFICATION AND REGRESSION TREES (CART) REQUIRE LESS COMPUTATIONAL COST. IN THIS PAPER, SWAN AND ANN MODELS WERE USED FOR WAVE HEIGHT HINDCASTING IN THE PERSIAN GULF. BOTH MODELS WERE CALIBRATED (TRAINED) AND VERIFIED (TESTED) USING THE WIND SPEED INPUTS AND THE RESULTS WERE COMPARED.