Abstract Prediction of tidal current in the coastal region is an important activity in marine science. It is useful in taking operation- and planning-related decisions such as towing of vessels and monitoring of oil slick movements. It is also useful for fisheries and recreational activities. General practice is to carry out this prediction using harmonic analysis or numerical hydrodynamic models. However, both these methods have their own limitations and nonlinear data adaptive approaches are gaining increasing acceptance. In this paper, such an approach, known as genetic algorithm (GA), has been employed for this prediction. A preliminary empirical orthogonal function (EOF) analysis has been used to compress the spatial variability into a few eigenmodes, so that GA could be applied to the time series of the dominant principal components (PC). The multivariate version of GA has been used to carry out the forecast using a few tidal levels at the boundary of the domain of study as inputs. The performance of this combined technique has been found to be quite satisfactory.