The present work reports the development of a nonlinear technique based on genetic algorithm (GA) for the prediction of wave heights in the north Indian Ocean. Time series observations of surface wind speed and significant wave height at three locations in the Arabian Sea (AS) and the Bay of Bengal (BOB) have been used for developing and testing the technique. The predictions have been compared with persistence forecasts and it has been found that the prediction by GA is always superior to persistence forecast and thus represents a net information gain. The predicted wave heights have been found to be consistent with the results obtained with autocorrelation analysis applied on the respective time series of wave heights.