This paper presents an overview of the development of the numerical wave prediction models and recently used neural networks for wave hindcasting and forecasting. The numerical wave models express the physical concepts of the phenomena. The performance of the numerical wave model depends on how best the phenomena are expressed into the numerical schemes, so that more accurate wave parameters could be estimated. There are still scopes for improving the numerical wave models. When exact input-output parameters are known for the same phenomenon, it can be well defined by the neural network. Hindcasting of ocean wave parameters using neural networks shows its potential usefulness. It is observed that the short-term wave predictions using neural networks are very close to the actual ones. Ocean wave forecasting using recurrent network yields very high correlation coefficients. The correlation coefficients for the lead of 3, 6, 12 and 24 hours are 0.95, 0.90, 0.87 and 0.73 respectively. It is also observed that the neural network simplifies not only the complex phenomena, but also predicts fairly accurate wave parameters.