We address the forecasting of spot electric prices governed under a competitive open market. Considering normalized hourly returns for the NordPool data corresponding to spot prices during 1992-2004, it is observed that a reasonably good forecast of consumption and prices within the next 24 hours plays a very important role. Evidently the time series of prices show marked periodicities interspersed with rather pronounced chaotic patterns, and hence we have found it useful to separate the deterministic trend from the random fluctuations using a wavelet based approach. It is then shown that the deterministic part can be modeled quite meaningfully using a Genetic Programming based approach where the embedded time delayed technique for reconstructing the state of the system has been incorporated. The fluctuations may then be modeled separately, if required, using a statistical approach. The building up of the analytical model of the real system is accomplished by carrying out the two steps, namely 1) embedding the time delayed vectors by extracting the relevant dimension and the time lag of the attractor, and 2) building up the predictive map that connects the future of the time series in terms of its components in the immediate past.