Genetic algorithms (GAs) are a class of general optimization procedures randomized optimization heuristics based loosely on the biological paradigm of natural evolution. Artificial neural networks (ANNs) are well established optimization procedures in the domains of pattern recognition and function approximation, whose properties and training methods have been well studied. Recently there has been some successful applications of ANNs in sequential decision making under uncertainty (or stochastic control), where one's goal is the cost-to-go or cost function, which evaluates and guides management or control decisions in an organization. In this work we report on the integration of GAs and ANNs in terms of a new paradigm, the genetic algorithm based neural networks, taking the advantages of both approaches for time series forecasting of sunspots, airlines and production yields.