I have some experience with neural networks and genetic programming although its been a few years since I've worked with them. You may want to look through the paper at the following link:
Time Series Prediction Using Evolving Polynomial Neural Networks
The paper applies those techniques for time series forecasting. I don't think it would be too difficult to evolve a NN to predict one or two days out for a particular stock. An easier approach would be to evolve a NN that would generate a buy/sell waveform like an oscillator. Since it would be more closely tailored to the characteristics of the stock you are looking at it may be more accurate than traditional indicators.