In this chapter, we started with a slight disdain for the impossible, trying to beat the market by using TensorFlow and Keras RNN APIs to predict stock prices. We first discussed what RNN and LSTM models are and how to use them to make stock price predictions. Then we built two RNN models from scratch with TensorFlow and Keras, reaching close to 60% of testing correctness. Finally, we covered how to freeze the models and use them on iOS and Android, fixing a possible runtime error on iOS with a custom TensorFlow library.
If you're a little bit disappointed that we haven't built a model with an 80% or 90% correct prediction ratio, you may want to continue the "try and iterate" process to see whether predicting stock prices with that correct ratio is possible at all. But the skills you have learned from RNN model building, training, and testing using...