Summary

In this chapter, we started by creating a baseline model to predict stock prices. To do this, we used an ARIMA model. Based on this model, we explored some important components of machine learning with time-series data, including using lagging variable values to predict a current variable value and the importance of stationarity. From there, we built a deep learning solution using Keras to assemble LSTM and then tuned this model further. In the process, we observed that this deep learning approach has some marked advantages compared to other traditional models, such as ARIMA. In the next chapter, we will use a generative adversarial network to create a synthetic face image.