On building LSTM and CNN architectures for modeling time series data

Abstract

Stock price prediction is one very challenging and desirable real-world task. The challenge comes from the very dynamic nature of stock movement that is triggered by many different known and unknown factors. An accurate prediction is naturally connected to money gain. In this tutorial, two deep learning architectures will be employed to model such time series data, namely the long short-term memory networks and the temporal convolutional neural networks. The implementation will be performed in Python, using Keras API with Tensorflow backend.