Time Series Prediction and Text Generation using RNN

Introduction

This project has two parts:

• Part 1: Perform time series prediction using a Recurrent Neural Network (RNN) regressor. In particular we will forecast the stock price of Apple 7 days in advance.
• Part 2: Create an English language sequence generator capable of building semi-coherent English sentences from scratch by building them up character-by-character. This will require a substantial amount of parameter tuning on a large training corpus (at least 100,000 characters long). In particular for this project we will be using a complete version of Sir Arthur Conan Doyle's classic book The Adventures of Sherlock Holmes.

Code

• RNN_project.ipynb - Code to perform time series prediction and create a sequence generator
• my_answers.py - Helper code to be used in the above notebook

Setup

• Python 3
• Install the packages in requirements.txt

Build your Own Deep Learning Workstation

If you have access to a GPU, you should follow the Keras instructions for running Keras on GPU.

Amazon Web Services

Instead of a local GPU, you could use Amazon Web Services to launch an EC2 GPU instance. (This costs money.)

Data

All the data for the two parts are in the subdirectory datasets.

Run

To run any script file, use:

python <script.py>

To open a notebook, use:

jupyter notebook <notebook.ipynb>