Get Started

This is the 'Get Started' page for this project. Everything you will need to follow this project can be found here. We cover the requirements, the Jupyter Environment, the Python libraries we use and the datasets.

Requirements

These are the basic requirements which will be required.

1. A PC/Server which will contain a Tensorflow docker container and will do the processing. We are using an Intel NUC for this project, but as you can read here, there are no minimum requirements to use a CPU for tensorflow. We were using the following hardware specifications: Intel Core i7, 16GB RAM, SSD storage, but I am sure you could use lower specifications as well.
   It's also possible to install the Tensorflow docker container on your laptop, to take the neural network with you.
2. A PC/Laptop for the programming work. This can also be the PC/Server if you are planning on taking the same computer.
3. An Internet connection to download docker, the tensorflow container and your datasets.
4. (Optional) If you want to be able to work from around the globe and leave your server at home, you will need the ability to port forward. We won't cover that in this wiki, but you can google the terms 'docker container port forward' and read more about container networking here.

Environment

I have decided to use docker for this project, because it allows the Tensorflow virtual environment to be isolated from my other applications and only a part of it to be accessed. We have used the official tensorflow/tensorflow:latest-py3-jupyter docker container (more information about the different containers available can be found here).

The Tensorflow container uses Jupyter which can be used to develop open-source software and services for interactive computing across dozens of programming languages. The most usefull features are:

• Creating Python3 Notebooks, where snippets of code can be run immediately
• Terminal access to the docker container lets you install new libraries and run python files.
• File management with folders, upload, download, move and rename functions.

The installation instructions for docker and the tensorflow docker container can be found here.

Python Libraries

A lot of libraries came preinstalled with the installation of the Tensorflow (Jupyter) container. The most important libraries (with versions) are listed below and every new library installed will be added below.

• some of the preinstalled libraries: jupyter (1.0.0), Keras-Applications (1.0.8) and Keras-Preprocessing (1.1.0), tensorflow (2.0.0), virtualenv (16.7.8)
• custom libraries: pandas (0.25.3)

Extra information about Custom Libraries

• We are using the pandas (0.25.3) library to make the .csv datasets readable and usable for the neural network to read. More information about the pandas library can be found here.

Datasets

We are using two datasets for this project: total-traffic.csv and volatility.csv. These two datasets contain entries (every minute) of the network load in MBit/s and the market volatility in percent. These datasets are/will not be available online, so you should generate your own datasets to continue with this project.