Level 3 Fake News Identification Project

Big Data Summative Assignment

Author: JRhodes95 - Z0954757

Environment Set-Up

Full Dataset not included in repo due to large size.

This project was created in an environment with the following packages:

• Python 2.7,
• Numpy,
• Pandas,
• Matplotlib
• NLTK - and data installed by entering the python interpreter typing the following and installing the data associated with the NLTK book: >> import nltk >> nltk.download()
• spaCy
• Scikit learn
• Keras

Shallow Learning approach

The shallow learning approach is encapsulated in shallow.py.

Deep Learning approach

C) Feature Extraction: The required word2vec function is available as word2vec.py and implements spaCy's prebuilt vectorizer.

D) i) The LSTM is encapsulated in deep.py D) ii) Due to hardware and time limitations, there is no implementation for D) ii).

Results presentation

The module main.py is included. Running $ python main.py will start a command line interface to select what code you wish to run. The layout of the function follows the order of the report.

It first asks if you would like to run the shallow approaches. Typing y will execute this step, or any other input not including y will skip it.

Next it asks if you would like to run the further investigations into optimising the the shallow learning models. If y is entered, you are then prompted if you would like to regenerate the data or load it from a saved csv file of the outcomes.

In both cases, it will then genrate the contour plots for the grid search optimisation results.

Report

The report is included as report.pdf.