This is a further development of the kdnuggets article on fake news classification by George McIntyre:

https://www.kdnuggets.com/2017/04/machine-learning-fake-news-accuracy.html

In his article McIntyre approaches document classification from a very classical perspective: applying a vector-model to the corpus and then using a Naive Bayes classifier. Here we take it into the deep learning realm: we apply a deep convolutional network with a traininable word-embedding layer.

We compare the performances of both approaches. The notebook fake-news-classification.ipynb contains our computational results. The markdown document report.md reports our results in a more accessible manner. The results were also posted at opendatascience.