Hello everyone,

I made a project that tries to predict outcomes of Premier League matches.

In short, the predictions are modeled as time series classification in an unconventional way. Basically, a neural network model is created for every team and are all trained simultaneously according to matches as they were successively played.

I did not publish the dataset used, but I put full source code on my github. It is quite restrictive due to missing dataset, so please take it mostly as an inspiration or to take a look how I implemented various things.

Anyway, if you are interested in a bit crazy approach, check it out!

Link to blog post: Predicting soccer matches outcomes with machine learning as time series