Use Deep learning on tabular data by training Entity Embeddings of Categorical Variables. - Chandrasekhar's blog

#artificialintelligence 

Kaggle Elo merchant category recommendation being my first competition, my expectations weren't sky high and I'd be very happy if I managed to standout amongst the top 10%. I am trailing at 570 of 4000 odd data scientists in the competition. I have tried all the ML best practices and tricks known to me. I have done monstrous aggregates of aggregates, bevy of models (hehehe..) like LGBM, XG Boost, Random forests, Catboost and model post processing, parameter tuning, model blending, ensembling, feature permutation, elimination, recursive feature selection, Boruta and many more. I've written about this in detail here.