Perfectly Awesome

* Automatic Feature Creation using featuretools
* Categorical Features: Label/Binary/Hashing and Target/Mean Encoding
* log loss clipping
* Latitude and Longitude techniques
* AutoEncoders

1) using INDICATOR VARIABLES to isolate key information (thresholds, class groupings, special events, multiple features)
2) flagging INTERACTION FEATURES (sum, diff, product, quotients)
3) improved FEATURE REPRESENTATION (time/date, numbers->category maps, sparse class groups, dummy vars)
4) using EXTERNAL DATA (apis, scraping, time series, geocoding, ...
5) ERROR ANALYSIS after initial training

1) Continuous Numeric Data
2) Categorical Data
3) Text Data (Traditional Methods)
4) Text Data (Deep Learning Methods)

Artificial Dataset Builder (GitHub)
Missing Data Imputation (GitHub)
Preprocessing Toolkit (GitHub)
File IO - CSV, HDF5, H5py, Pytables, JSON, Serialization (GitHub)