HN Explorer

1 comment

[ 5.1 ms ] story [ 12.8 ms ] thread
LrnByTeach
It seems the ability of this tool to extract features from regular enterprise databases will save lots of manual effort

Github repo url https://github.com/Featuretools/featuretools

Python notebook for NYC Taxi 1.5 million trips training dataset https://github.com/Featuretools/NYC-Taxi-Demo/blob/master/NY...

How this automated feature extraction compares:

    Received a score of 0.45288 on the Kaggle competition.
    Placed 685 out of 1257.
    Beat 45% of competitors on the Kaggle competition.
    Scored 4% better than the baseline solution
    Had a modeling RMSLE of 0.40196