rhiever / sklearn-benchmarks

A centralized repository to report scikit-learn model performance across a variety of parameter settings and data sets.
MIT License
210 stars 54 forks source link

Correlation stats, Symbol stats and Class probablities #24

Closed harshnisar closed 8 years ago

harshnisar commented 8 years ago

Three new families of stats:

Added skew and kurtosis as stats we are capturing. Any suggestions to capture characteristics of the distribution of values? Currently doing min, max, std, mean along with skew and kurt.

Next: Before I add new stats, I'll write the monkeyrunner script to test all metafeatures on our dataset collection to make sure nothing funny is happening.