Speedml Machine Learning Speed Start

Author of Speedml explains data science solutions for beginners. Includes chapter on Top 20 most voted Kaggle solution.
Get Data Science Solutions Book

Speedml 0.9.0 Consistent Api And Support

The Speedml release 0.8.1 and 0.9.0 are focusing on documentation and consistent API related changes.

The API documentation is available on PyPi website.

Changes to existing API

feature.regex_extract We are renaming to a simpler alternative feature.extract making use of the prefix to signify the method. Of course the parameter is still called regex so we know how we are extracting the feature.

feature.ordinal_to_numeric This method was incorrectly named in release 0.8.0 as it operates on categorical features regardless of these being ordinal. We now call the methond mapping signifying it can take more general use cases. We still recommend using this method sparingly when you have 2-5 categories. You should consider the feature.labels method as a first preference alternative in most cases.

outliers_fix is renaming to outliers as the feature. prefix indicates feature engineering aspect for the method.

crosstab is now part of plot.crosstab API instead of core Speedml. We are treating plot component generally for EDA including dataframes as outputs as well as charts.

sample_accuracy is now appropriately part of Xgb component as it returns accuracy specific to XGBoost model.

Created Model component. Renamed and moved model.data, model.evaluate, model.ranks to the new component.

Moved feature_selection to Xgb component as it returns feature thresholds and accuracy for XGBoost model.

return values are now more consitent and return printable strings or dataframes which can output directly to a Notebook. The print method is applicable when returning strings during a loop to determine progress, like in case of the feature_selection method.

Supporting the API

We also went live with the Speedml.com website detailing API documentation and use cases.

We are updating the Speedml notebooks to include the latest API. We are also adding shorter single feature notebooks as demos for the Speedml linear workflow.