skoot

Skoot is a lightweight python library of machine learning transformer classes that interact with scikit-learn and pandas. Its objective is to expedite data munging and pre-processing tasks that can tend to take up so much of data science practitioners' time. See the documentation for more info.

__Note that skoot is the preferred alternative to the now deprecated skutil library__

Two minutes to model-readiness

Real world data is nasty. Most data scientists spend the majority of their time tackling data cleansing tasks. With skoot, we can automate away so much of the bespoke hacking solutions that consume data scientists' time.

In this example, we'll examine a common dataset (the adult dataset from the UCI machine learning repo) that requires significant pre-processing.

from skoot.datasets import loadadultdf
from skoot.feature_selection import FeatureFilter
from skoot.decomposition import SelectivePCA
from skoot.preprocessing import DummyEncoder
from skoot.utils.dataframe import getnumericcolumns
from skoot.utils.dataframe import getcategoricalcolumns
from sklearn.modelselection import traintest_split
from sklearn.pipeline import Pipeline
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score
load the dataset with the skoot-native loader & split it
adult = loadadultdf(tgt_name="target")
y = adult.pop("target")
Xtrain, Xtest, ytrain, ytest = traintestsplit(
    adult, y, randomstate=42, testsize=0.2)
    
get numeric and categorical feature names
numcols = getnumericcolumns(Xtrain).columns
objcols = getcategoricalcolumns(Xtrain).columns
remove the education-num from the num_cols since we're going to remove it
numcols = numcols[~(num_cols == "education-num")]
    
build a pipeline
pipe = Pipeline([
    # drop out the ordinal level that's otherwise equal to "education"
    ("dropper", FeatureFilter(cols=["education-num"])),
    
    # decompose the numeric features with PCA
    ("pca", SelectivePCA(cols=num_cols)),
    
    # dummy encode the categorical features
    ("dummy", DummyEncoder(cols=objcols, handleunknown="ignore")),
    
    # and a simple classifier class
    ("clf", RandomForestClassifier(nestimators=100, randomstate=42))
])
pipe.fit(Xtrain, ytrain)
produce predictions
preds = pipe.predict(X_test)
print("Test accuracy: %.3f" % accuracyscore(ytest, preds))

For more tutorials, check out the documentation.