Kaggle Toolbox

Koolbox is a collection of helper functions and utilities designed to simplify training machine learning models in Kaggle competitions. This library abstracts away repetitive boilerplate code, allowing competitors to focus on more important tasks.

Installation

pip install koolbox

Usage

Trainer

import pandas as pd
from sklearn.model_selection import KFold
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import roc_auc_score

from koolbox import Trainer


X = pd.DataFrame(...)
y = pd.Series(...)

trainer = Trainer(
    estimator=RandomForestClassifier(random_state=42),
    cv=KFold(n_splits=5, shuffle=True, random_state=42),
    metric=roc_auc_score,
    task="binary",
    verbose=True
)

trainer.fit(X, y)

X_test = pd.DataFrame(...)
preds = trainer.predict(X_test)

oof_preds = trainer.oof_preds
overall_score = trainer.overall_score
fold_scores = trainer.fold_scores

SequentialFeatureSelector

from sklearn.linear_model import Ridge
from sklearn.model_selection import KFold
from sklearn.metrics import root_mean_squared_error
import pandas as pd

from koolbox import SequentialFeatureSelector


X = pd.DataFrame(...)
y = pd.Series(...)
X_test = pd.DataFrame(...)

sfs = SequentialFeatureSelector(
    Ridge(),
    cv=KFold(n_splits=5, random_state=42, shuffle=True),
    objective="minimize",
    direction="backward",
    metric=root_mean_squared_error
)

X = sfs.fit_transform(X, y)
X_test = sfs.transform(X_test)

selected_features = sfs.selected_features

WeightedEnsemble[Regressor, Classifier]

from sklearn.metrics import root_mean_squared_error
import pandas as pd

from koolbox import WeightedEnsembleRegressor


X = pd.DataFrame(...)
y = pd.Series(...)
X_test = pd.DataFrame(...)

model = WeightedEnsembleRegressor(
    objective="minimize",
    metric=root_mean_squared_error
)

model.fit(X, y)
preds = model.predict(X_test)