kappaml-core

Core library for the KappaML project.

This library implements experimental online automated machine learning algorithms for the KappaML project.

Examples

MetaClassifier

from river.tree import HoeffdingTreeClassifier
from river.linear_model import LogisticRegression
from river.metrics import Accuracy
from kappaml_core.meta import MetaClassifier

# Create base models
models = [
    HoeffdingTreeClassifier(weighted=True),
    HoeffdingTreeClassifier(weighted=False),
    LogisticRegression()
]

# Initialize meta-classifier
model = MetaClassifier(
    models=models,
    meta_learner=HoeffdingTreeClassifier(),
    metric=Accuracy(),
    mfe_groups=["general"],
    window_size=200,
    meta_update_frequency=50
)

for x, y in stream:
    # Make prediction
    y_pred = model.predict_one(x)

    # Update the model
    model.learn_one(x, y)

MetaRegressor

from river.linear_model import LinearRegression
from river.tree import HoeffdingTreeRegressor
from river.preprocessing import StandardScaler
from river.metrics import MAPE
from kappaml_core.meta import MetaRegressor

# Create base models
models = [
    LinearRegression(),
    StandardScaler() | LinearRegression(),
    [LinearRegression(l2=l2) for l2 in range(0, 1, 0.1)],
    HoeffdingTreeRegressor()
]

# Initialize meta-regressor
model = MetaRegressor(
    models=models,
    meta_learner=HoeffdingTreeClassifier(),
    metric=MAPE(),
    mfe_groups=["general"],
    window_size=200,
    meta_update_frequency=50
)

for x, y in stream:
    # Make prediction
    y_pred = model.predict_one(x)
    # Update the model
    model.learn_one(x, y)

Note

This project has been set up using PyScaffold 4.1.4. For details and usage information on PyScaffold see https://pyscaffold.org/.