API Documentation

This is the API documentation for scikit-multiflow.

Core: skmultiflow.core

The skmultiflow.core module covers core elements of scikit-multiflow.

core.base.BaseEstimator

Base Estimator class for compatibility with scikit-learn.

core.BaseSKMObject

Base class for most objects in scikit-multiflow

core.ClassifierMixin

Mixin class for all classifiers in scikit-multiflow.

core.RegressorMixin

Mixin class for all regression estimators in scikit-multiflow.

core.MetaEstimatorMixin

Mixin class for all meta estimators in scikit-multiflow.

core.MultiOutputMixin

Mixin to mark estimators that support multioutput.

core.Pipeline

[Experimental] Holds a set of sequential operation (transforms), followed by a single estimator.

Data: skmultiflow.data

The skmultiflow.data module contains data stream methods including methods for batch-to-stream conversion and generators.

data.base_stream.Stream

Base Stream class.

data.DataStream

Creates a stream from a data source.

data.FileStream

Creates a stream from a file source.

Stream Generators

data.AGRAWALGenerator

Agrawal stream generator.

data.HyperplaneGenerator

Hyperplane stream generator.

data.LEDGenerator

LED stream generator.

data.LEDGeneratorDrift

LED stream generator with concept drift.

data.MIXEDGenerator

Mixed data stream generator.

data.RandomRBFGenerator

Random Radial Basis Function stream generator.

data.RandomRBFGeneratorDrift

Random Radial Basis Function stream generator with concept drift.

data.RandomTreeGenerator

Random Tree stream generator.

data.SEAGenerator

SEA stream generator.

data.SineGenerator

Sine stream generator.

data.STAGGERGenerator

STAGGER concepts stream generator.

data.WaveformGenerator

Waveform stream generator.

data.MultilabelGenerator

Creates a multi-label stream.

data.RegressionGenerator

Creates a regression stream.

data.ConceptDriftStream

Generates a stream with concept drift.

Learning methods

Anomaly detection methods

The skmultiflow.anomaly_detection module includes anomaly detection methods.

anomaly_detection.HalfSpaceTrees

Half–Space Trees.

Bayes methods

The skmultiflow.bayes module includes Bayes learning methods.

bayes.NaiveBayes

Naive Bayes classifier.

Lazy learning methods

The skmultiflow.lazy module includes lazy learning methods in which generalization of the training data is delayed until a query is received, this is, on-demand.

lazy.KNN

K-Nearest Neighbors classifier.

lazy.KNNAdwin

K-Nearest Neighbors classifier with ADWIN change detector.

lazy.SAMKNN

Self Adjusting Memory coupled with the kNN classifier.

Ensemble methods

The skmultiflow.meta module includes meta learning methods.

meta.AccuracyWeightedEnsemble

Accuracy Weighted Ensemble classifier

meta.AdaptiveRandomForest

Adaptive Random Forest classifier.

meta.AdditiveExpertEnsemble

Additive Expert ensemble classifier.

meta.BatchIncremental

Batch Incremental ensemble classifier.

meta.ClassifierChain

Classifier Chains for multi-label learning.

meta.ProbabilisticClassifierChain

Probabilistic Classifier Chains for multi-label learning.

meta.MonteCarloClassifierChain

Monte Carlo Sampling Classifier Chains for multi-label learning.

meta.DynamicWeightedMajority

Dynamic Weighted Majority ensemble classifier.

meta.LearnNSE

Learn++.NSE ensemble classifier.

meta.LearnPP

Learn++ ensemble classifier.

meta.LeverageBagging

Leverage Bagging ensemble classifier.

meta.MultiOutputLearner

Multi-Output Learner for multi-label classification.

meta.OnlineAdaC2

Online AdaC2 ensemble classifier.

meta.OnlineBoosting

Online Boosting ensemble classifier.

meta.OnlineCSB2

Online CSB2 ensemble classifier.

meta.OnlineRUSBoost

Online RUSBoost ensemble classifier.

meta.OnlineSMOTEBagging

Online SMOTEBagging ensemble classifier.

meta.OnlineUnderOverBagging

Online Under-Over-Bagging ensemble classifier.

meta.OzaBagging

Oza Bagging ensemble classifier.

meta.OzaBaggingAdwin

Oza Bagging ensemble classifier with ADWIN change detector.

meta.RegressorChain

Regressor Chains for multi-output learning.

Neural Networks

The skmultiflow.neural_networks module includes learning methods based on Neural Networks.

neural_networks.PerceptronMask

Mask for sklearn.linear_model.Perceptron.

Prototype based methods

The skmultiflow.prototype module includes prototype-based learning methods.

prototype.RobustSoftLearningVectorQuantization

Robust Soft Learning Vector Quantization for Streaming and Non-Streaming Data.

Rules based methods

The skmultiflow.rules module includes rule-based learning methods.

rules.VFDR

Adaptive Very Fast Decision Rules.

Trees based methods

The skmultiflow.trees module includes learning methods based on trees.

trees.HoeffdingTree

Hoeffding Tree or Very Fast Decision Tree.

trees.HAT

Hoeffding Adaptive Tree.

trees.HATT

Hoeffding Anytime Tree or Extremely Fast Decision Tree.

trees.LCHT

Label Combination Hoeffding Tree for multi-label learning.

trees.RegressionHoeffdingTree

Regression Hoeffding Tree or Fast Incremental Model Tree with Drift Detection.

trees.RegressionHAT

An adaptation of the Hoeffding Adaptive Tree for regression.

trees.MultiTargetRegressionHoeffdingTree

Multi-target Regression Hoeffding Tree.

trees.StackedSingleTargetHoeffdingTreeRegressor

Stacked Single-target Hoeffding Tree Regressor.

Drift Detection: skmultiflow.drift_detection

The skmultiflow.drift_detection module includes methods for Concept Drift Detection.

drift_detection.ADWIN

Adaptive Windowing method for concept drift detection.

drift_detection.DDM

Drift Detection Method.

drift_detection.EDDM

Early Drift Detection Method.

drift_detection.PageHinkley

Page-Hinkley method for concept drift detection.

Evaluation: skmultiflow.evaluation

The skmultiflow.evaluation module includes evaluation methods for stream learning.

evaluation.EvaluateHoldout

The holdout evaluation method or periodic holdout evaluation method.

evaluation.EvaluatePrequential

The prequential evaluation method or interleaved test-then-train method.

Transform: skmultiflow.transform

The skmultiflow.transform module covers methods that perform data transformations.

transform.MissingValuesCleaner

Fills missing values with some defined value.

transform.OneHotToCategorical

Transforms one-hot encoded data into categorical feature(s).

Misc:

Data structure

lazy.KDTree

A K-dimensional tree implementation, adapted for k dimensional problems.

Utilities

core.clone

Constructs a new estimator with the same parameters.