hoeztutar / conceptdrift Goto Github PK

ADWIN is a change detector and estimator that solves in a well-specified way the problem of tracking the average of a stream of bits or real-valued numbers.

ADWIN keeps a variable-length window of recently seen items, with the property that the window has the maximal length statistically consistent with the hypothesis “there has been no change in the average value inside the window”. More precisely, an older fragment of the window is dropped if and only if there is enough evidence that its average value differs from that of the rest of the window.

Based on: "Learning from Time-Changing Data with Adaptive Windowing" by Albert Bifet and Ricard Gavaldà

class concept_drift.adwin.Adwin
	(delta=0.002, max_buckets=5, min_clock=32, min_length_window=10, min_length_sub_window=5)

Methods

set_input(value)

Set input value to the drift detector - ADWIN.

Example

from concept_drift.adwin import Adwin

adwin = Adwin()
for i in range(1000):
    if adwin.set_input(i):
	print("Here is a drift")

The Page-Hinkley test which is a sequential adaptation of the detection of an abrupt change of the average of a Gaussian signal. It allows efficient detection of changes in the normal behaviour of a process which is established by a model.

Based on: "Knowledge Discovery from Data Streams" by João Gama (p.76)

 class concept_drift.page_hinkley.PageHinkley
 	(delta_=0.005, lambda_=50, alpha_=1 - 0.0001)

Methods

set_input(value)

Set input value to the drift detector - PageHinkley.

Used the elecNormNew dataset;
Used GaussianNB as based classification;
Used accuracy_score as input date for change detector;
Used the prequential evaluation;

GaussianNB :
Mean acc within the window 1000: 0.7289912189511405

Page-Hinkley :
Drift detection: 8
Mean acc within the window 1000: 0.7342849306584686

ADWIN :
Drift detection: 139
Mean acc within the window 1000: 0.7496421003738032