Concept Drift

In predictive analytics and machine learning, concept drift means that the statistical properties of the target variable, which the model is trying to predict, change over time in unforeseen ways. This causes problems because the predictions become less accurate as time passes.

The term ''concept'' refers to the quantity to be predicted. More generally, it can also refer to other phenomena of interest besides the target concept, such as an input, but, in the context of concept drift, the term commonly refers to the target variable.

Examples

In a fraud detection application the target concept may be a binary attribute fraudulent with values "yes" or "no" that indicates whether a given transaction is fraudulent. Or, in a weather prediction application, there may be several target concepts such as temperature, pressure, and humidity.

The behavior of the customers in an online shop may change over time. For example, if weekly merchandise sales are to be predicted, and a predictive model has been developed that works satisfactorily. The model may use inputs such as the amount of money spent on advertising, promotions being run, and other metrics that may affect sales. The model is likely to become less and less accurate over time – this is concept drift. In the merchandise sales application, one reason for concept drift may be seasonality, which means that shopping behavior changes seasonally. Perhaps there will be higher sales in the winter holiday season than during the summer, for example. Concept drift generally occurs when the covariates that comprise the data set begin to explain the variation of your target set less accurately — there may be some confounding variables that have emerged, and that one simply cannot account for, which renders the model accuracy to progressively decrease with time. Generally, it is advised to perform health checks as part of the post-production analysis and to re-train the model with new assumptions upon signs of concept drift.

Possible remedies

To prevent deterioration in prediction accuracy because of concept drift, ''reactive'' and ''tracking'' solutions can be adopted. Reactive solutions retrain the model in reaction to a triggering mechanism, such as a change-detection test, to explicitly detect concept drift as a change in the statistics of the data-generating process. When concept drift is detected, the current model is no longer up-to-date and must be replaced by a new one to restore prediction accuracy. A shortcoming of reactive approaches is that performance may decay until the change is detected. Tracking solutions seek to track the changes in the concept by continually updating the model. Methods for achieving this include online machine learning, frequent retraining on the most recently observed samples, and maintaining an ensemble of classifiers where one new classifier is trained on the most recent batch of examples and replaces the oldest classifier in the ensemble.

Contextual information, when available, can be used to better explain the causes of the concept drift: for instance, in the sales prediction application, concept drift might be compensated by adding information about the season to the model. By providing information about the time of the year, the rate of deterioration of your model is likely to decrease, concept drift is unlikely to be eliminated altogether. This is because actual shopping behavior does not follow any static, finite model. New factors may arise at any time that influence shopping behavior, the influence of the known factors or their interactions may change.

Concept drift cannot be avoided for complex phenomena that are not governed by fixed laws of nature. All processes that arise from human activity, such as socioeconomic processes, and biological processes are likely to experience concept drift. Therefore, periodic retraining, also known as refreshing, of any model is necessary.

Software

NannyML
An open-source Python library for detecting univariate and multivariate distribution drift and estimating machine learning model performance without ground truth labels.
* RapidMiner: Formerly ''Yet Another Learning Environment'' (YALE): free open-source software for knowledge discovery, data mining, and machine learning also featuring data stream mining, learning time-varying concepts, and tracking drifting concept. It is used in combination with its data stream mining plugin (formerly concept drift plugin).
* EDDM
Early Drift Detection Method
: free open-source implementation of drift detection methods in Weka.
* MOA (Massive Online Analysis): free open-source software specific for mining data streams with concept drift. It contains a prequential evaluation method, the EDDM concept drift methods, a reader of ARFF real datasets, and artificial stream generators as SEA concepts, STAGGER, rotating hyperplane, random tree, and random radius based functions. MOA supports bi-directional interaction with Weka.

Datasets

Real

* USP Data Stream Repository, 27 real-world stream datasets with concept drift compiled by Souza et al. (2020)
Access
* Airline, approximately 116 million flight arrival and departure records (cleaned and sorted) compiled by E. Ikonomovska. Reference: Data Expo 2009 Competitio
Access
* Chess.com (online games) and Luxembourg (social survey) datasets compiled by I. Zliobaite
Access
* ECUE spam 2 datasets each consisting of more than 10,000 emails collected over a period of approximately 2 years by an individual

from S.J.Delany webpage
* Elec2, electricity demand, 2 classes, 45,312 instances. Reference: M. Harries, Splice-2 comparative evaluation: Electricity pricing, Technical report, The University of South Wales, 1999

from J.Gama webpage
Comment on applicability

* PAKDD'09 competition data represents the credit evaluation task. It is collected over a five-year period. Unfortunately, the true labels are released only for the first part of the data
Access
* Sensor stream and Power supply stream datasets are available from X. Zhu's Stream Data Mining Repository.

* SMEAR is a benchmark data stream with a lot of missing values. Environment observation data over 7 years. Predict cloudiness
Access
* Text mining, a collection of text mining datasets with concept drift, maintained by I. Katakis
Access
* Gas Sensor Array Drift Dataset, a collection of 13,910 measurements from 16 chemical sensors utilized for drift compensation in a discrimination task of 6 gases at various levels of concentrations
Access

Other

* KDD'99 competition data contains ''simulated'' intrusions in a military network environment. It is often used as a benchmark to evaluate handling concept drift

Synthetic

* Extreme verification latency benchmark
Access
from Nonstationary Environments – Archive.
* Sine, Line, Plane, Circle and Boolean Data Sets
Access
from L.Minku webpage.

* SEA concepts

from J.Gama webpage.
* STAGGER
* Mixed

Data generation frameworks

*
Download
from L.Minku webpage.
*

Projects

INFER
Computational Intelligence Platform for Evolving and Robust Predictive Systems (2010–2014), Bournemouth University (UK), Evonik Industries (Germany), Research and Engineering Centre (Poland)

HaCDAIS
Handling Concept Drift in Adaptive Information Systems (2008–2012), Eindhoven University of Technology (the Netherlands)

KDUS
Knowledge Discovery from Ubiquitous Streams, INESC Porto and Laboratory of Artificial Intelligence and Decision Support (Portugal)

ADEPT
Adaptive Dynamic Ensemble Prediction Techniques, University of Manchester (UK), University of Bristol (UK)

ALADDIN
autonomous learning agents for decentralised data and information networks (2005–2010)

GAENARI
C++ incremental decision tree algorithm. it minimize concept drifting damage. (2022)

Benchmarks

NAB
The Numenta Anomaly Benchmark, benchmark for evaluating algorithms for anomaly detection in streaming, real-time applications. (2014–2018)

Meetings

*2014
*

Special Session on "Concept Drift, Domain Adaptation & Learning in Dynamic Environments" @IEEE IJCNN 2014
*2013
*
RealStream
Real-World Challenges for Data Stream Mining Workshop-Discussion at the ECML PKDD 2013, Prague, Czech Republic.
*
LEAPS 2013
The 1st International Workshop on Learning stratEgies and dAta Processing in nonStationary environments
*2011
*

Special Session on Learning in evolving environments and its application on real-world problems at ICMLA'11
*
HaCDAIS 2011
The 2nd International Workshop on Handling Concept Drift in Adaptive Information Systems
*
ICAIS 2011
Track on Incremental Learning
*
IJCNN 2011
Special Session on Concept Drift and Learning Dynamic Environments
*

Symposium on Computational Intelligence in Dynamic and Uncertain Environments
*2010
*
HaCDAIS 2010
International Workshop on Handling Concept Drift in Adaptive Information Systems: Importance, Challenges and Solutions
*

Special Session on Dynamic learning in non-stationary environments
*
SAC 2010
Data Streams Track at ACM Symposium on Applied Computing
*
SensorKDD 2010
International Workshop on Knowledge Discovery from Sensor Data
*
StreamKDD 2010
Novel Data Stream Pattern Mining Techniques
** Concept Drift and Learning in Nonstationary Environments a
IEEE World Congress on Computational Intelligence
*
MLMDS’2010
Special Session on Machine Learning Methods for Data Streams at the 10th International Conference on Intelligent Design and Applications, ISDA’10

Bibliographic references

Many papers have been published describing algorithms for concept drift detection. Only reviews, surveys and overviews are here:

Reviews

*
*
*
*
*
*
*
* Zliobaite, I., Learning under Concept Drift: an Overview. Technical Report. 2009, Faculty of Mathematics and Informatics, Vilnius University: Vilnius, Lithuania
PDF
*
*
*
*
*

See also

* Data stream mining
* Data mining
* Machine learning

References

{{reflist

Data mining
Machine learning