Issues in evaluation of stream learning algorithms

By João Gama, Raquel Sebastião, Pedro Pereira Rodrigues. Knowledge Discovery and Data Mining, 2009

394 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/bf22b3489072b51b5b4c6059088be014a0371cbb

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Sentiment analysis on evolving social streams: how self-report imbalances can help (2014, background): “algorithms to constantly update and refine data description models, in face of the time-changing characteristics of the data [14, 6].”
Sentiment Knowledge Discovery in Twitter Streaming Data (2010, methodology): “In data stream mining, the most frequently used measure for evaluating predictive accuracy of a classifier is prequential accuracy [10].”
Scalable and efficient multi-label classification for evolving data streams (2012, methodology): “Gama et al. (2009) propose a forgetting mechanism for estimating holdout accuracy using prequential accuracy: either a sliding window of size w with the most recent observations can be used, or fading factors that weigh observations using a decay factor α.”
On evaluating stream learning algorithms (2012, background): “This paper is a substantial extension of a previous one (Gama et al. 2009), published in a top-level data mining conference, where these issues were firstly enunciated.”
Learning model trees from evolving data streams (2010, methodology): “Concerning the algorithm: We consider a new approach for the split selection criteria; We include a memory saving method for disabling bad split points; We improve the adaptation mechanism by using the Q statistic with a fading factor (Gama et al. 2009); We discuss and propose several memory managem”
Online Machine Learning in Big Data Streams (2018, background): “The effect of the beginning of the stream can be mitigated, for example, by forgetting mechanisms [97, 98].”
Speeding-Up Hoeffding-Based Regression Trees With Options (2011, methodology): “The weighted prequential error wPreqMSE( T i , t) is an improved variant of the prequential error originally proposed by Dawid (1984) and using a fading factor as proposed by Gama et al. (2009).”