MINAS: multiclass learning algorithm for novelty detection in data streams

Abstract

Data stream mining is an emergent research area that aims at extracting knowledge from large amounts of continuously generated data. Novelty detection (ND) is a classification task that assesses if one or a set of examples differ significantly from the previously seen examples. This is an important task for data stream, as new concepts may appear, disappear or evolve over time. Most of the works found in the ND literature presents it as a binary classification task. In several data stream real life problems, ND must be treated as a multiclass task, in which, the known concept is composed by one or more classes and different new classes may appear. This work proposes MINAS, an algorithm for ND in data streams. MINAS deals with ND as a multiclass task. In the initial training phase, MINAS builds a decision model based on a labeled data set. In the online phase, new examples are classified using this model, or marked as unknown. Groups of unknown examples can be used later to create valid novelty patterns (NP), which are added to the current model. The decision model is updated as new data come over the stream in order to reflect changes in the known Responsible editor: Charu Aggarwal.

Novelty detection in data streams

“…Recently, several authors extended this framework to a multiclass context (Faria et al 2013a;de Faria et al 2015b;Farid and Rahman 2012;Masud et al 2010aMasud et al , 2011aAl-Khateeb et al 2012a, b). In this case, the previous formalization must be generalized to the multiclass context.…”

“…Thus, the induced decision model is able to distinguish between three or more classes. This strategy is adopted by algorithms like Faria et al (2013a), de Faria et al (2015b), Al-Khateeb et al (2012a), Masud et al (2011a), Masud et al (2010a), Farid and Rahman (2012) and Farid et al (2013).…”

“…The second aspect is the number of classifiers. The techniques proposed in Faria et al (2013a), de Faria et al (2015b), Spinosa et al (2009), Hayat and Hashemi (2010), Albertini and de Mello (2007), Rusiecki (2012), Farid and Rahman (2012) and Krawczyk and Michal (2013) create only one classifier in the offline phase, which will be update over the stream incrementally. Techniques, found in Masud et al (2010a), Masud et al (2011a), Tan et al In all these studies, a chunk of labeled examples is used for the induction of each classifier.…”

“…In the last years, several researchers have proposed different approaches for ND in DSs (Masud et al 2011a;Spinosa et al 2009;Hayat and Hashemi 2010;Faria et al 2013a;de Faria et al 2015b;Farid and Rahman 2012;Al-Khateeb et al 2012a, b). A review of the main aspects of these studies may provide an useful inspiration for the development of new techniques and highlight the main issues to be addressed in future researches.…”

Exploração do uso de short-term memory na construção de métodos de acesso métricos dinâmicos sobre a perspectiva de diferentes políticas de divisão de nós

“…Estas bases foram utilizados nos trabalhos (FARIA et al, 2016b;CARVALHO;GAMA, 2016;MASUD et al, 2011), para avaliar métodos de classificação e agrupamento de fluxos contínuos de dados simulando distribuição não estacionária de dados.…”