Introduction - Data Preprocessing in Data Mining - page 9

number of examples representing the class of interest is much lower than that of
the other classes. Its presence in many real-world applications has brought along a
growth of attention from researchers.
1.5.2 Multi-instance Learning [ 5 ]
This paradigm constitutes an extension based on imposed restrictions on models
in which each example consists of a bag of instances instead of an unique instance.
There are twomain ways of addressing this problem, either convertingmulti-instance
into single-instance by data transformations or by means of upgrade of single-case
algorithms.
1.5.3 Multi-label Classification [ 8 ]
It is generalization of traditional classification, in which each processed instance is
associated not with a class, but with a subset of them. In recent years different tech-
niques have appeared which, through the transformation of the data or the adaptation
of classic algorithms, aim to provide a solution to this problem.
1.5.4 Semi-supervised Learning [ 33 ]
This paradigm arises as an hybrid between the classification predictive task and the
clustering descriptive analysis. It is a learning paradigm concerned with the design
of models in the presence of both labeled and unlabeled data. Essentially, the de-
velopments in this field use unlabeled samples to either modify or re-prioritize the
hypothesis obtained from the labeled samples alone. Both semi-supervised classifi-
cation and semi-supervised clustering have emerged extending the traditional para-
digms by including unlabeled or labeled examples, respectively. Another paradigm
called Active Learning, with the same objective as Semi-supervised Learning, tries
to select the most important examples from a pool of unlabeled data, however these
examples are queried by an human expert.
1.5.5 Subgroup Discovery [ 17 ]
Also known as Contrast Set Mining and Emergent Pattern Mining, it is formed as the
result of another hybridization between supervised and unsupervised learning tasks,