Types of Cost in Inductive Concept Learning

By Peter D. Turney. arXiv.org, 2002

426 citations recorded by the source corpus.

Inductive concept learning is the task of learning to assign cases to a discrete set of classes. In real-world applications of concept learning, there are many different types of cost involved. The majority of the machine learning literature ignores all types of cost (unless accuracy is interpreted as a type of cost measure). A few papers have investigated the cost of misclassification errors. Very few papers have examined the many other types of cost. In this paper, we attempt to create a taxonomy of the different types of cost that are involved in inductive concept learning. This taxonomy may help to organize the literature on cost-sensitive learning. We hope that it will inspire researchers to investigate all types of cost in inductive concept learning in more depth.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/95da1bff50185494bd747bead09958f2e34bf50a

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

Get another label? improving data quality and data mining using multiple, noisy labelers (2008, background): “Turney [31] provides a short but comprehensive survey of the diﬀerent sorts of costs that should be considered, including data acquisition costs and labeling costs.”
Cost-Sensitive Learning and the Class Imbalance Problem (2008, background): “( Turney, 2000 ) provides a comprehensive survey of a large variety of different types of costs in data mining and machine learning, including misclassification costs, data acquisition cost (instance costs and attribute costs), active learning costs, computation cost, human-computer interaction cost”
Cost-conscious comparison of supervised learning algorithms over multiple data sets (2012, background): “Various types of cost can be used [20], for example, the space and/or time complexity during training and/or testing, interpretability, ease of programming, etc.”
Decision trees with minimal costs (2004, background): “As pointed out by (Turney 2000) it is obviously an oversight.”
Cost-Sensitive Decision Trees Applied to Medical Data (2007, background): “Misclassification and test costs are on the most important costs, but there are also other types of costs [ 5 ].”
Finding Cost-Efficient Decision Trees (2014, background): “Although this thesis only concentrates on the cost of a test, there are many other ways to include costs when building decision trees[43].”
Cluster-based under-sampling approaches for imbalanced data distributions (2009, positive): “At the algorithmic level, developed methods include cost-sensitive learning (Drummond & Holte, 2003; Elkan, 2001; Turney, 2000) and rec-ognition-based learning (Chawla, Bowyer, Hall, & Kegelmeyer, 2002; Manevitz & Yousef, 2001).”
Data Mining for Imbalanced Datasets: An Overview (2005, background): “Cost-sensitive measures usually assume that the costs of making an error are known (Turney, 2000; Domingos, 1999; Elkan, 2001).”
A Fast Way to Produce Optimal Fixed-Depth Decision Trees (2008, background): “Turney (2000) discusses the general challenge of learning the best classi(cid:2)er subject to some explicit cost constraint.”
A Fast Way to Produce Optimal Fixed-Depth Decision Trees. (2008, background): “low different tests to have different costs (Turney 2000; Greiner, Grove, & Roth 2002).”
Information, Divergence and Risk for Binary Experiments (2009, background): “Examples of existing pluralistic attempts include limited problem catalogs such as for different notions of cost (Turney, 2000) or a restricted set of problems (Raudys, 2001).”
Cost-sensitive boosting for classification of imbalanced data (2007, background): “The cost value of a sample may depend on the nature of the particular case [59].”
Unsupervised named-entity extraction from the Web: An experimental study (2005, background): “The discriminator ordering problem is identical to the problem of active classification, in which an object to be classified has a set of unknown attributes, each of which can be obtained by performing tests of varying costs [26, 45].”
Financial credit risk assessment: a recent review (2015, background): “With respect to the category of Turney (2000), there are different categories of misclassification cost: constant error cost of class, conditional error cost dependent on individual instance, Real class Predicted class time cost, feature cost and test cost.”
Quantifying counts and costs via classification (2008, methodology): “ Turney (2000) developed an extensive taxonomy of types of cost in machine learning, but none of the costs are related to the cost attribute in cost quantification.”
Editorial: special issue on learning from imbalanced data sets (2004, background): “Cost-sensitive learning or measures assume that a cost-matrix is known for different types of errors or even examples, which can be used at classification time [53][15].”