Database management and analysis tools of machine induction

Fisher, Douglas H.; Hapanyengwi, Gilford

doi:10.1007/bf01066545

Cited by 11 publications

(10 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Since every internal node of T is a predictive attribute, the hidden variable C appears in every component BN. This fact implies that the component BN at each leaf of T does not represent only one cluster as Fisher and Hapanyengwi (1993) propose but a context-specific data clustering. That is, the data clustering encoded by each component BN is totally unrelated to the data clusterings encoded by the rest.…”

Section: Rbmns For Data Clusteringmentioning

confidence: 99%

“…A previous work with the same aim is where Fisher and Hapanyengwi (1993) propose to perform data clustering based upon a decision tree. The measure used to select the divisive attribute at each node during the decision tree construction consists of the computation of the sum of information gains over all attributes, while in the supervised paradigm the measure is limited to the information gain over a single specified class attribute.…”

Section: Rbmns For Data Clusteringmentioning

confidence: 99%

“…RBMNs extend BNs and BMNs, but RBMNs also extend partitional clustering systems (Fisher & Hapanyengwi, 1993). RBMNs can be considered as extensions of BNs because, like BMNs and mixtures of BNs, RBMNs allow us to encode context-specific conditional (in)dependencies.…”

Section: Rbmns For Data Clusteringmentioning

confidence: 99%

See 2 more Smart Citations

Untitled

2002

View full text Add to dashboard Cite

Abstract. This paper introduces and evaluates a new class of knowledge model, the recursive Bayesian multinet (RBMN), which encodes the joint probability distribution of a given database. RBMNs extend Bayesian networks (BNs) as well as partitional clustering systems. Briefly, a RBMN is a decision tree with component BNs at the leaves. A RBMN is learnt using a greedy, heuristic approach akin to that used by many supervised decision tree learners, but where BNs are learnt at leaves using constructive induction. A key idea is to treat expected data as real data. This allows us to complete the database and to take advantage of a closed form for the marginal likelihood of the expected complete data that factorizes into separate marginal likelihoods for each family (a node and its parents). Our approach is evaluated on synthetic and real-world databases.

show abstract

Section: Rbmns For Data Clusteringmentioning

confidence: 99%

Section: Rbmns For Data Clusteringmentioning

confidence: 99%

See 1 more Smart Citation

Untitled

2002

View full text Add to dashboard Cite

show abstract

“…One of them is chosen as the root of multi-level taxonomy. To make that choice we use the partition utility function (see Fisher (1996), Fisher and Hapanyengwi (1993) for details). Our taxonomy formation algorithm picks the one-level taxonomy with the greatest partition utility value as the one which guides the split at this level of the tree.…”

Section: Partition Utility and Creating Many-level Taxonomiesmentioning

confidence: 99%

“…In particular, 1. due to their simplicity and arguably clear interpretation, measures of association introduced by Goodman and Kruskal (1954) are used in lieu of χ 2 -like Cramer's V statistics applied in 49er (Troxel et al, 1994;Żytkow and Zembowicz, 1996). 2. instead of using the original heuristic approach intended to minimize the description length of the taxonomy, the choice of a one-level taxonomy to be attached under a given node is based on maximization of the so-called partition utility (as suggested, e.g., by Fisher and Hapanyengwi (1993) and based either on entropy or on the Gini index);…”

Section: Introductionmentioning

confidence: 99%

Taxonomy Formation by Approximate Equivalence Relations, Revisited

El-Mouadib

Koronacki

Żytkow

1999

Principles of Data Mining and Knowledge Discovery

View full text Add to dashboard Cite

Abstract. Unsupervised classification of objects involves formation of classes and construction of one or more taxonomies that include those classes. Meaningful classes can be formed in feedback with acquisition of knowledge about each class. We demonstrate how contingency tables can be used to construct one-level taxonomy elements by relying only on approximate equivalence relations between attribute pairs, and how a multi-level taxonomy formation can be guided by a partition utility functions. Databases with different types of attributes and large number of records can be dealt with.

show abstract

A Survey of Evolutionary Algorithms for Data Mining and Knowledge Discovery

Freitas

2003

Natural Computing Series

260

144

View full text Add to dashboard Cite

Abstract:This chapter discusses the use of evolutionary algorithms, particularly genetic algorithms and genetic programming, in data mining and knowledge discovery. We focus on the data mining task of classification. In addition, we discuss some preprocessing and postprocessing steps of the knowledge discovery process, focusing on attribute selection and pruning of an ensemble of classifiers. We show how the requirements of data mining and knowledge discovery influence the design of evolutionary algorithms. In particular, we discuss how individual representation, genetic operators and fitness functions have to be adapted for extracting high-level knowledge from data.

show abstract

Database management and analysis tools of machine induction

Cited by 11 publications

References 9 publications

Untitled

Untitled

Taxonomy Formation by Approximate Equivalence Relations, Revisited

A Survey of Evolutionary Algorithms for Data Mining and Knowledge Discovery

Contact Info

Product

Resources

About