A branch and bound incremental conceptual clusterer

Nevins, Arthur J.

doi:10.1007/bf00993819

Cited by 4 publications

(6 citation statements)

References 18 publications

(24 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In contrast, hierarchical redistribution resorts each cluster, regardless of its initial location in the tree, through the root of the entire tree, thus more vigorously pursuing migration and more globally evaluating the merits of such moves. 1 The idea of hierarchical redistribution is also closely related to strategies found in the Bridger (Reich & Fenves, 1991) and Hierarch (Nevins, 1995) systems. In particular, Bridger identi es `misplaced' clusters in a hierarchical clustering using a criterion speci ed, in part, by a domain expert, whereas hierarchical redistribution simply uses the objective function.…”

Section: Iterative Hierarchical Redistributionmentioning

confidence: 99%

“…Ideally, clustering quality as measured by the objective function should be well correlated with clustering utility as determined by a performance task: the higher the quality of a clustering as judged by the objective function, the greater the performance improvement (e.g., reduction of error rate), and the lower the quality, the less that performance improves. However, several authors (Fisher et al, 1992;Nevins, 1995;Devaney & Ram, 1993) have pointed out that PU scores do not seem well-correlated with error rates. More precisely, hierarchical clusterings (constructed by hierarchical sorting) in which the top-level partition has a low PU score lead to roughly the same error rates as hierarchies in which the top-level partition has a high PU score, when variable-value predictions are made at leaves (singleton clusters).…”

Section: A Closer Look At External Validation Criteriamentioning

confidence: 99%

See 1 more Smart Citation

Iterative Optimization and Simplification of Hierarchical Clusterings

Fisher¹

1996

jair

140

View full text Add to dashboard Cite

Clustering is often used for discovering structure in data. Clustering systems di er in the objective function used to evaluate clustering quality and the control strategy used to search the space of clusterings. Ideally, the search strategy should consistently construct clusterings of high quality, but be computationally inexpensive as well. In general, we cannot have it both ways, but we can partition the search so that a system inexpensively constructs a`tentative' clustering for initial examination, followed by iterative optimization, which continues to search in background for improved clusterings. Given this motivation, we evaluate an inexpensive strategy for creating initial clusterings, coupled with several control strategies for iterative optimization, each of which repeatedly modi es an initial clustering in search of a better one. One of these methods appears novel as an iterative optimization strategy in clustering contexts. Once a clustering has been constructed it is judged by analysts { often according to task-speci c criteria. Several authors have abstracted these criteria and posited a generic performance task akin to pattern completion, where the error rate over completed patterns is used to`externally' judge clustering utility. Given this performance task, we adapt resampling-based pruning strategies used by supervised learning systems to the task of simplifying hierarchical clusterings, thus promising to ease post-clustering analysis. Finally, we propose a number of objective functions, based on attribute-selection measures for decision-tree induction, that might perform well on the error rate and simplicity dimensions.

show abstract

Section: Iterative Hierarchical Redistributionmentioning

confidence: 99%

Section: A Closer Look At External Validation Criteriamentioning

confidence: 99%

Iterative Optimization and Simplification of Hierarchical Clusterings

Fisher¹

1996

jair

140

View full text Add to dashboard Cite

show abstract

“…Similar cluster shapes are also formed by the INC system [2]. HIERARCH"s constraints exhibit bias toward certain cluster shapes [7].…”

Section: Discussionmentioning

confidence: 99%

“…HOMOGEN"s approach that uses a set of conceptual constraints (e.g., the homogeneity and monotonicity properties) as the guiding principles during the hierarchy restructuring can be related to the ARACHNE [6] and the HIERARCH [7] systems. Unlike these systems that rely exclusively on their constraints as the only guiding principles, HOMOGEN also explicitly detects and rectifies structural problems that cannot be recovered by satisfying the imposed constraints.…”

Section: Related Workmentioning

confidence: 99%

Exploiting Homogeneity of Density in Incremental Hierarchical Clustering

Widyantoro¹

2006

itbj.eng.sci.

View full text Add to dashboard Cite

Abstract. Hierarchical clustering is an important tool in many applications. As it involves a large data set that proliferates over time, reclustering the data set periodically is not an efficient process. Therefore, the ability to incorporate a new data set incrementally into an existing hierarchy becomes increasingly demanding. This article describes HOMOGEN, a system that employs a new algorithm for generating a hierarchy of concepts and clusters incrementally from a stream of observations. The system aims to construct a hierarchy that satisfies the homogeneity and the monotonicity properties. Working in a bottom-up fashion, a new observation is placed in the hierarchy and a sequence of hierarchy restructuring processes is performed only in regions that have been affected by the presence of the new observation. Additionally, it combines multiple restructuring techniques that address different restructuring objectives to get a synergistic effect. The system has been tested on a variety of domains including structured and unstructured data sets. The experimental results reveal that the system is able to construct a concept hierarchy that is consistent regardless of the input data order and whose quality is comparable to the quality of those produced by non incremental clustering algorithms.

show abstract

“…4.4. In Nevins (1995), an incremental branch-and-bound clusterer for the formation of hierarchies was introduced. Since addition of new observations can have a severe effect on the existing hierarchy, re-insertion of instances and clusters is performed during the formation process.…”

Section: Related Workmentioning

confidence: 99%

Cluster-grouping: from subgroup discovery to clustering

Zimmermann

Raedt

2009

Mach Learn

View full text Add to dashboard Cite

We introduce the problem of cluster-grouping and show that it can be considered a subtask in several important data mining tasks, such as subgroup discovery, mining correlated patterns, clustering and classification. The algorithm CG for solving cluster-grouping problems is then introduced, and it is incorporated as a component in several existing and novel algorithms for tackling subgroup discovery, clustering and classification. The resulting systems are empirically compared to state-of-the-art systems such as CN2, CBA, Ripper, Autoclass and CobWeb. The results indicate that the CG algorithm can be useful as a generic local pattern mining component in a wide variety of data mining and machine learning algorithms.

show abstract

A branch and bound incremental conceptual clusterer

Cited by 4 publications

References 18 publications

Iterative Optimization and Simplification of Hierarchical Clusterings

Iterative Optimization and Simplification of Hierarchical Clusterings

Exploiting Homogeneity of Density in Incremental Hierarchical Clustering

Cluster-grouping: from subgroup discovery to clustering

Contact Info

Product

Resources

About