Year: 2014
Type: editorial
Abstract: This issue of the IDA journal consists of eleven articles, all covering various topics related to the theoretical and applied research in the field of Intelligent Data Analysis.The first five articles of this issue are about data understanding and search for patterns.Zimmerman in the first article of this issue discusses episode mining techniques and propose a process to compare these techniques using data simulation that mimics real world applications.This approach is based on recovering patterns that are embedded in noise.The experiments included in this article indicate that temporal constraints are more important in affecting the effectiveness of episode mining than occurrence semantics.Aghabozorgi and Wah in the second article of this issue discuss clustering of time series data and propose a new multi-step approach to accurately cluster time series data.From their experimental results that include data from different domains, the authors claim that their multi-step approach generates accurate clusters that are based on similarity in shape in very large time series data sets.Balasch-Masoliver et al. in the next article of this issue discuss the topic of multivariate aggregation in high dimensional data and argue that depending on the attributes chosen and their correlation, the amount of information loss and disclosure risk vary.The authors present a genetic algorithm based approach for multivariate microaggregation with its goal being finding optimal or near optimal attribute groupings taking into account both information loss and disclosure risk.The approach is extensively experimented with results included in this article.Visconti et al. in the fourth article of this issue discuss the topic of efficient gene expression profiling through integration of additional information such as functional annotations.The authors propose a new bi-clustering approach that includes gene associations.Their approach supports the extraction of biologically relevant bi-clusters through leveraging additional knowledge.This is demonstrated through experiments performed on several gene expression data reported in this article.The fifth article of this group, by Akdemir and Jannink is about the last step in search for patterns in data analysis.The authors propose some new methods for post-processing a large ensemble of conjunctive rules discovered through supervised, semi-supervised and unsupervised methods.They show through various examples that for high dimensional regression problems the models constructed through postprocessing the rules with partial least square regression have significantly better prediction performance than the ones produced by the random forest.The next three articles focus on a different aspect of learning and classification.Pevec and Kononenko compare and examine two families of non-parametric approaches that are common in construction of prediction intervals for arbitrary regression models in the supervised learning framework.The first of the families introduced in this article is based on the idea of explaining the total prediction error as a sum of model's error and the second is based on the assumption of similarity of data.Their results on a large set of real and artificial data demonstrate one of these proposed methods always generates valid and superior results comparing to others.Similarly, the seventh article of this issue by Kutsuna and Yamamoto is about a parameter-free approach to one-class classification where they argue that parameter tuning is a difficult process in all classes of learning, especially when one class is the subject of study.
Source: Intelligent Data Analysis
DOI: 10.3233/ida-140667
Author: A. Famili
Language: Spanish
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Keywords: episode mining, time series clustering, disclosure risk, gene expression profiling, biclustering, partial least squares regression, prediction intervals, nonparametric regression, pattern search, high-dimensional data, time constraints, shape similarity, functional annotation, information loss
Open Access status: bronze
OpenAlex ID: W4232856078
Cites:
Cited by:
Related to:
