Search

Outlier detection being an important data mining problem has attracted a lot of research interest in the recent past. As a result, various methods for outlier detection have been developed particularly for dealing with numerical data, whereas categorical data needs some attention. Addressing this requirement, we propose a two-phase algorithm for detecting outliers in categorical data based on a novel definition of outliers. In the first phase, this algorithm explores a clustering of the given data, followed by the ranking phase for determining the set of most likely outliers. The proposed algorithm is expected to perform better as it can identify different types of outliers, employing two independent ranking schemes based on the attribute value frequencies and the inherent clustering structure in the given data. Unlike some existing methods, the computational complexity of this algorithm is not affected by the number of outliers to be detected. The efficacy of this algorithm is demonstrated through experiments on various public domain categorical data sets. [ABSTRACT FROM AUTHOR]

Abstract: Nearest neighbor (NN) classifier is a popular non-parametric classifier. It is conceptually a simple classifier and shows good performance. Due to the curse of dimensionality effect, the size of training set needed by it to achieve a given classification accuracy becomes prohibitively large when the dimensionality of the data is high. Generating artificial patterns can reduce this effect. In this paper, we propose a novel pattern synthesis method called partition based pattern synthesis which can generate an artificial training set of exponential order when compared with that of the given original training set. We also propose suitable faster NN based methods to work with the synthetic training patterns. Theoretically, the relationship between our methods and conventional NN methods is established. The computational requirements of our methods are also theoretically established. Experimental results show that NN based classifiers with synthetic training set can outperform conventional NN classifiers and some other related classifiers. [Copyright &y& Elsevier]

Abstract: Kullback–Leibler relative-entropy has unique properties in cases involving distributions resulting from relative-entropy minimization. Tsallis relative-entropy is a one-parameter generalization of Kullback–Leibler relative-entropy in the nonextensive thermostatistics. In this paper, we present the properties of Tsallis relative-entropy minimization and present some differences with the classical case. In the representation of such a minimum relative-entropy distribution, we highlight the use of the q-product, an operator that has been recently introduced to derive the mathematical structure behind the Tsallis statistics. One of our main results is the generalization of triangle equality of relative-entropy minimization to the nonextensive case. [Copyright &y& Elsevier]

This paper introduces a novel clustering scheme employing a combination of rough set theory and fuzzy set theory to generate meaningful abstractions from web access logs. Our experimental results show that the proposed scheme is capable of capturing the semantics involved in web access logs at an acceptable computational expense. [Copyright &y& Elsevier]

Presents a study that proposed a stochastic connectionist approach for solving function optimization problems with application to pattern clustering. Aim of cluster analysis; Information on the squared-error criterion function; Advantages of the stochastic connectionist approach; Experimental results.

Proposes a novel hybrid genetic algorithm that finds a globally optimal partition of a given data into a specified number of clusters. Importance of cluster analysis; Partitional clustering; Experimental study.

Proposed two algorithms, along with their neural network implementations, to overcome certain limitations of the earlier learning subspace methods. Extended subspace methods of pattern recognition; Growing subspace method; Growing learning subspace method; Neural network model for the growing subspace methods.

Abstract: Shannon entropy of a probability measure P, defined as on a measure space , is not a natural extension from the discrete case. However, maximum entropy (ME) prescriptions of Shannon entropy functional in the measure-theoretic case are consistent with those for the discrete case. Also it is well known that Kullback–Leibler relative entropy can be extended naturally to measure-theoretic case. In this paper, we study the measure-theoretic aspects of nonextensive (Tsallis) entropy functionals and discuss the ME prescriptions. We present two results in this regard: (i) we prove that, as in the case of classical relative-entropy, the measure-theoretic definition of Tsallis relative-entropy is a natural extension of its discrete case, and (ii) we show that ME-prescriptions of measure-theoretic Tsallis entropy are consistent with the discrete case with respect to a particular instance of ME. [Copyright &y& Elsevier]

Abstract: In this paper a novel kernel-based soft clustering method is proposed. This method incorporates rough set theoretic flavour in support vector clustering paradigm to achieve soft clustering. Empirical studies show that this method can find soft clusters having arbitrary shapes. [Copyright &y& Elsevier]