Intelligent Data Analysis: An International Journal

The ever-increasing number of fielded Data Mining applications is evidence that the technology works and produces added value in a variety of business areas. Most of the research-lab generated algorithms have found their way under various guises in a number of commercial software packages. When considering the use of Data Mining, the average business user is now faced with a plethora of DM software to choose from. In order to be informed, such a choice requires a standard basis from which to compare and contrast alternatives along relevant, business-focused dimensions, as well as the location of candidate tools within the space outlined by these dimensions. This paper aims at meeting this business requirement. It presents a standard schema for the characterisation of Data Mining software tools and the results of a recent survey of 41 popular Data Mining tools described within this schema.

This paper addresses the issue of reducing the storage requirements on instance-based learning algorithms. Algorithms proposed by other researches use heuristics to prune instances of the training set or modify the instances themselves to achieve a reduced set of instances. This paper presents an alternative way. The presented approach proposes to induce a reduced set of prototypes (partially-defined instances) with evolutionary algorithms. Experiments were performed with GALE, a fine-grained parallel evolutionary algorithm, and other well-known reduction techniques on several data sets. Results suggest that GALE is competitive and robust for inducing sets of partially-defined instances. Moreover, it achieves better reduction rates in storage requirements without losses in generalization accuracy. Simultaneously, if the partially-defined instances induced by GALE are post-processed, results can also be used for attribute selection.

This paper proposes a new similarity measure between basket datasets based on associations. The new measure is calculated from support counts using a formula inspired by information entropy. Experiments on both real and synthetic datasets show the effectiveness of the measure. This paper then investigates the applications of the similarity measure. It first studies the problem of finding a mapping between categorical database attribute sets using similarity measures. A generic approach for identifying such a mapping is proposed. The approach is implemented based on the similarity measure proposed in the paper and its performance has been evaluated and validated. Moreover, this paper also explores the applications of using the similarity measure to mine distributed datasets.

The goal of Web page categorization is to classify Web documents into a certain number of predefined categories. Previous works in this area employed a large number of labeled training documents for supervised learning. The problem is that, it is difficult to create labeled training documents. Though it is not so easy to manually categorize unlabeled documents for creating training data, it is easy to collect unlabeled ones. Therefore, a new machine learning algorithm is investigated to overcome these difficulties and effectively utilize unlabeled documents. We propose a novel approach called Iterative Cross-Training (ICT). In this paper, we applied the algorithm to Web page categorization on three data sets. The performance of ICT was evaluated and analyzed with the supervised learning algorithms, Co-Training and Expectation Maximization. We found that ICT is considered to be an effective approach for the Web page categorization task.

The paper discusses disparity issues in sensing tasks involving the production of a ‘high-level’ signal from ‘low-level’ signal sources. It introduces an abstraction theory which helps to explain the nature of the problem and point the way to a solution. It proposes a solution based on the use of supervised adaptive methods drawn from artificial intelligence. Finally, it describes a set of empirical experiments which were carried out to evaluate the efficacy of the method.