Elisabet Golobardes

Computer aided diagnosis with case-based reasoning and genetic algorithms

This article addresses breast cancer diagnosis using mammographic images. Throughout, the diagnosis is done using the mammographic microcalcifications. The aim of the work presented here is twofold. First, we introduce a back-end phase, based on machine learning techniques, in a previous computer aided diagnosis system. The two machine learning techniques incorporated are case-based reasoning and genetic algorithms. These algorithms look for improving the results obtained by human experts and the previous statistical model. On the other hand, we analyse the obtained results comparing them with the ones provided by other well-known machine learning techniques. The breast cancer dataset used in the experiments come from Girona Health Area. This database contains 216 images previously diagnosed by surgical biopsy.

Large-Scale Experimental Evaluation of Cluster Representations for Multiobjective Evolutionary Clustering

Multiobjective evolutionary clustering algorithms are based on the optimization of several objective functions that guide the search following a cycle based on evolutionary algorithms. Their capabilities allow them to find better solutions than with conventional clustering algorithms if the suitable individual representation is selected. This paper provides a detailed analysis of the three most relevant and useful representations-prototype-based, label-based, and graph-based-through a wide set of synthetic data sets. Moreover, they are also compared to relevant conventional clustering algorithms. Experiments show that multiobjective evolutionary clustering is competitive with regard to other clustering algorithms. Furthermore, the best scenario for each representation is also presented.

Unsupervised case memory organization: analysing computational time and soft computing capabilities

There are problems that present a huge volume of information or/and complex data as imprecision and approximated knowledge. Consequently, a Case-Based Reasoning system requires two main characteristics. The first one consists of offering a good computational time without reducing the accuracy rate of the system, specially when the response time is critical. On the other hand, the system needs soft computing capabilities in order to construct CBR systems more tractable, robust and tolerant to noise. The goal of this paper is centred on achieving a compromise between computational time and complex data management by focusing on the case memory organization (or clustering) through unsupervised techniques. In this sense, we have adapted two approaches: 1) neural networks (Kohonen Maps); and 2) inductive learning (X-means). The results presented in this work are based on datasets acquired from medical and telematics domains, and also from UCI repository.

Improving data partition schemes in Smart Grids via clustering data streams

Abstract Data mining techniques are traditionally divided into two distinct disciplines depending on the task to be performed by the algorithm: supervised learning and unsupervised learning. While the former aims at making accurate predictions after deeming an underlying structure in data—which requires the presence of a teacher during the learning phase—the latter aims at discovering regular-occurring patterns beneath the data without making any a priori assumptions concerning their underlying structure. The pure supervised model can construct a very accurate predictive model from data streams. However, in many real-world problems this paradigm may be ill-suited due to (1) the dearth of training examples and (2) the costs of labeling the required information to train the system. A sound use case of this concern is found when defining data replication and partitioning policies to store data emerged in the Smart Grids domain in order to adapt electric networks to current application demands (e.g., real time consumption, network self adapting). As opposed to classic electrical architectures, Smart Grids encompass a fully distributed scheme with several diverse data generation sources. Current data storage and replication systems fail at both coping with such overwhelming amount of heterogeneous data and at satisfying the stringent requirements posed by this technology (i.e., dynamic nature of the physical resources, continuous flow of information and autonomous behavior demands). The purpose of this paper is to apply unsupervised learning techniques to enhance the performance of data storage in Smart Grids. More specifically we have improved the eXtended Classifier System for Clustering (XCSc) algorithm to present a hybrid system that mixes data replication and partitioning policies by means of an online clustering approach. Conducted experiments show that the proposed system outperforms previous proposals and truly fits with the Smart Grid premises.

Explanations of unsupervised learning clustering applied to data security analysis

Network security tests should be periodically conducted to detect vulnerabilities before they are exploited. However, analysis of testing results is resource intensive with many data and requires expertise because it is an unsupervised domain. This paper presents how to automate and improve this analysis through the identification and explanation of device groups with similar vulnerabilities. Clustering is used for discovering hidden patterns and abnormal behaviors. Self-organizing maps are preferred due to their soft computing capabilities. Explanations based on anti-unification give comprehensive descriptions of clustering results to analysts. This approach is integrated in Consensus, a computer-aided system to detect network vulnerabilities.

Rough Sets Reduction Techniques for Case-Based Reasoning

Case Based Reasoning systems are often faced with the problem of deciding which instances should be stored in the case base. An accurate selection of the best cases could avoid the system being sensitive to noise, having a large memory storage requirements and, having a slow execution speed. This paper proposes two reduction techniques based on Rough Sets theory: Accuracy Rough Sets Case Memory (AccurCM) and Class Rough Sets Case Memory (ClassCM). Both techniques reduce the case base by analysing the representativity of each case of the initial case base and applying a different policy to select the best set of cases. The first one extracts the degree of completeness of our knowledge. The second one obtains the quality of approximation of each case. Experiments using different domains, most of them from the UCI repository, show that the reduction techniques maintain accuracy obtained when not using them. The results obtained are compared with those obtained using well-known reduction techniques.

A Methodology for Analyzing Case Retrieval from a Clustered Case Memory

Case retrieval from a clustered case memory consists in finding out the clusters most similar to the new input case, and then retrieving the cases from them. Although the computational time is improved, the accuracy rate may be degraded if the clusters are not representative enough due to data geometry. This paper proposes a methodology for allowing the expert to analyze the case retrieval strategies from a clustered case memory according to the required computational time improvement and the maximum accuracy reduction accepted. The mechanisms used to assess the data geometry are the complexity measures. This methodology is successfully tested on a case memory organized by a Self-Organization Map.

Data classification through an evolutionary approach based on multiple criteria

Real-world problems usually present a huge volume of imprecise data. These types of problems may challenge case-based reasoning systems because the knowledge extracted from data is used to identify analogies and solve new problems. Many authors have focused on organizing case memory in patterns to minimize the computational burden and deal with uncertainty. The organization is usually determined by a single criterion, but in some problems, a single criterion can be insufficient to find accurate clusters. This work describes an approach to organize the case memory in patterns based on multiple criteria. This new approach uses the searching capabilities of multiobjective evolutionary algorithms to build a Pareto set of solutions, where each one is a possible organization based on the relevance of objectives. The system shows promising capabilities when it is compared with a successful system based on self-organizing maps. Due to the data set geometry influences, the clustering building process results are analyzed taking into account it. For this reason, some complexity measures are used to categorize data sets according to their topology.

Analysis of vulnerability assessment results based on CAOS

Abstract: Information system security must battle regularly with new threats that jeopardize the protection of those systems. Security tests have to be run periodically not only to identify vulnerabilities but also to control information systems, network devices, services and communications. Vulnerability assessments gather large amounts of data to be further analyzed by security experts, who recently have started using data analysis techniques to extract useful knowledge from these data. With the aim of assisting this process, this work presents CAOS, an evolutionary multiobjective approach to be used to cluster information of security tests. The process enables the clustering of the tested devices with similar vulnerabilities to detect hidden patterns, rogue or risky devices. Two different types of metrics have been selected to guide the discovery process in order to get the best clustering solution: general-purpose and specific-domain objectives. The results of both approaches are compared with the state-of-the-art single-objective clustering techniques to corroborate the benefits of the clustering results to security analysts.

Integration of a Methodology for Cluster-Based Retrieval in jColibri

One of the key issues in Case-Based Reasoning (CBR) systems is the efficient retrieval of cases when the case base is huge and/or it contains uncertainty and partial knowledge. Although many authors have focused on proposing case memory organizations for improving the retrieval performance, there is not any free open source framework which offers this kind of capabilities. This work presents a plug-in called Thunder for the j colibri framework. Thunder provides a methodology integrated in a graphical environment for managing the case retrieval from cluster based organizations. A case study based on tackling a Textual CBR problem using Self-Organizing Maps as case memory organizing technique is successfully tested.