Fernando Berzal

TBAR: An efficient method for association rule mining in relational databases

Abstract In this paper, we propose a new algorithm for efficient association rule mining, which we apply in order to discover interesting patterns in relational databases. Our algorithm, which is called Tree-Based Association Rule mining (TBAR), redefines the notion of item and employs an effective tree data structure. It can also use techniques such as Direct Hashing and Pruning (DHP). Experiments with real-life datasets show that TBAR outperforms Apriori, a well-known and widely used algorithm.

Building multi-way decision trees with numerical attributes

Decision trees are probably the most popular and commonly used classification model. They are recursively built following a top-down approach (from general concepts to particular examples) by repeated splits of the training dataset. When this dataset contains numerical attributes, binary splits are usually performed by choosing the threshold value which minimizes the impurity measure used as splitting criterion (e.g. C4.5 gain ratio criterion or CART Ginis index). In this paper we propose the use of multi-way splits for continuous attributes in order to reduce the tree complexity without decreasing classification accuracy. This can be done by intertwining a hierarchical clustering algorithm with the usual greedy decision tree learning.

ART: A Hybrid Classification Model

This paper presents a new family of decision list induction algorithms based on ideas from the association rule mining context. ART, which stands for ‘Association Rule Tree’, builds decision lists that can be viewed as degenerate, polythetic decision trees. Our method is a generalized “Separate and Conquer” algorithm suitable for Data Mining applications because it makes use of efficient and scalable association rule mining techniques.

AN ALTERNATIVE APPROACH TO DISCOVER GRADUAL DEPENDENCIES

In this paper we propose a new definition of gradual dependence as a special kind of association rule. We propose a way to adapt existing association rule mining algorithms for the new task of mining such dependencies, and we discuss about its complexity. Some experiments in a real database illustrate the usefulness of the approach.

A Survey of Link Prediction in Complex Networks

Networks have become increasingly important to model complex systems composed of interacting elements. Network data mining has a large number of applications in many disciplines including protein-protein interaction networks, social networks, transportation networks, and telecommunication networks. Different empirical studies have shown that it is possible to predict new relationships between elements attending to the topology of the network and the properties of its elements. The problem of predicting new relationships in networks is called link prediction. Link prediction aims to infer the behavior of the network link formation process by predicting missed or future relationships based on currently observed connections. It has become an attractive area of study since it allows us to predict how networks will evolve. In this survey, we will review the general-purpose techniques at the heart of the link prediction problem, which can be complemented by domain-specific heuristic methods in practice.

On the quest for easy-to-understand splitting rules

Decision trees are probably the most popular and commonly used classification model. They are built recursively following a top-down approach (from general concepts to particular examples) by repeated splits of the training dataset. The chosen splitting criterion may affect the accuracy of the classifier, but not significantly. In fact, none of the proposed splitting criteria in the literature has proved to be universally better than the rest. Although they all yield similar results, their complexity varies significantly, and they are not always suitable for multi-way decision trees. Here we propose two new splitting rules which obtain similar results to other well-known criteria when used to build multi-way decision trees, while their simplicity makes them ideal for non-expert users.

Component-based data mining frameworks

OLAP Vs. OLTP in the middle tier.

A GENERAL FRAMEWORK FOR COMPUTING WITH WORDS IN OBJECT-ORIENTED PROGRAMMING

Computing with words (CWW) techniques have been shown to be useful in the management of imperfect information. From the programmers standpoint, new tools are necessary to ease the use of these techniques within current programming platforms. This paper presents a step in this direction by describing a general framework that supports the implementation of applications dealing with fuzzy objects. We pay special attention to the study of the object comparison problem by offering both a theoretical analysis and a simple and transparent way to use our theoretical results in practice.

Managing Fuzziness on Conventional Object-Oriented Platforms

During the last few years, many database researchers have aimed their efforts at extending the object‐oriented model for dealing with different kinds of imperfect information. Some of these scholars have used the Fuzzy Set Theory to deal with imperfection because it has proved to be useful in problems where imprecision and uncertainty play important roles. This article describes an architecture that can be used to develop a fuzzy object‐oriented system on top of an existing classical one. This article also introduces a general framework as the basis for managing fuzziness in conventional object‐oriented systems. Foodbi, a fuzzy object‐oriented database interface, is presented as a prototype that allows the creation of fuzzy object‐oriented schemata that can be translated into sets of standard Java classes.

A definition for fuzzy approximate dependencies

In the analysis of data stored in databases, a very interesting issue is the detection of possible existing relations between attribute values and, at an upper level, relations between attributes themselves. In case uncertainty is present in data, or it is introduced in a pre-processing step, specific data mining and knowledge discovery techniques and methodologies must be provided. The theory of fuzzy subsets is a helpful tool to reach this goal. In this paper we introduce a new definition and an algorithm for computing fuzzy approximate dependencies, a type of relations that can be found between attributes in a fuzzy database, on the basis of a previous definition of fuzzy association rule. We will discuss about possible applications of this new tool.