Saso D Zeroski And Nada Lavra C (eds).

Relational data mining is the study of methods for knowledge discovery in databases when the database contains information about several types of objects. This, of course, is usually the case when the database has more than one table. Hence, there is little doubt as to the relevance of the area. Relational data mining has its roots in inductive logic programming, an area in the intersection of machine learning and programming languages. Early work in this area aimed at the synthesis of non-trivial programs from examples and background knowledge. The results were quite fascinating, but the true applicability of the techniques became clear only when the focus changed to the discovery of useful pieces of information from large collections of data, i.e. when the techniques started to be applied to data mining issues. This book is a collection of contributions from several authors who worked in the field. It provides quite an extensive overview of different techniques and strategies used in knowledge discovery from multi-relational data, and describes several interesting applications. The book is divided into four parts. The first part places relational data mining in the wider context of data mining and knowledge discovery, with contributions by Džeroski (overview of data mining), Fayyad (overview of knowledge discovery), Džeroski and Lavrač (introduction to inductive logic programming) and Wrobel (inductive logic programming for knowledge discovery in databases). The second part of the book presents a number of relational data mining approaches, including the learning of relational decision trees, relational classification and association rules, and distance-based approaches to relational learning and clustering. The contributions are by De Raedt et al. (a description of three systems based on the ILP framework to induce classification rules, decision trees and integrity constraints), Kramer and Widmer (classification and regression trees), Muggleton and Firth (relational classification rules), Dehaspe and Toivonen (discovery of relational association rules) and Kirsten et al. (relational upgrades of the k-NN method, hierarchical agglomerative clustering and k-means clustering). The third part discusses the links between propositional and relational learning by showing how a single table data mining approach can be upgraded to a relational data mining context, and vice versa, how a propositional approach can be used in a relational context after transforming the multi-table data to a single table. Contributions to this part are from: Van Laer and De Raedt, who describe a generic approach of upgrading a propositional learner to a relational one; Kramer et al., who describe how a relational data mining problem can be transformed to a single table data mining problem; Quinlan, who shows how the technique of boosting can fruitfully be applied to a relational learner to improve his performances; Getoor et al., who describe an extension of probabilistic modelling to probabilistic relational models and present techniques for finding such models from multi-table data. The fourth and last part of the book is concerned with the practice of relational data mining, and presents several applications of relational data mining in a number of areas, such as drug design, protein structure and function, medicine and engineering. The book provides quite a narrow view of relational data mining, mainly focusing on inductive logic programming techniques and their successful applications, ignoring contributions from other fields, such as evolutionary computation, connectionist learning and pattern recognition, in which several approaches have been developed. Moreover, computational complexity does not seem to be an important issue in the material presented, and so it is unclear as to what the scaling capabilities of the described approaches to very large databases are. Anyhow, the book may stimulate the interest for practical applications of relational data mining and further research in the development of relational data mining techniques.