Data mining and knowledge discovery via logic-based methods. Theory, algorithms, and applications. (Q986032)

The book presents theoretical foundations, algorithms, and applications of the field of data mining and knowledge discovery (DM\&KD). It is divided into two parts. The first part (chapters 1 through 8) addresses the theory and algorithmic issues, while the second part (chapters 9 through 17) discusses applications of data mining and presents a number of examples and case studies. Chapter 1 provides an introduction to the DM\&KD field by examining the DM\&KD process and reviewing some of its applications. Chapter 2 deals with the core issue in data mining, namely how to derive Boolean functions from positive and negative examples. After reviewing some of the current developments in the field, it presents a method for transferring a nonbinary data into equivalent binary notation and discusses data processing. Chapter 3 presents a revised branch-and-bound algorithm for inferring a single clause from two disjoint sets of binary training examples. Performance characteristics of the method are studied and evaluated. In Chapter 4, a heuristic approach for inferring Boolean functions from examples with polynomial time complexity is presented. This approach encompasses two closely related problems, namely how to infer a Boolean function fast from two disjoint collections of positive and negative examples, and from examples which we have partial knowledge of. Chapter 5 deals with the guided learning problem, while Chapter 6 studies the problem of inferring a Boolean function in an incremental way and introduces a new incremental learning from example algorithms. Chapter 7 discusses a relationship between the CNF and DNF forms of Boolean functions derivable from the same training data, while Chapter 8 presents the motivation and definition of a special graph derived from positive and negative examples. Chapter 9 discusses reliability issues in data mining in the context of computer-aided breast cancer diagnosis. Chapter 10 deals with the problem of learning monotone Boolean functions with the underlying objective to efficiently acquire simple and intuitive knowledge which can be validated and has a general representation power. Chapter 11 discusses general application issues of monotone Boolean functions to DM\&KD problems. It presents a simple design problem in the car industry and discusses the accuracy of diagnostic systems. Chapter 12 deals with mining of association rules from databases and presents a new approach to it, while Chapter 13 discusses data mining of text documents as a classification problem in which a document must be classified into one of two disjoint classes. The next chapter presents a case study on predicting muscle fatigue from EMG signals. The data used in the study are available for download on the Web. Chapter 15 presents the second case study on inference of diagnostic rules for breast cancer in which data from a number of clinical cases of breast cancer diagnoses are used. Chapter 16 describes a fuzzy logic approach for quantifying some of the attributes involved in diagnosing breast cancer. Conclusions and the future of DM\&KD research are discussed in the final chapter.