Three approaches to data analysis. Test theory, rough sets and logical analysis of data. (Q2429140)

The aim of this book is to present three approaches to data analysis: test theory (TT), rough sets (RS) and logical analysis of data (LAD). The first part, by Yablonskii and Chegis, consists of seven sections. The first three contain a general study of decision trees, tests and rules (mainly notions, results related to bounds on complexity, algorithms for construction and optimization), results for tests and reducts obtained in TT (length of tests, number of reducts, algorithms for construction), local and global approaches for the study of decision trees and rule systems for problems over infinite or large finite sets of attributes. The next tree sections treat TT applications: pattern recognition, control and diagnosis of faults, discrete optimization (traveling salesman problem, unconditional and conditional optimization), recognition of words of fixed length in a regular language. The second part, proposed by Zdzisław I. Pawlak, contains 11 sections. The first three of them give an introduction to rough set theory (vague concepts, indiscernibility and discernibility relations, dependences, reducts, discernibility of Boolean reasoning, rough membership). Further, two extensions of rough sets are presented: one is based on approximation spaces and the other is an extension of the concept approximation to approximation of ontologies. In the next two sections, the rough set approach to scalability in data mining, based on the combination of rough sets and Boolean reasoning, and the relation between rough sets and logic are discussed. Finally, interactive rough granular computing as an approach for modeling interactive computation is proposed; in this way, three topics are analyzed: context inducing, process mining and perception based computing. The last part by Peter Hammer is devoted to the logical analysis of data. The theory section defines the main notions such as partially defined Boolean functions (theory, bi-theories, extension with missing bits) and patterns (Pareto-optimal patterns, maximum patterns, pattern generation). The next section presents the main steps which are used in LAD: binarization, attribute selection, pattern generation, model construction, validation and accuracy. Finally, various applications of LAD are analyzed. The book can be used successfully by researchers interested in data analysis.