Uncertain Data Analysis

This book studies different classification, detection and decision fusion algorithms, and helps practitioners deal with uncertainty in their data sets. Data uncertainties are considered as a collection of linguistic / fuzzy vectors, or a vector of fuzzy numbers and fuzzy algorithms are used to analyze these data sets. There are many theories and applications developed based on fuzzy set theory.

The topics of classification and prediction using fuzzy algorithms are introduced in the chapters on K-nearest prototype, clustering and neural networks. The Linguistic K-Nearest Prototype algorithm is designed to work with linguistic data represented by fuzzy vectors. This algorithm is particularly useful in fields where data is inherently imprecise or fuzzy, such as in management questionnaire analysis, where responses may not be strictly quantitative. The reader also learns about clustering algorithms such as Linguistic Hard C-means, Linguistic Fuzzy C-means, for single and multiple clusters respectively. The author explores the integration of Fuzzy Multilayer Perceptrons (FMLPs) with the Cuckoo Search (CS) algorithm to enhance the performance and applicability of neural networks in handling complex, fuzzy data. Two commonly used fuzzy integrals, covered are the Choquet integral and the Sugeno integral. Mathematical analysis of these algorithms is included in the study of the difference approaches it takes to aggregation of data. Both integrals are powerful tools for handling fuzzy data and their use to improve decision-making and analysis, is demonstrated using real-world application examples from both these algorithms. Very importantly, the topic of decision fusion is studied using Fuzzy Dempster-Shafer Theory with example of a real application provided.

This book serves as a guide for practitioners, such as robotics engineers, computer scientists and researchers working on computational intelligence. It is also suitable for graduate courses on fuzzy theories and fuzzy techniques.