Paradoxically, in these days of information glut, there is a concurrent problem of data loss--missing and incomplete data. Statisticians have generated a wealth of knowledge on the methods of handling missing data. While solving differential equations, it is common to encounter such problems as missing initial conditions, missing boundary conditions, and unspecified location of the boundary contour. Such inverse problems, said to be improperly posed, are often solved by regularization--a method of systematically guessing the missing values. This book makes an attempt to summarize the state of the art in the use of computational intelligence techniques to estimate the values of missing entries, but it fails to mention the possibility of searching the Web as a possible solution. Except for this omission, the book is a decent attempt at assembling the available tools under one cover.
A list of the book’s chapters should give potential readers an idea of the scope of the material covered: “Introduction to Missing Data,” “Estimation of Missing Data Using Neural Networks and Genetic Algorithms,” “A Hybrid Approach to Missing Data: Bayesian Neural Networks, Principal Component Analysis and Genetic Algorithms,” “Maximum Expectation Algorithms for Missing Data Estimation,” “Missing Data Estimation Using Rough Sets,” “Support Vector Regression for Missing Data Estimation,” “Committee of Networks for Estimating Missing Data,” “Online Approaches to Missing Data Estimation,” “Missing Data Approaches to Classification,” “Optimization Methods for Estimation of Missing Data,” “Estimation of Missing Data Using Neural Networks and Decision Trees,” “Control of Biomedical Systems Using Missing Data Approaches,” and “Emerging Missing Data Estimation Problems: Heteroskedasticity, Dynamic Programming and Impact of Missing Data.”
Although the coverage is broad, I got the impression that the chapters were independently written; many ideas are repeated from one chapter to the next.
Marwala repeatedly refers to an HIV/AIDS dataset, but does not provide a Web address where this dataset can be found. If he did, the book would have served a better purpose. What good are the results if the reader who wishes to repeat the experiments described cannot locate the dataset?
The author never claims this to be a reference volume, but it is definitely not a textbook and it is not suitable for self-learning either.