Relationships between members of a social network come in many shapes, forms, and intensities. This book exploits new methods for analyzing and modeling such relationships, as it presents models that take into consideration the many qualifiers among individual relationships in a group. It is not a book for beginners, but rather targeted at experienced modelers and software engineers interested in the inner workings of modeling and analyzing social networks. Furthermore, this book is about comparing the authors’ proposed approach to the existing state of the art and demonstrating how their approach is better in terms of correctness. If this is your area, then this is a book for you.

The book is divided into 11 chapters and six appendices, and provides a more practical “how to” view of modeling social networks. The reader interested in formal proofs will find them in the first five appendices or must consult the references cited.

After a compelling preface and introduction, chapter 2 introduces a few data structure alternatives to represent social networks with rich semantics. Next, chapter 3 covers the required background and notation.

The next five chapters present and explain different approaches to modeling relationships of different types (for example, friend, family, or work), asymmetric relationships, relationships that change over time, and networks with positive and negative relationships. Each chapter includes examples or real-life applications of the network type just covered, and compares their proposed heuristics with the previous state of the art in terms of “correctness” or insight quality, that is, how their modeled graphs and results are either a better fit to known patterns or provide better information than previous methods. Other than brief mentions here and there that one method is more (computationally or representatively) efficient than another, there is no complexity analysis of any of the models presented.

Chapter 9 takes a detour into machine learning. The problem considered is how to add labels to nodes in a network in which only a few nodes have been labeled, with the disclaimer that this will only work if the nodes that are similar for other network properties are likely to be similar for the new property being considered. Again, the proposed method is compared to the current state of the art.

In chapter 10, we are back to modeling networks with rich semantics. This time, combining directed and signed edge properties.

Appendices A to E cover formal proofs of some of the theorems used, and appendix F provides MATLAB code for the functions discussed. There is a glossary at the beginning of the book for the mathematical functions defined and most abbreviations.