Customer Reviews

Biologically Inspired Algorithms for Financial Modelling (Natural Computing Series)

5 star:		(0)
4 star:		(3)
3 star:		(0)
2 star:		(1)
1 star:		(0)

 

 

In the ceaseless search for better modelling of financial instruments and economic events, one approach is to look for methods from mathematical biology as inspiration. Here, the main approaches studied include neural networks, genetic algorithms and ant colony modelling. The first two are perhaps the most widely used.

The key inspiration is to look into the future. The later sections of the book involve predicting various events, like a corporate failure. The efficacy of the biological methods for doing predictions is unclear. The book's results are intriguing, though.

There appear to be 2 audiences for the book. One is biologists or programmers already using those methods in biology, and who are looking at applying these to finance. The other audience is mathematicians in finance wanting more tools. Consequently, each audience will find different portions of the book useful. The explanation of conventional financial modelling is for the biologist, for example.

Although this is a good book on using bio-inspired algorithms for financial modelling, I cannot give it five stars for a simple reason: Its price currently being somewhere around 90$ is way over the top for a book containing only 275 pages. (Even the Kindle-edition of the book is approaching 90$.) The books content is just not worth that much. Compare it for instance to Barry Johnson's highly recommendable book "Algorithmic Trading and DMA: An introduction to direct access trading strategies", consisting of nearly 600 pages (though paperback only), having a price of around 40$ - 50$. Or Stephen Marsland's book "Machine Learning: An Algorithmic Perspective (Chapman & Hall/Crc Machine Learning & Pattern Recognition)" (406 pages) at a price of roughly 60$ - 70$.

Anyway, let's discuss the books contents. The book is well written and clearly structured, so it's fun to read it. It basically consists of three parts: The first part explains the theories and introduces several bio-inspired algorithms and how they work. The second part covers implementation details and gives hints about typical implementation problems you'll run into. It also gives you a really brief introduction to technical analysis. And the third part contains case studies where for most algorithm types one implementation example is given.

Algorithms covered by the book are 1. neural networks, 2. genetic algorithms, 3. grammatical evolution, 4. particle swarm models, 5. ant colony models and 6. artificial immune systems. These (and only these) are all well explained, and pseudo-code examples are often given. You'll however still have to bring your own programming knowledge to be able to implement these algorithms yourself and you'll certainly have to have a basic understanding of statistics. Not covered in the book however are for instance P-systems/membrane systems and cellular automata.

What is important to know also is that the book mainly targets at modelling stock exchange trading situations - not financial situations in general. Typical problems addressed include portfolio management of stocks or bonds including classification decisions, trend analysis and prediction and therefore buy/sell/out decisions for traders.

So to sum it up all: It's a good book, it's just too expensive.

I usually don't spend time writing reviews, but I did want to comment on the very
negative review made in Dec 2009 because I found this review misleading.

I am interested in both Biologically Inspired Algorithms (BIA) and Financial Modelling (FM).
I have done a lot of research myself using Neural Networks, and I am currently working
on a project regarding Financial Modelling.

I have read only parts of the book yet, but I already found those chapters very
helpful. The fields of BIA and FM are very broad so, of course, no text book can
exhaustively cover both fields. Nevertheless, the authors explain very nicely the fundamental
steps that one needs to consider when building a system which uses BIA for FM.
The first parts of the book introduce the fundamental aspects (BIA and FM) whereas the
third part of the book provides case studies.

I liked the book (as far as I have read it yet) very much: It is well structured and
well written and it provides a good overview. For in depth information it provides a lot
of references.

So for everyone interested in getting an overview of the topic this book is a clear
recommendation. I would have given 5 stars, but I have to admit that at some points a bit
more detail would not have gone amiss.

"Biologically inspired optimization algorithms with financial applications" would be a better title. Apart from the models embedded in the algorithms themselves, the models on hand are statistical - regression and classification - used to predict stock prices and corporate defaults.

The first problem is initially handled with multilayer perceptrons: having settled on a specific network structure, one employs backpropagation to search for optimal network weights, effectively coefficients in a non-linear regression. Next, one makes the network structure part of the optimization problem: a genetic algorithm (GA) experiments with different configurations (and starting weights), while backpropagation continues to tune coefficients for each profile. A distinct approach employs GA alone, developing it into genetic programming (GP), which performs a "smart" search over sequences of operator/value strings, forming transformation-defining expressions.

Genetic programming accounts for a third of Part I and 6 out of 10 case studies in Part III. It seems fair to judge the book by how well it covers its central topic. It fails. Things get difficult to follow just as GP is introduced on p. 54: a "syntax tree" is shown without any explanation - then again, the section on radial-basis-function networks never said what a radial-basis function is - and the plausible question about how GA can handle the valid-syntax constraint is unanswered. Details pile on, onto a foundation that's not there. Implementation remains unclear, and if a book about GP does not tell you how to build GP, what good is it?

Subpar writing is found elsewhere in Part I, especially in the sections on radial-basis-function networks and ant-colony optimization (ACO). Two recurring annoyances are failure to spell out an algorithm - ACO again, not detailed until Part III - and gratuitous biology coverage. (Artificial immune system algorithms: biology 5 pages, algorithms 2 pages). Scattered typos give extra evidence of limited editorial effort.

The short Part II will be interesting for those new to finance; others will find the information decently presented but familiar. Part II reviews "technical indicators", and Part III, of course, presents 10 finance-themed case studies. In my view, these are unremarkable - imagine a short project in a master's computer-science course - and at best clarify the points left out in Part I. (Chapter 18, for example; impressively, the case study forgets to finish its analysis).

"BIAFM" is not up to the standards that I expect from Springer, and does not justify the investment of $90.