When I was originally asked to write this book I said no, several times. My reticence was a consequence of workload and not because I thought there was no need for the book. However, Katrina Halliday from Cambridge University Press persisted and eventually I gave in. However, I can only blame myself for any errors or omissions in the text.

Katrina asked me because of my editorship of an earlier machine learning book and favourable comments about some of my web material, in particular a postgraduate multivariate statistics unit. My interest in multivariate statistics arose from my research as a conservation biologist. As part of this research I spent time trying to develop predictive species distribution models, which led to my exploration of two additional topics: machine learning methods as alternatives to statistical approaches; and how to measure the accuracy of a model's predictions.

If you are not an ecologist you may be thinking that there will be little of value for you in the book. Hopefully, the contents will alleviate these fears. My multivariate statistics unit was delivered to a diverse group of students including biomedical scientists, so I am used to looking beyond ecology and conservation. In my experience there is much to be gained by straying outside of the normal boundaries of our research. Indeed my own research into the accuracy of ecological models drew greatly on ideas from biomedical research.