Another great book missing in the article is "Mining Massive Datasets" [1] by Jure Leskovec, Anand Rajaraman and Jeff Ullman.

[1] http://infolab.stanford.edu/~ullman/mmds.html

I think the book rank 2 in this article (The Elements of Statistical Learning) is too mathematical.

i agree that it is mathematical. i would not have said "too", but i guess that is relative.

i reckon it is useful to have some base level of understanding of statistics, mathematics, and applied mathematics (eg solving (regularised?) systems of linear equations or numerically optimising functions) before attempting to understand something like machine learning.

understanding the theory behind the approaches gives you some basis for assessing if a given technique will be appropriate in a given application.

(eg there's no data and no ability to get more? probably can't do anything. there is data, but it was collected in a haphazard uncontrolled way, and no ability to get more? probably can't do anything...)

There should be a data mining textbook derived from data-mined data mining textbooks.