Tree-based Machine Learning Algorithms: Decision Trees, Random Forests, and Boosting

A good book if you want to program decision trees from scratch. Sadly this isn't what I needed. Entirely my as I should have read the description more carefully, it does clearly state: "it isn't about ... using pre-built machine learning libraries"

This book is only codes with no explanation on what a decision tree is. This is not a problem per se because the tile is tree based machine learning "algorithms". The problem with the book is its poor explanations. There are blocks of codes without showing the output. What I expected was to see some high level explanations first and then going to details. This book does not have high level concept at all and explains the details terribly.

Good effort in explaining DT / RF from scratch.Often, the problem with traditional programmers (and academics) is that they know how to write an algorithm from scratch and then believe they know how to use it in the real world.Maybe worse, beginners believe that too.Don't get me wrong, there's certainly a lot of value in knowing how an algorithm works from scratch, but that's only 50%.In the real world, you use an API. Many programmers/academics see using an API as below them.They often don't see that an API needs technical knowledge as well. Just different knowledge like:- what to do if the dataset is unbalanced?- how heuristic post pruning works - fast?- how to modify the most important parameter: max_leaf_nodes?- how does the tree make decisions (graphviz)?- how to calculate probabilities by hand (graphviz)?- at what lead nodes depth does the algorithm start overfitting?- how to inspect the behavior of DT vs RF?- can your random state distort your accuracy (attention: small dataset)?- ...As Jason Brownlee stated perfectly: in ML, work backward. It's simply too much.In other words, if you fancy writing ML algorithm from scratch, do it if you understand how to apply it. Of course, if you're a researcher like Andrew Trask (he writes DL code from scratch), then go ahead...

This book would allow the intermediate and above programmer to craft their own tree based learning algorithms and begin applying them to problems. I believe it is a pretty good start for someone that just wants to get up running without a lot of theory. I think it could be a little bit more in depth and needs another chapter to really finish off boosting. That said it is the best I have encountered for concisely delivering enough information to get started with machine learning.