export ruleset to ruby conditional blocks and method

[newx]

Generated ruby code example: \ this editor removes original code's indentation

def classify(hunger, happiness) if (hunger >= 4.5 and happiness >= 4.0) then 'angry' elsif (hunger >= 4.5 and happiness < 4.0) then 'not angry' elsif (hunger < 4.5) then 'not angry' else nil end end

def classify(hunger, color) if (hunger == 'no') then 'not angry' elsif (hunger == 'yes' and color == 'blue') then 'not angry' elsif (hunger == 'yes' and color == 'red') then 'angry' else nil end end

[newx]

@igrigorik do you think it is good enough? Let me know if you have any comment on it... thanks

[igrigorik]

Silly question.. Have you verified that this works for nested trees?

[newx]

@igrigorik Do you have an example of nested trees? I'm not sure if you mean "bagging". So if you point me an example of nested tree I'll test it and work on it if it breaks :)

[igrigorik]

@newx any tree with multiple variables would do the trick. For example, the one in this post: https://www.igvita.com/2007/04/16/decision-tree-learning-in-ruby/

[newx]

@igrigorik unfortunately it breaks with multiple variables ): . But I'll work on it. Thanks

[newx]

@igrigorik here it goes. This is my first interaction to make it work with nested trees (or value ranges). Now it works and generates the code below. Let me know if you think something that can be improved or is just wrong. Thanks...

def my_classify_method(age, education, income, marital_status) if ((age.to_f >= 18 and age.to_f <= 35)) then 'will not buy' elsif ((age.to_f >= 36 and age.to_f <= 55) and marital_status == 'married') then 'will not buy' elsif ((age.to_f >= 36 and age.to_f <= 55) and marital_status == 'single') then 'will buy' elsif (age.to_f <= 18) then 'will buy' elsif (age.to_f >= 55) then 'will buy' else nil end end

[igrigorik]

I'm not sure this makes sense.. You shouldn't be looking for "value ranges". The range is an arbitrary and opaque attribute (i.e. it's just a string). The code generator should be a recursive function that generates a nested if that matches the output decision tree. E.g:

[igrigorik]

(feel free to reopen once/if previous comments are addressed)