We are trying to do naive bayes classification in Ruby.
Currently we are using http://ai4r.rubyforge.org/
We couldn't get it working for float values and have about 20% points lower accuracy with string values. With float/integers we get a [] no implicit conversion from nil to integer. We convert floats with to_s.
Is there a way to get float values working? If not what are alternative gems to ai4r for ruby or alternative algorithms to naive bayes?
The "...implicit conversion..." error is coming from within the Ruby interpreter in a context where an Integer is required but a nil is found.
It's a bit hard to get this error but one way to see it is:
Array.new(nil)
...so that's probably what is happening in your case.
And in fact, gems/ai4r-1.9/lib/ai4r/classifiers/naive_bayes.rb does actually use the rather rare construct of Array.new(...).
The value passed to Array.new() depends on what Ai4r::Data::DataSet returns from #build_domains.
I know this doesn't solve your problem but perhaps it will help you proceed further with your analysis.
I know this question is an older one, but I wanted to point out a more robust Ruby gem for Naive Bayes classification:
