search

Rambatino / CHAID

A python implementation of the common CHAID algorithm
Apache License 2.0
150 stars 50 forks source link

Refactored into classes and adding stopping rules #36

Closed Rambatino closed 7 years ago

Rambatino commented 8 years ago

Not sure if splitting the classes out in this PR should really happen, but the file was becoming gigantic and I was adding another method.

xulaus commented 8 years ago

Im not sure I get what rules is doing, could you either reword the doc sting or add some explanatory comments?

Rambatino commented 8 years ago

Cool yeah. It was a way I could avoid traversing the tree. Essentially, if you capture the heterogeneity in the respondents of that tree, (i.e. get the uniq rows) you then have all the different unique attributes that define that node. There's reason to this madness, it's so i can work out the FP, TP... by merging the new dataset with these rules and working out the accuracy