Module BC_BDTree

module BC_BDTree: sig  end

Binary decision trees

type ('b, 'a) btree =

`\|`	`Leaf of 'a`	`(*`	`a leaf of type 'b`	`*)`
`\|`	`Node of 'b * ('b, 'a) btree * ('b, 'a) btree`	`(*`	`a node consists of a (parameter for a) decision function of type 'a, and two child nodes`	`*)`

the type for binary decision trees

val bdtree : ('a, 'b) btree -> ('c -> 'a -> bool) -> 'c -> 'b

bdtree spec matcher creates a function for classification. spec is the structure of the decision tree, and matcher is a function for discrimination, using the element of type 'a at the node as a parameter

'a : parameter type at node of tree
'b : classification class type
'c : input type to binary decision tree (data type)

val prune_tree : ('a, 'b) btree -> ('a, 'b) btree

prune useless branches from a tree

val binary_id3 : ?maxdepth:int ->
       ?scoref:(int -> int -> int -> int -> float) ->
       'a list ->
       'a list ->
       ((int -> int -> int -> int -> float) -> 'a list -> 'a list -> 'b) ->
       ('a -> 'b -> bool) -> ('b, bool) btree

ID3 algorithm (Quinlan, 19??) for binary classes. arguments in following order:

maxdepth: maximum depth of the tree. If negative, don't care. If positive, subtract 1. If 0 don't go any further. Default is -1
scoref: score function. Arguments in order of: posmatch, negmatch, postotal, negtotal
pd positive data
nd negative data
bestfinder algorithm for finding the best parameter ('a in Node 'a) given scoref, pd, and nd
matcher interpreter of the parameter

val bdt_to_string : ('a, bool) btree -> ('a -> string) -> string

convert a boolean tree to a string. the second argument is a function which converts the parameters at the node to a string

val serialize : ('a, 'b) btree -> ('a -> string) -> ('b -> string) -> string

convert binary decision trees to strings which can be used as internal representation

first arg: tree
second arg: conversion function for parameters at Nodes ('a -> string)
third arg: conversion function for Leaves ('b -> string)

val entropy_gain : int -> int -> int -> int -> float

entropy gain score function: for ID3