func (self *PredictionForest) Train(inputLog il.Interface) {
for _, key := range inputLog.Keys() {
var newTree dt.Interface
newTree = dt.New()
newTree.Train(inputLog, key)
self.Trees[key] = newTree
}
}
maxEntropyCall = inputLog.EXPECT().MaxEntropy(targetKey)
})
Describe("when there are no features left to decide on", func() {
BeforeEach(func() {
maxEntropyCall.Return(targetKey)
inputLog.EXPECT().Len().Return(7)
keyCountCall = inputLog.EXPECT().KeyCount(targetKey)
})
Context("when there are more occurances of the target key than absences", func() {
It("decides with a value of true", func() {
keyCountCall.Return(5)
mockBinaryTree.EXPECT().Set(true)
myTree.Train(inputLog, targetKey)
})
})
Context("when there are more absences of the target key than occurances", func() {
It("decides with a value of false", func() {
keyCountCall.Return(3)
mockBinaryTree.EXPECT().Set(false)
myTree.Train(inputLog, targetKey)
})
})
})
Describe("when there are still features to decide on", func() {
BeforeEach(func() {
nextKey = "the next key to decide on"
var myTree dt.Interface
BeforeEach(func() {
id3InputSet = il.InputLog{
{"1", "3"},
{"0", "1", "3"},
{"1", "3"},
{"0", "1", "3"},
{"1", "2"},
{"0", "1", "2"},
{},
{"0"},
}
myTree = dt.New()
myTree.Train(il.Interface(id3InputSet), "3")
})
Describe("Training a Decision Tree", func() {
It("it uses ID3 to maximize entropy", func() {
Expect(myTree.Value()).To(Equal("2"))
Expect(myTree.Left().Value()).To(Equal("1"))
})
It("doesn't continue building a branch if a decision can be made", func() {
Expect(myTree.Left().Right().Right()).To(BeNil())
})
It("builds a tree with the right decisions", func() {
Expect(myTree.Left().Right().Value()).To(Equal(true))
Expect(myTree.Left().Left().Value()).To(Equal(false))
