golang-decisiontree

A simple decision tree algorithm that should definitely be used by no one

As a first exercise, while learning golang, I decided to implement a regular categorical decision tree. I'm sure the style is horrible and many golang conventions are violated. Additionally, I didn't teach myself how to package golang files appropriately, so everything is under package main.

intSample.go

• intSample - this struct holds a single data point for use in prediction
  • NewIntSample - init function (not a method)
  • Feature - method returns the value associated with the named feature

intFeature.go

• intFeature - this is just an int slice
  • Filter - this method takes in an equal length []bool mask and returns the filtered feature
  • NumSamples - Returns the length of the feature
  • Get - gets the element in position idx
  • Unique - returns a dictionary mapping the unique elements to their counts
  • Entropy - calculates the entropy of the feature using log2
  • String - converts the feature to a string for printing

intDataset.go

• intDataset - This is the main dataset type used as an input to the decision tree
  • Note - I might want to consider breaking out the target from the features.
  • NewIntDataset - initialization function (not a method)
  • NumFeatures - returns the number of features (including the target) included in the dataset
  • NumSamples - the number of training samples provided
  • GetFeature - returns the intFeature corresponding to the named feature
  • SubsetByFeature - takes in a feature name, returns a partition of the dataset based on the feature
  • MutualInformation - calculates the mutual information or information gan between the target and the supplied feature
  • MajorityClass - returns the most commonly occurring class in the target
  • String - String method for printing

decisionTree.go

• decisionTree - this is a helper type used to recursively form the tree, you can mostly ignore it
  • NewDecisionTree - init function (not method), called by NewDecisionTreeLearner
  • Predict - takes in an intSample and provides a prediction
  • Depth - returns the depth of the tree

DecisionTreeLearner.go

• DecisionTreeLearner - the main type of the project
  • NewDecisionTreeLearner - takes in tree parameters and a dataset and initializes a new trained tree
  • Predict - predicts the class of an intSample
  • Depth - returns the depth of the decision tree