public Node initRoot(List<Record> data) {
Node root = DecisionTree.INSTANCE.getRoot();
SampleDistribute sampleDis = root.getSampleDistribute();
for (Record record : data) {
sampleDis.acceptOneRecord(record);
}
root.calculteEntropy();
root.setPotentialClassifierAttrs(new HashSet<>(Data.INSTANCE.getDeterminingAttributes()));
return root;
}
private double calcEntropy(Node parent, Collection<Node> children) {
int total = parent.getTotal();
return children
.stream()
.map(e -> ((double) e.getTotal() / (double) total) * e.getEntropy())
.reduce(0.0, (x, y) -> x + y);
}
private Collection<DecisionTree.Node> getChildrenForBestClassifierAttribute(Node node) {
Collection<DecisionTree.Node> result = null;
Set<String> sets = new HashSet<>(node.getPotentialClassifierAttrs());
if (sets.isEmpty()) return null;
String selectedAttri = "";
double min = Double.MAX_VALUE;
for (String attribute : sets) {
Collection<DecisionTree.Node> children = generateChildren(node, attribute);
double entropy = calcEntropy(node, children);
if (entropy < min) {
min = entropy;
selectedAttri = attribute;
result = children;
}
}
node.setBestClassifierAttribute(selectedAttri);
return result;
}
public Collection<Node> generateChildren(Node node, String attribute) {
Collection<Node> children = node.generateChildren(attribute);
node.setChildren(new ArrayList<Node>(children));
return children;
}
