def make_tree(train_array):
global current_instances
global blocked_features
global tree_nodes
current_instances = []
while (calculate_global_entropy(current_instances) == 0):
current_instances.append(train_array.pop(0))
choose_root()
for ins in train_array:
(should_change, new_node_place, child_node,
parent_node) = tree_root.add_instance(ins)
current_instances.append(ins)
if (should_change):
#blocked_features = list(set().union(blocked_features , list(child_node.get_path_to_root().keys()) ))
#print(child_node.get_path_to_root())
#print(blocked_features)
related_instances = query(child_node.get_path_to_root())
new_feature_name = get_best_feature(parent_node, related_instances)
if (new_feature_name != None):
#blocked_features.append(new_feature_name) ####
(new_featre_total_labels,
new_feature_attr_labels) = get_feature_labels(
new_feature_name, related_instances)
new_node = TreeNode(parent_node, new_feature_name,
new_feature_attr_labels, None, True, True,
new_featre_total_labels)
parent_node.children_nodes[
new_node_place].visual_node.parent = None
parent_node.children_nodes[new_node_place] = new_node
try:
tree_nodes.remove(tree_nodes.index(child_node))
except:
pass
new_node_children_nodes = {}
new_node_children_labels = {}
tree_nodes.append(new_node)
for attr_name in list(new_node.children_labels.keys()):
temp_node = TreeNode(new_node, None, None, None, True,
False,
new_node.children_labels[attr_name])
new_node_children_nodes[str(attr_name)] = temp_node
tree_nodes.append(temp_node)
new_node.children_nodes = new_node_children_nodes
else:
continue
check_for_pull_up()
for pre, fill, node in RenderTree(tree_root.visual_node):
print("%s%s" % (pre, node.name))
def choose_root():
global current_labels
global current_instances
global tree_nodes
global visual_nodes
global tree_root
global blocked_features
tree_nodes = []
feature_dict = {}
for feature in list(current_instances[0].feature_dict.keys()):
if (feature not in blocked_features):
feature_dict[feature] = calculate_global_entropy(
current_instances) - get_feature_entropy(feature)
#print(feature_dict)
feature_name = max(feature_dict, key=feature_dict.get)
node = TreeNode(parent=None,
feature_name=str(feature_name),
children_labels=None,
children_nodes=None,
is_leaf=True,
is_feature=True,
labels=get_global_labels())
root = Node(node.feature_name)
blocked_features.append(node.feature_name)
visual_nodes.append(root)
tree_nodes.append(node)
tree_root = node
node_children_labels = {}
node_children_nodes = {}
for attr in get_feature_attrs(node.feature_name):
temp_node = TreeNode(parent=node,
feature_name=None,
children_labels=None,
children_nodes=None,
is_leaf=True,
is_feature=False,
labels=get_feature_attrs(node.feature_name)[attr])
temp_visual_node = Node(str(attr) + "[" + str(temp_node.labels[0]) +
"," + str(temp_node.labels[1]) + "]",
parent=root)
node_children_labels[str(attr)] = temp_node.labels
node_children_nodes[str(attr)] = temp_node
tree_nodes.append(temp_node)
node.children_labels = node_children_labels
node.children_nodes = node_children_nodes
current_instances = []
#print(node.feature_name)
return
def create_tree(ins_array):
global tree_root
global blocked_features
for ins in ins_array:
(should_change, new_node_place, child_node,
parent_node) = tree_root.add_instance(ins)
if (should_change):
blocked_features = list(set().union(
blocked_features, list(child_node.get_path_to_root().keys())))
#print(child_node.get_path_to_root())
#print(blocked_features)
related_instances = query(child_node.get_path_to_root())
new_feature_name = get_best_feature(parent_node, related_instances)
if (new_feature_name != None):
blocked_features.append(new_feature_name) ####
(new_featre_total_labels,
new_feature_attr_labels) = get_feature_labels(
new_feature_name, related_instances)
new_node = TreeNode(parent_node, new_feature_name,
new_feature_attr_labels, None, True, True,
new_featre_total_labels)
parent_node.children_nodes[
new_node_place].visual_node.parent = None
parent_node.children_nodes[new_node_place] = new_node
try:
tree_nodes.remove(tree_nodes.index(child_node))
except:
pass
new_node_children_nodes = {}
new_node_children_labels = {}
tree_nodes.append(new_node)
for attr_name in list(new_node.children_labels.keys()):
temp_node = TreeNode(new_node, None, None, None, True,
False,
new_node.children_labels[attr_name])
new_node_children_nodes[str(attr_name)] = temp_node
tree_nodes.append(temp_node)
new_node.children_nodes = new_node_children_nodes
else:
continue
return
def check_for_pullup():
global tree_nodes
global tree_root
for key in tree_root.children_nodes.keys():
n = tree_root.children_nodes[key]
now_instances = tree_root.seen_instances
if ((n.is_feature)):
print(n.feature_name)
if (n.entropy() < n.parent.entropy()):
node = TreeNode(parent=None,
feature_name=str(n.feature_name),
children_labels=None,
children_nodes=None,
is_leaf=True,
is_feature=True,
labels=n.labels)
tree_root = node
node_children_labels = {}
node_children_nodes = {}
for attr in n.children_nodes.keys():
print(attr)
temp_node = TreeNode(parent=node,
feature_name=n.parent.feature_name,
children_labels=None,
children_nodes=None,
is_leaf=True,
is_feature=False,
labels=get_feature_attrs(
node.feature_name)[attr])
node_children_labels[str(attr)] = temp_node.labels
node_children_nodes[str(attr)] = temp_node
tree_nodes.append(temp_node)
node.children_labels = node_children_labels
node.children_nodes = node_children_nodes
create_tree(now_instances)
print("------------")
return
