def iterative_reduced_error_pruning(training_set_size, validation_set_range,
data):
"""
Implements reduced error pruning on a tree build by the id3 algorithm.
Gets training, validation, and testing error for each iteration by naively creating a new tree for
all training set sizes between 1 and training_set_size.
Returns lists of training, validation, and testing error vs number of instances trained on from 1 to
training_set_size.
"""
train_error_points, validation_error_points, test_error_points = [], [], []
for cur_training_set_size in range(1, training_set_size):
decision_tree = id3(data.iloc[:cur_training_set_size])
validation_error = _calc_error(validation_set_range, decision_tree,
data)
_mark_nodes(decision_tree, decision_tree, data, validation_error,
validation_set_range)
train_error_points.append(
_calc_error(range(cur_training_set_size), decision_tree, data))
validation_error_points.append(
_calc_error(validation_set_range, decision_tree, data))
test_error_points.append(
_calc_error(range(80, len(data)), decision_tree, data))
if cur_training_set_size % 5 == 0:
print("Iterations left: ",
training_set_size - cur_training_set_size)
return train_error_points, validation_error_points, test_error_points
def tennis():
"""Builds, displays, and tests the tennis.csv on the id3 algorithm."""
data = read_csv('tennis.csv')
decision_tree = id3(data)
print(decision_tree)
for idx in range(len(data)):
print("Correct?: ", _is_correct_label(decision_tree, data, idx),
" Value: ", decision_tree.traverse(data.iloc[idx]))
def zoo():
"""Builds, displays, and tests the zoo.csv on the id3 algorithm."""
data = read_csv('zoo.csv').drop('animal_name', 1)
decision_tree = id3(data.iloc[:80])
print(decision_tree)
for idx in range(80, len(data)):
print("Correct?: ", _is_correct_label(decision_tree, data, idx),
" Value: ", decision_tree.traverse(data.iloc[idx]))
def zoo_iterative_id3(training_set_size):
"""
Builds and tests the zoo.csv on the id3 algorithm.
Gets training and testing error for each iteration by naively creating a new tree for all training set sizes
between 1 and training_set_size.
Returns lists of training and testing error vs number of instances trained on from 1 to training_set_size.
"""
train_error_points, test_error_points = [], []
data = read_csv('zoo.csv').drop('animal_name', 1)
for cur_training_set_size in range(1, training_set_size):
decision_tree = id3(data.iloc[:cur_training_set_size])
train_error_points.append(
_calc_error(range(cur_training_set_size), decision_tree, data))
test_error_points.append(
_calc_error(range(80, len(data)), decision_tree, data))
if cur_training_set_size % 5 == 0:
print("Iterations left: ",
training_set_size - cur_training_set_size)
return train_error_points, test_error_points
from ID3 import id3
from InfoGain import Gain
from BuildFromTraining import loadFromFile
gain = {}
##################################################
print("Cargando ejemplos de entrenamiento...")
S = loadFromFile("training_examples.txt")
Attributes = list(S[0])
##################################################
###
##################################################
#print("Calculando ganancia de informacion...")
#for A in Attributes[:-1]:
# gain[A] = Gain(S, A)
##################################################
print("Atributos: {}".format(Attributes))
id3(S, "", Attributes[:-1])
N = 5
one = True
two = True
three = True
if one:
print("=====================================")
print("========== Experiment 3.2.1 =========")
print("=====================================")
print("")
trees = []
for i in range(N):
e = ensemble_data(hand_train_data, hand_train_label)
random.shuffle(hand_train_attr)
t = id3(e['d'], e['l'], hand_train_attr[:8])
trees.append(t)
data_set = get_data(trees, hand_train_data)
W_b = run_svm(hand_train_label, data_set, epochs, 2, 0.001)
evaluation = evaluate_svm(hand_train_label, hand_train_data, W_b['W'],
W_b['b'])
print("=========== Training Data ===========")
print_eval(evaluation)
evaluation = evaluate_svm(hand_test_label, hand_test_data, W_b['W'],
W_b['b'])
print("============= Test Data =============")
print_eval(evaluation)
if two:
print("=====================================")
'''
Created on Sep 23, 2016
@author: Leland Stenquist
'''
from ID3 import scan, id3, print_tree, test_id3
from ID3_Helper import ensemble_data
r = scan("res/test/train.labels","res/test/train.data",50)
#r = scan("res/madelon/madelon_train.labels","res/madelon/madelon_train.data",25)
e = ensemble_data(r['d'],r['l'])
t = id3(e['d'],e['l'],r['a'],8)
print_tree(t)
test = test_id3(t,r['d'],r['l'],r['a'])
print(test)