def test_id3_gain_criteria(defs, db, chi_table, CI="0%"):
print("\n==== Testing id3 using gain criteria =====\n")
gain_criteria = InformationGainCriteria()
#pdb.set_trace()
tree = id3(gain_criteria, mydb, 'class', mydefs.attr_set,
mydefs, chi_table, CI)
tree.print_entire_tree()
def Run_over_dataset(file_path): # run each algorithm 5 times with same datasets for i in range(5): dataset1 = Load_dataset(file_path) # switch train and test dataset2 = deepcopy(dataset1) dataset2.train_dataset = dataset1.test_dataset dataset2.test_dataset = dataset1.train_dataset k_nearest_neighbor(dataset1) k_nearest_neighbor(dataset2) # naive_bayes(dataset1) # naive_bayes(dataset2) # tan(dataset1) # tan(dataset2) id3(dataset1) id3(dataset2) print()
def test_classify(testfilename, chi_table):
print("\n\n==== Test classification====")
testdb = ShroomDatabase([], testfilename)
record = testdb.records[0]
gain_criteria = InformationGainCriteria()
gain_tree = id3(gain_criteria, mydb, 'class',
mydefs.attr_set, mydefs, chi_table)
gain_classification = gain_tree.classify(record)
misclass_error_criteria = ClassificationErrorCriteria()
misclass_tree = id3(misclass_error_criteria, mydb,
'class', mydefs.attr_set, mydefs, chi_table)
misclass_classification = misclass_tree.classify(record)
print("\nClassification under gain and misclassification:")
print("record to classify: " + record.get_raw_string())
print("(gain) classification: " + gain_classification)
print("(misclass) classification: " + misclass_classification)
def generate_id3_tree(self, selection_criteria,
datadef_filename, training_filename,
chi_table, CI="0%"):
"""Generates an ID3 classification tree."""
data_defs = ShroomDefs(datadef_filename)
training_db = ShroomDatabase([], training_filename)
classification_tree = id3(selection_criteria, training_db,
'class', data_defs.attr_set,
data_defs, chi_table, CI)
return classification_tree
def test_id3_misclass_criteria(defs, db, chi_table, CI="0%"):
print("\n==== Testing id3 using classifcation-error criteria =====\n")
misclass_error_criteria = ClassificationErrorCriteria()
tree = id3(misclass_error_criteria, mydb, 'class',
mydefs.attr_set, mydefs, chi_table, CI)
tree.print_entire_tree()
from results import *
#%% use training set to build decision trees
trees = {}
depths = [1, 2, 4, 8]
t_st = time.time()
for maxDepth in depths:
trees[maxDepth] = {}
print('\nMax Depth:', maxDepth)
for i in np.arange(200):
df = dataTrn.sample(int(0.1 * len(dataTrn)), replace=True)
# input id3(data, maxDepth, init_depth)
trees[maxDepth][i] = id3(df.values, maxDepth, 0)
if np.mod(i, 10) == 0:
print('.', end=" ")
t_en = time.time()
t_RunTrees = np.round((t_en - t_st) / 60, 3)
print('\nRuntime (m):', t_RunTrees)
#% transformer data
print('\n\nEnsemble Training Data')
dataTrfm_trn = transformData(dataTrn, trees, depths)
print('\n\nEnsemble Testing Data')
dataTrfm_tst = transformData(dataTst, trees, depths)
print('\nEnsemble Cross-Validation Data')
from pre_processing import *
from id3 import *
import random
input = pre_process_ratings()
data = input[::1000]
test = input
tree = id3(data)
#questions = ["Genero", "Idade", "Ocupação", "Categoria"]
#printtree(tree, questions)
"""
num_pos = 0
num_neg = 0
for t in test:
b = t[-1]
if b == 1:
num_pos += 1
else:
num_neg += 1
"""
correct = 0
fp_id3, fn_id3, tp_id3, tn_id3 = 0, 0, 0, 0
fp_priori, fn_priori, tp_priori, tn_priori = 0, 0, 0, 0
fp_random, fn_random, tp_random, tn_random = 0, 0, 0, 0
for t in test:
# Decision Tree
a = predict(t[:-1], tree)
