def mode2(examples, target_attr):
v = []
db = ShroomDatabase(examples)
if target_attr == 'class':
v = db.fetch_class_vector()
else:
v = db.fetch_class_vector(target_attr)
return mode(v)
def mode2(examples, target_attr):
v = []
db = ShroomDatabase(examples)
if target_attr=='class':
v = db.fetch_class_vector()
else:
v = db.fetch_class_vector(target_attr)
return mode(v)
def id3(criteria, db, target_attr, attributes, defs):
v = db.fetch_class_vector()
homogeneous, label = is_homogeneous(v)
if (homogeneous):
return ID3Tree(ID3LeafNode(label))
if len(attributes) == 0:
label = mode2(examples, target_attr)[0]
return ID3Tree(ID3LeafNode(label))
A, stat = criteria.recommend_next_attr(attributes, db, defs)
decision_node = None
if isinstance(criteria, InformationGainCriteria):
decision_node = ID3DecisionNode(A, stat, 0.0)
else:
decision_node = ID3DecisionNode(A, 0.0, stat)
tree = ID3Tree(decision_node)
for v in defs.attr_values[A]:
edge = ID3Edge(A, v)
subset_records = []
for x in db.records:
if ((x.attributes[A] == v) and (x.attributes[A] != '?')):
subset_records.append(x)
if len(subset_records) == 0:
label = mode2(db.records, 'class')[0]
leaf_node = ID3LeafNode(label)
tree.add_node(decision_node, edge, leaf_node)
else:
subattributes = attributes - set([A])
subset_db = ShroomDatabase(subset_records)
subtree = id3(criteria, subset_db, target_attr, subattributes,
defs)
tree.add_tree(decision_node, edge, subtree)
return tree
def calc_all_class_error(attributes, db, defs):
"""Calculates the classification error for all attributes of a ShroomDatabase."""
tot_len = len(db.fetch_class_vector())
class_error_table = dict()
class_error_before = calc_class_error(db.fetch_class_vector())
for attr in attributes:
interim_db_list = []
for symbol in defs.attr_values[attr]:
interim_list = []
for r in db.records:
if r.attributes[attr] == symbol:
interim_list.append(r)
interim_db_list.append(ShroomDatabase(interim_list))
interim_class_error = 0
for idb in interim_db_list:
vec = idb.fetch_class_vector()
interim_class_error += (len(vec) / tot_len * calc_class_error(vec))
class_error_after = class_error_before - interim_class_error
class_error_table[attr] = class_error_after
return class_error_table
#!/usr/bin/env python
#
# filename: test_database.py
# authors: Jon David and Jarrett Decker
# date: Thursday, February 10, 2016
#
import pdb
from database import ShroomDatabase
trainfilename = "./data/training.dat"
testfilename = "./data/testing.dat"
outfilename = "./output.dat"
print("\n\n==== Test database ====")
print("training file: ", trainfilename)
print("testing file: ", testfilename)
print("output file: ", outfilename)
training_db = ShroomDatabase([], trainfilename)
test_db = ShroomDatabase([], testfilename)
out_db = ShroomDatabase(test_db.records)
out_db.save_data(outfilename)
print("\nVerify {} and {} are equal.".format(testfilename, outfilename))
print("Use diff -Z, to ignore whitespace at end of line")
#!/usr/bin/env python
#
# filename: test_database.py
# authors: Jon David and Jarrett Decker
# date: Thursday, February 10, 2016
#
import pdb
from database import ShroomDatabase
trainfilename = "./data/training.dat"
testfilename = "./data/testing.dat"
outfilename = "./output.dat"
print("\n\n==== Test database ====")
print("training file: ", trainfilename)
print("testing file: ", testfilename)
print("output file: ", outfilename)
training_db = ShroomDatabase([], trainfilename)
test_db = ShroomDatabase([], testfilename)
out_db = ShroomDatabase(test_db.records)
out_db.save_data(outfilename)
print("\nVerify {} and {} are equal.".format(testfilename,
outfilename))
print("Use diff -Z, to ignore whitespace at end of line")
from datadef import ShroomDefs
from database import ShroomDatabase
from id3_v2 import *
deffilename = "_shroom.data.definition"
trainfilename = "./data/training.dat"
testfilename = "./data/testing.dat"
print("\n\n==== Test id3_v2 ====")
print("definition file: ", deffilename)
print("training set: ", trainfilename)
mydefs = ShroomDefs(deffilename)
mydb = ShroomDatabase([], trainfilename)
#pdb.set_trace()
gain_criteria = InformationGainCriteria()
tree = id3(gain_criteria, mydb, 'class', mydefs.attr_set, mydefs)
tree.print_entire_tree()
print("\n\n==== Test classification====")
testdb = ShroomDatabase([], testfilename)
record = testdb.records[0]
classification = tree.classify(record)
print("record to classify: " + record.get_raw_string())
print("classification: " + classification)
#!/usr/bin/env python
#
# filename: test_gain.py
# authors: Jon David and Jarett Decker
# date: Thursday, February 4, 2016
#
from datadef import ShroomDefs
from database import ShroomDatabase
from id3 import *
deffilename = "_shroom.data.definition"
trainfilename = "./data/training.dat"
print("\n")
print("definition file: ", deffilename)
print("training set: ", trainfilename)
mydefs = ShroomDefs(deffilename)
mydb = ShroomDatabase([], trainfilename)
gain_table = calc_all_gain(mydefs.attr_set, mydb, mydefs)
print("\nGain table:")
print("===========================")
for attr in gain_table:
gain = gain_table[attr]
print(attr, ": ", gain)
rmend_attr, gain = recommend_next_attr(gain_table)
print("\nRecommend: ", rmend_attr, ", ", gain)