def cross_validate(fold, data):
def make_chunks(seq, num_chunks):
avg = len(seq) / float(num_chunks)
out = []
last = 0.0
while last < len(seq):
out.append(seq[int(last):int(last + avg)])
last += avg
return out
random.shuffle(data)
chunks = make_chunks(data, fold)
err = []
err_prune = []
for i in range(fold):
test_data = chunks[i]
train_data = chunks[:i] + chunks[i+1:]
root = gen_tree(training_data)
err.append(test_accuracy(root, test_data)[1])
err_prune.append(test_accuracy(root, test_data, 2)[1])
return (err , err_prune)
def validation_error(test_data, train_data_orig, validate_ratio):
random.shuffle(train_data_orig)
# Make validation set
split = int(len(train_data_orig) * validate_ratio)
validate_data = train_data_orig[:split]
train_data = train_data_orig[split:]
# Initial tree
root = gen_tree(train_data)
# Find best pruning
least_err = float("inf")
best_prune = -1
for i in range(max_depth(root)):
err = test_accuracy(root, validate_data, i)[1]
if err < least_err:
least_err = err
best_prune = i
# Return err of best pruning against test data
return test_accuracy(root, test_data, best_prune)[1]
best_prune = -1
for i in range(max_depth(root)):
err = test_accuracy(root, validate_data, i)[1]
if err < least_err:
least_err = err
best_prune = i
# Return err of best pruning against test data
return test_accuracy(root, test_data, best_prune)[1]
training_data = load_training("wifi.train")
test_data = load_training("wifi.test")
root = gen_tree(training_data)
print "\n"
err_locs, err_all = test_accuracy(root, training_data)
print "TEST ACCURACY ON TRAINING SET:"
print "\tLOCATION SPECIFIC ERRORS:"
for loc, val in err_locs.iteritems():
print "\t\t{0} : {1}".format(loc,val)
print "\tOVERALL ERROR:\n\t\t{0}".format(err_all)
print "\n"
err_locs, err_all = test_accuracy(root, test_data)
print "TEST ACCURACY ON TEST SET:"
print "\tLOCATION SPECIFIC ERRORS:"
