import magellan as mg

DEBUG:Cloud:Log file (/Users/aliHitawala/.picloud/cloud.log) opened

mg.init_jvm()

True

# read tables
A = mg.read_csv('/Users/aliHitawala/Documents/UW-2015/Fall 2015/784 - Data model/Phase2revisited/bikedekho_clean.csv', key='id')
B = mg.read_csv('/Users/aliHitawala/Documents/UW-2015/Fall 2015/784 - Data model/Phase2revisited/bikewale_clean.csv', key='id')

A.head(2)

B.head(2)

# Blocking - already done and stored in CC.csv
C = mg.read_csv('C_after_fix.csv', ltable=A, rtable=B)
len(C)

4932

# Plan

# 1. Sample candidate set --> S
# 2. Label S --> G
# 3. Split G into development set I, and evaluation set J
# 4. Select best learning-based matcher Y, using I 
# 5. Add triggers to Y ---> Z
# 6. Evaluate Z using J

# Sample the candidate set
S = mg.sample_table(C, 450)

# Label S
G = mg.label_table(S, 'gold')

G.to_csv('golden_3.csv')

True

# use pre-labeled data if present
G = mg.read_csv('golden.csv', ltable=A, rtable=B)
len(G)

450

# Split G into development (I) and evaluation (J)
IJ = mg.train_test_split(G, train_proportion=0.7)
I = IJ['train']
J = IJ['test']

# checking the number of tuples in I and J
(len(I), len(J))

(315, 135)

# Selecting the best learning-based matcher using I

# Plan 

# 1. Create a set of ML-matchers
# 2. Generate features --> feature_table
# 3. Extract feature vectors using I and feature_table
# 4. Select best learning-based matcher using CV
# 5. Debug the selected matcher (and repeat the above steps)

# Create a set of ML-matchers
dt = mg.DTMatcher(name='DecisionTree', random_state=0)
svm = mg.SVMMatcher()
rf = mg.RFMatcher(name='RF', random_state=0)
nb = mg.NBMatcher(name='NB')
lg = mg.LogRegMatcher(name='LogReg')
ln = mg.LinRegMatcher(name='LinReg')

# check names of the matchers
(dt.name, svm.name)

('DecisionTree', 'SVM_61402071609014438441')

# Generate features
feat_table = mg.get_features_for_matching(A, B)
# We faced this problem where the internal type of 'color' from both the table didn't match.

# To fix this, we explicitly initiated the types for 'color' attribute and passed 
# in the get_feature api

WARNING:magellan.feature.autofeaturegen:Magellan types: color type (str_eq_1w) and color type (str_bt_1w_5w) are different.If you want to set them to be same and generate features, update output from get_attr_types and use get_features command.

mg._atypes1['color'] = mg._atypes2['color']

feat_table = mg.get_features(A, B, mg._atypes1, mg._atypes2, mg._match_c,mg._match_t, mg._match_s)

feat_table
# feature generated after the 'color' fix

mg._match_s

{'abs_norm': <function magellan.feature.simfunctions.abs_norm>,
 'cosine': <function magellan.feature.simfunctions.cosine>,
 'exact_match': <function magellan.feature.simfunctions.exact_match>,
 'jaccard': <function magellan.feature.simfunctions.jaccard>,
 'jaro': <function magellan.feature.simfunctions.jaro>,
 'jaro_winkler': <function magellan.feature.simfunctions.jaro_winkler>,
 'lev': <function magellan.feature.simfunctions.lev>,
 'monge_elkan': <function magellan.feature.simfunctions.monge_elkan>,
 'needleman_wunsch': <function magellan.feature.simfunctions.needleman_wunsch>,
 'rel_diff': <function magellan.feature.simfunctions.rel_diff>,
 'smith_waterman': <function magellan.feature.simfunctions.smith_waterman>,
 'smith_waterman_gotoh': <function magellan.feature.simfunctions.smith_waterman_gotoh>,
 'soundex': <function magellan.feature.simfunctions.soundex>}

mg._match_t

{'dlm_dc0': <function magellan.feature.tokenizers.tok_delim>,
 'qgm_2': <function magellan.feature.tokenizers.tok_qgram>,
 'qgm_3': <function magellan.feature.tokenizers.tok_qgram>}

mg._match_c['corres']

[('id', 'id'),
 ('bike_name', 'bike_name'),
 ('city_posted', 'city_posted'),
 ('km_driven', 'km_driven'),
 ('color', 'color'),
 ('fuel_type', 'fuel_type'),
 ('price', 'price'),
 ('model_year', 'model_year'),
 ('owner_type', 'owner_type'),
 ('url', 'url')]

# Features used:
# Adding features related to 'bike_name' and 'color' features to get the best learning-based matcher 
# and using extra triggers on top of it to take care of boundary cases.

feat_subset_iter1 = (feat_table[3:11].append(feat_table[20:28]))

# Get feature vectors
K = mg.extract_feature_vecs(I, feature_table=feat_subset_iter1, attrs_after='gold')

K.head()

# select the best ML matcher using CV
result = mg.select_matcher([dt, rf, svm, nb, lg, ln], table=K, 
        exclude_attrs=['_id', 'ltable.id', 'rtable.id', 'gold'],
        target_attr='gold', metric='f1', random_state=1)

result['selected_matcher']

<magellan.matcher.rfmatcher.RFMatcher at 0x124e1e290>

result['cv_stats']

# Debug decision tree

# Split feature vectors to train and test
UV = mg.train_test_split(K, train_proportion=0.5, random_state=0)
U = UV['train']
V = UV['test']

mg.vis_debug_dt(dt, U, V, 
        exclude_attrs=['_id', 'ltable.id', 'rtable.id', 'gold'],
        target_attr='gold')

# Add features for 'bike_name' and 'color' to be used in ML algorithm and triggers
# 1. create two features for triggers - both are blackbox features
#   a. bike_name_dont_match: to label the output as not match if the bikename present in one but not in the other. Eg: Pulsar in left and not in right
#   b. bike_name_do_match: to label the output as match if the bikename present in both left and right, also the color should match
# 2. create another feature on jaro-winkler for ML algorithm on bike_name
# 3. Add the features create above to the feature table.

# define blackbox features

BIKENAME = ['pulsar', "avenger", 'passion', 'hunk', 'cbr', 'fazer', 'xtreme', 'karizma zmr',
           'yamaha fz16', 'stunner', 'splendor plus', 'splendor pro', 'discover', 'enfield thunderbird',
           'enfield electra', 'unicorn', 'dazzler', 'activa', 'enfield bullet']

def bike_name_dont_match(x, y):
    x_bike_name = x['bike_name'].lower()
    y_bike_name = y['bike_name'].lower()
    for bike in BIKENAME:
        if (bike in x_bike_name) != (bike in y_bike_name):
            return 0
    return 1

def bike_name_do_match(x, y):
    x_bike_name = x['bike_name'].lower()
    y_bike_name = y['bike_name'].lower()
    x_color = x['color'].lower()
    y_color = y['color'].lower()
    for bike in BIKENAME:
        if (bike in x_bike_name and bike in y_bike_name) and (x_color == y_color):
            return 1
    return 0

# Create a feature on bike_name to be used in the ML algorithm 
feature = mg.get_feature_fn("jaro_winkler(ltuple['bike_name'], rtuple['bike_name'])", mg._match_t, mg._match_s)

# Add features to feature table
mg.add_feature(feat_table, 'bike_name_bike_name_jw', feature)
mg.add_blackbox_feature(feat_table, 'bike_name_dont_match', bike_name_dont_match)
mg.add_blackbox_feature(feat_table, 'bike_name_do_match', bike_name_do_match)

True

feat_table

# Select all bike_name feature but not the jac0 and all the color features. Apart from those add
# newly created feature on bike_name (jaro-winkler)
feat_subset_iter2 = (feat_table[3:5].append(feat_table[6:10]).append(feat_table[20:28]).append(feat_table[52:53]))

# printing the selected features
feat_subset_iter2

# Get new set of feature vectors
K = mg.extract_feature_vecs(I, feature_table=feat_subset_iter2, attrs_after='gold')

# Split feature vectors into U and V
UV = mg.train_test_split(K, train_proportion=0.5, random_state=0)
U = UV['train']
V = UV['test']

# Check whether the added features improves the accuracy in the test set.
# Steps
# 1. Train DT using U
# 2. Predict V using DT
# 3. Evaluate predictions

# Train dt using U (use other algorithms also to get the predictions on them)
dt.fit(table=U, 
       exclude_attrs=['_id', 'ltable.id', 'rtable.id', 'gold'], 
       target_attr='gold')

# Predict V using dt
P = dt.predict(table=V, exclude_attrs=['_id', 'ltable.id', 'rtable.id', 'gold'], 
              append=True, target_attr='predicted', inplace=False)

# Writing it to the csv file to check the output
P.to_csv('test0.csv')

True

# Evaluate the predictions
eval_result = mg.eval_matches(P, 'gold', 'predicted')
mg.print_eval_summary(eval_result)

Precision : 88.24% (45/51)
Recall : 97.83% (45/46)
F1 : 92.78%
False positives : 6 (out of 51 positive predictions)
False negatives : 1 (out of 107 negative predictions)

# Apply cross validation to find if there is a better matcher
result = mg.select_matcher([dt, rf, svm, nb, lg, ln], table=K, 
        exclude_attrs=['_id', 'ltable.id', 'rtable.id', 'gold'],
        target_attr='gold', metric='f1')

result['cv_stats']

# Select DT as the best matcher -- Y
# Use bike_name + color related features

# Add triggers on top of Y

# 1. Split K into U and V
# 2. Use U,V + Y to write triggers (examine fp, fn).

# Split feature vectors to U and V
UV = mg.train_test_split(K, train_proportion=0.5, random_state=0)
U = UV['train']
V = UV['test']

# Invoke debug interface to check FP and FN
mg.vis_debug_dt(dt, U, V, 
        exclude_attrs=['_id', 'ltable.id', 'rtable.id', 'gold'],
        target_attr='gold')

# Add trigger - target false positives: using the bike_name blackbox feature created to check non matching bike_name in left and right. 
no_match_trigger = mg.MatchTrigger()
no_match_trigger.add_cond_rule('bike_name_dont_match(ltuple, rtuple) < 0.5', feat_table)
no_match_trigger.add_cond_status(True)
no_match_trigger.add_action(0)

# Add trigger - target false negatives: using the bike_name blackbox feature created to check matching bike_name in left and right. 
match_trigger = mg.MatchTrigger()
match_trigger.add_cond_rule('bike_name_do_match(ltuple, rtuple) > 0.5', feat_table)
match_trigger.add_cond_status(True)
match_trigger.add_action(1)

True

# Check whether the added trigger improves the accuracy in the test set.
# Steps
# 1. Train DT using U
# 2. Predict V using DT
# 3. Apply trigger
# 4. Evaluate the result

# Train dt using U
dt.fit(table=U, 
       exclude_attrs=['_id', 'ltable.id', 'rtable.id', 'gold'], 
       target_attr='gold')

# Predict V using dt
P = dt.predict(table=V, exclude_attrs=['_id', 'ltable.id', 'rtable.id', 'gold'], 
              append=True, target_attr='predicted', inplace=False)

P.to_csv('before_trigger.csv')

True

# Apply trigger
Q = no_match_trigger.execute(P, 'predicted', inplace=False)
R = match_trigger.execute(Q, 'predicted', inplace=False)

R.to_csv('after_trigger.csv')

True

# Evaluate the result
eval_result = mg.eval_matches(R, 'predicted', 'gold')
mg.print_eval_summary(eval_result)

Precision : 97.83% (45/46)
Recall : 100.0% (45/45)
F1 : 98.9%
False positives : 1 (out of 46 positive predictions)
False negatives : 0 (out of 112 negative predictions)

# Do cross-validation for matcher + triggers using I (K)
result = mg.cv_matcher_and_trigger(dt, [no_match_trigger, match_trigger], table = K, 
                                   exclude_attrs=['_id', 'ltable.id', 'rtable.id', 'gold'],
                                  target_attr='gold')

0%  100%
[#####] | ETA[sec]: 0.000 
Total time elapsed: 0.972 sec

result['cv_stats']

# Recall the CV for just the matcher (without triggers) was
result = mg.cv_matcher_and_trigger(dt, [], table = K, 
                                   exclude_attrs=['_id', 'ltable.id', 'rtable.id', 'gold'],
                                  target_attr='gold')

0%  100%
[#####] | ETA[sec]: 0.000 
Total time elapsed: 0.205 sec

result['cv_stats']

# Now Z is DT (features: feat_subset_iter2) + no_match_trigger, match_trigger
# Validate Z using J
# Steps
# 1. Extract feature vectors (using feat_subset_iter2) -- > M
# 2. Train DT using H (feature vectors generated using I)
# 3. Predict M using DT
# 4. Apply triggers
# 5. Evaluate the result

# Extract feature vectors
M = mg.extract_feature_vecs(J, feature_table=feat_subset_iter2, attrs_after='gold')

# Train using feature vectors from I 
dt.fit(table=K, 
       exclude_attrs=['_id', 'ltable.id', 'rtable.id', 'gold'], 
       target_attr='gold')

# Predict M 
N = dt.predict(table=M, exclude_attrs=['_id', 'ltable.id', 'rtable.id', 'gold'], 
              append=True, target_attr='predicted', inplace=False)

# Apply trigger
T = no_match_trigger.execute(N, 'predicted', inplace=False)
T = match_trigger.execute(T, 'predicted', inplace=False)

# Evaluate the result
eval_result = mg.eval_matches(T, 'gold', 'predicted')
mg.print_eval_summary(eval_result)

Precision : 100.0% (34/34)
Recall : 100.0% (34/34)
F1 : 100.0%
False positives : 0 (out of 34 positive predictions)
False negatives : 0 (out of 101 negative predictions)

T.to_csv('final_matches.csv')

True

	id	bike_name	city_posted	km_driven	color	fuel_type	price	model_year	owner_type	url
0	36082	Royal Enfield Bullet Electra Twinspark	Delhi	900	black	Petrol	110000	2015	FirstOwner	http://www.bikedekho.com/
1	36115	Royal Enfield Thunderbird 350	Mumbai	15000	black	Petrol	110000	2013	FirstOwner	http://www.bikedekho.com/

	id	bike_name	city_posted	km_driven	color	fuel_type	price	model_year	owner_type	url
0	12	TVS Apache RTR 160 Rear Drum Brake	Mumbai	28500	grey	Petrol	40000	2010	First	http://www.bikewale.com/used/bikes-in-mumbai/t...
1	20	Bajaj Discover 125 Disc	Mumbai	21300	green	Petrol	27000	2009	First	http://www.bikewale.com/used/bikes-in-mumbai/b...

	feature_name	left_attribute	right_attribute	left_attr_tokenizer	right_attr_tokenizer	simfunction	function	function_source
0	id_id_exm	id	id	None	None	exact_match	<function id_id_exm at 0x125024938>	from magellan.feature.simfunctions import *\nf...
1	id_id_anm	id	id	None	None	abs_norm	<function id_id_anm at 0x125024a28>	from magellan.feature.simfunctions import *\nf...
2	id_id_lev	id	id	None	None	lev	<function id_id_lev at 0x125024aa0>	from magellan.feature.simfunctions import *\nf...
3	bike_name_bike_name_jac_qgm_3_qgm_3	bike_name	bike_name	qgm_3	qgm_3	jaccard	<function bike_name_bike_name_jac_qgm_3_qgm_3 ...	from magellan.feature.simfunctions import *\nf...
4	bike_name_bike_name_cos_dlm_dc0_dlm_dc0	bike_name	bike_name	dlm_dc0	dlm_dc0	cosine	<function bike_name_bike_name_cos_dlm_dc0_dlm_...	from magellan.feature.simfunctions import *\nf...
5	bike_name_bike_name_jac_dlm_dc0_dlm_dc0	bike_name	bike_name	dlm_dc0	dlm_dc0	jaccard	<function bike_name_bike_name_jac_dlm_dc0_dlm_...	from magellan.feature.simfunctions import *\nf...
6	bike_name_bike_name_mel	bike_name	bike_name	None	None	monge_elkan	<function bike_name_bike_name_mel at 0x125024c80>	from magellan.feature.simfunctions import *\nf...
7	bike_name_bike_name_lev	bike_name	bike_name	None	None	lev	<function bike_name_bike_name_lev at 0x125024cf8>	from magellan.feature.simfunctions import *\nf...
8	bike_name_bike_name_nmw	bike_name	bike_name	None	None	needleman_wunsch	<function bike_name_bike_name_nmw at 0x125024d70>	from magellan.feature.simfunctions import *\nf...
9	bike_name_bike_name_sw	bike_name	bike_name	None	None	smith_waterman	<function bike_name_bike_name_sw at 0x125024de8>	from magellan.feature.simfunctions import *\nf...
10	bike_name_bike_name_swg	bike_name	bike_name	None	None	smith_waterman_gotoh	<function bike_name_bike_name_swg at 0x125024e60>	from magellan.feature.simfunctions import *\nf...
11	city_posted_city_posted_lev	city_posted	city_posted	None	None	lev	<function city_posted_city_posted_lev at 0x125...	from magellan.feature.simfunctions import *\nf...
12	city_posted_city_posted_jar	city_posted	city_posted	None	None	jaro	<function city_posted_city_posted_jar at 0x125...	from magellan.feature.simfunctions import *\nf...
13	city_posted_city_posted_jwn	city_posted	city_posted	None	None	jaro_winkler	<function city_posted_city_posted_jwn at 0x125...	from magellan.feature.simfunctions import *\nf...
14	city_posted_city_posted_sdx	city_posted	city_posted	None	None	soundex	<function city_posted_city_posted_sdx at 0x125...	from magellan.feature.simfunctions import *\nf...
15	city_posted_city_posted_exm	city_posted	city_posted	None	None	exact_match	<function city_posted_city_posted_exm at 0x125...	from magellan.feature.simfunctions import *\nf...
16	city_posted_city_posted_jac_qgm_3_qgm_3	city_posted	city_posted	qgm_3	qgm_3	jaccard	<function city_posted_city_posted_jac_qgm_3_qg...	from magellan.feature.simfunctions import *\nf...
17	km_driven_km_driven_exm	km_driven	km_driven	None	None	exact_match	<function km_driven_km_driven_exm at 0x12502f230>	from magellan.feature.simfunctions import *\nf...
18	km_driven_km_driven_anm	km_driven	km_driven	None	None	abs_norm	<function km_driven_km_driven_anm at 0x12502f2a8>	from magellan.feature.simfunctions import *\nf...
19	km_driven_km_driven_lev	km_driven	km_driven	None	None	lev	<function km_driven_km_driven_lev at 0x12502f320>	from magellan.feature.simfunctions import *\nf...
20	color_color_jac_qgm_3_qgm_3	color	color	qgm_3	qgm_3	jaccard	<function color_color_jac_qgm_3_qgm_3 at 0x125...	from magellan.feature.simfunctions import *\nf...
21	color_color_cos_dlm_dc0_dlm_dc0	color	color	dlm_dc0	dlm_dc0	cosine	<function color_color_cos_dlm_dc0_dlm_dc0 at 0...	from magellan.feature.simfunctions import *\nf...
22	color_color_jac_dlm_dc0_dlm_dc0	color	color	dlm_dc0	dlm_dc0	jaccard	<function color_color_jac_dlm_dc0_dlm_dc0 at 0...	from magellan.feature.simfunctions import *\nf...
23	color_color_mel	color	color	None	None	monge_elkan	<function color_color_mel at 0x12502f500>	from magellan.feature.simfunctions import *\nf...
24	color_color_lev	color	color	None	None	lev	<function color_color_lev at 0x12502f578>	from magellan.feature.simfunctions import *\nf...
25	color_color_nmw	color	color	None	None	needleman_wunsch	<function color_color_nmw at 0x12502f5f0>	from magellan.feature.simfunctions import *\nf...
26	color_color_sw	color	color	None	None	smith_waterman	<function color_color_sw at 0x12502f668>	from magellan.feature.simfunctions import *\nf...
27	color_color_swg	color	color	None	None	smith_waterman_gotoh	<function color_color_swg at 0x12502f6e0>	from magellan.feature.simfunctions import *\nf...
28	fuel_type_fuel_type_lev	fuel_type	fuel_type	None	None	lev	<function fuel_type_fuel_type_lev at 0x12502f758>	from magellan.feature.simfunctions import *\nf...
29	fuel_type_fuel_type_jar	fuel_type	fuel_type	None	None	jaro	<function fuel_type_fuel_type_jar at 0x12502f7d0>	from magellan.feature.simfunctions import *\nf...
30	fuel_type_fuel_type_jwn	fuel_type	fuel_type	None	None	jaro_winkler	<function fuel_type_fuel_type_jwn at 0x12502f848>	from magellan.feature.simfunctions import *\nf...
31	fuel_type_fuel_type_sdx	fuel_type	fuel_type	None	None	soundex	<function fuel_type_fuel_type_sdx at 0x12502f8c0>	from magellan.feature.simfunctions import *\nf...
32	fuel_type_fuel_type_exm	fuel_type	fuel_type	None	None	exact_match	<function fuel_type_fuel_type_exm at 0x12502f938>	from magellan.feature.simfunctions import *\nf...
33	fuel_type_fuel_type_jac_qgm_3_qgm_3	fuel_type	fuel_type	qgm_3	qgm_3	jaccard	<function fuel_type_fuel_type_jac_qgm_3_qgm_3 ...	from magellan.feature.simfunctions import *\nf...
34	price_price_exm	price	price	None	None	exact_match	<function price_price_exm at 0x12502fa28>	from magellan.feature.simfunctions import *\nf...
35	price_price_anm	price	price	None	None	abs_norm	<function price_price_anm at 0x12502faa0>	from magellan.feature.simfunctions import *\nf...
36	price_price_lev	price	price	None	None	lev	<function price_price_lev at 0x12502fb18>	from magellan.feature.simfunctions import *\nf...
37	model_year_model_year_exm	model_year	model_year	None	None	exact_match	<function model_year_model_year_exm at 0x12502...	from magellan.feature.simfunctions import *\nf...
38	model_year_model_year_anm	model_year	model_year	None	None	abs_norm	<function model_year_model_year_anm at 0x12502...	from magellan.feature.simfunctions import *\nf...
39	model_year_model_year_lev	model_year	model_year	None	None	lev	<function model_year_model_year_lev at 0x12502...	from magellan.feature.simfunctions import *\nf...
40	owner_type_owner_type_lev	owner_type	owner_type	None	None	lev	<function owner_type_owner_type_lev at 0x12502...	from magellan.feature.simfunctions import *\nf...
41	owner_type_owner_type_jar	owner_type	owner_type	None	None	jaro	<function owner_type_owner_type_jar at 0x12502...	from magellan.feature.simfunctions import *\nf...
42	owner_type_owner_type_jwn	owner_type	owner_type	None	None	jaro_winkler	<function owner_type_owner_type_jwn at 0x12502...	from magellan.feature.simfunctions import *\nf...
43	owner_type_owner_type_sdx	owner_type	owner_type	None	None	soundex	<function owner_type_owner_type_sdx at 0x12502...	from magellan.feature.simfunctions import *\nf...
44	owner_type_owner_type_exm	owner_type	owner_type	None	None	exact_match	<function owner_type_owner_type_exm at 0x12502...	from magellan.feature.simfunctions import *\nf...
45	owner_type_owner_type_jac_qgm_3_qgm_3	owner_type	owner_type	qgm_3	qgm_3	jaccard	<function owner_type_owner_type_jac_qgm_3_qgm_...	from magellan.feature.simfunctions import *\nf...
46	url_url_lev	url	url	None	None	lev	<function url_url_lev at 0x12503c050>	from magellan.feature.simfunctions import *\nf...
47	url_url_jar	url	url	None	None	jaro	<function url_url_jar at 0x12503c0c8>	from magellan.feature.simfunctions import *\nf...
48	url_url_jwn	url	url	None	None	jaro_winkler	<function url_url_jwn at 0x12503c140>	from magellan.feature.simfunctions import *\nf...
49	url_url_sdx	url	url	None	None	soundex	<function url_url_sdx at 0x12503c1b8>	from magellan.feature.simfunctions import *\nf...
50	url_url_exm	url	url	None	None	exact_match	<function url_url_exm at 0x12503c230>	from magellan.feature.simfunctions import *\nf...
51	url_url_jac_qgm_3_qgm_3	url	url	qgm_3	qgm_3	jaccard	<function url_url_jac_qgm_3_qgm_3 at 0x12503c2a8>	from magellan.feature.simfunctions import *\nf...

	_id	ltable.id	rtable.id	bike_name_bike_name_jac_qgm_3_qgm_3	bike_name_bike_name_cos_dlm_dc0_dlm_dc0	bike_name_bike_name_jac_dlm_dc0_dlm_dc0	bike_name_bike_name_mel	bike_name_bike_name_lev	bike_name_bike_name_nmw	bike_name_bike_name_sw	bike_name_bike_name_swg	color_color_jac_qgm_3_qgm_3	color_color_cos_dlm_dc0_dlm_dc0	color_color_jac_dlm_dc0_dlm_dc0	color_color_mel	color_color_lev	color_color_nmw	color_color_sw	color_color_swg	gold
0	0	42500	1873	0.615385	0.800000	0.666667	0.900000	0.769231	0.769231	0.850000	0.900000	0	0	0	0.4	0	0.5	0.333333	0.4	0
1	1	44841	17451	0.378378	0.408248	0.250000	0.817391	0.562500	0.640625	0.739130	0.756522	1	1	1	1.0	1	1.0	1.000000	1.0	1
2	2	43816	7915	0.212121	0.447214	0.285714	0.473684	0.480000	0.620000	0.368421	0.505263	1	1	1	1.0	1	1.0	1.000000	1.0	0
3	3	47734	17347	0.243243	0.471405	0.285714	0.760000	0.400000	0.542857	0.733333	0.760000	1	1	1	1.0	1	1.0	1.000000	1.0	0
4	4	43387	8359	0.761905	0.894427	0.800000	1.000000	0.692308	0.692308	1.000000	1.000000	1	1	1	1.0	1	1.0	1.000000	1.0	1

	Name	Matcher	Num folds	Fold 1	Fold 2	Fold 3	Fold 4	Fold 5	Mean score
0	DecisionTree	<magellan.matcher.dtmatcher.DTMatcher object a...	5	0.960000	0.974359	0.962963	0.930233	0.937500	0.953011
1	RF	<magellan.matcher.rfmatcher.RFMatcher object a...	5	0.916667	1.000000	0.962963	0.926829	0.969697	0.955231
2	SVM_61402071609014438441	<magellan.matcher.svmmatcher.SVMMatcher object...	5	0.842105	0.844444	0.800000	0.785714	0.809524	0.816358
3	NB	<magellan.matcher.nbmatcher.NBMatcher object a...	5	0.842105	0.844444	0.800000	0.785714	0.809524	0.816358
4	LogReg	<magellan.matcher.logregmatcher.LogRegMatcher ...	5	0.857143	0.883721	0.823529	0.785714	0.850000	0.840021
5	LinReg	<magellan.matcher.linregmatcher.LinRegMatcher ...	5	0.901961	0.904762	0.848485	0.880000	0.864865	0.880014

	Name	Matcher	Num folds	Fold 1	Fold 2	Fold 3	Fold 4	Fold 5	Mean score
0	DecisionTree	<magellan.matcher.dtmatcher.DTMatcher object a...	5	0.914286	1.000000	1.000000	0.938776	0.933333	0.957279
1	RF	<magellan.matcher.rfmatcher.RFMatcher object a...	5	0.938776	1.000000	0.857143	0.875000	1.000000	0.934184
2	SVM_61402071609014438441	<magellan.matcher.svmmatcher.SVMMatcher object...	5	0.790698	0.865672	0.808511	0.750000	0.842105	0.811397
3	NB	<magellan.matcher.nbmatcher.NBMatcher object a...	5	0.821429	0.785714	0.810811	0.880000	0.777778	0.815146
4	LogReg	<magellan.matcher.logregmatcher.LogRegMatcher ...	5	0.792453	0.875000	0.850000	0.900000	0.844444	0.852379
5	LinReg	<magellan.matcher.linregmatcher.LinRegMatcher ...	5	0.830189	0.862745	0.833333	0.871795	0.736842	0.826981

	Metric	Num folds	Fold 1	Fold 2	Fold 3	Fold 4	Fold 5	Mean score
0	precision	5	1	1	1.000000	1	1	1.000000
1	recall	5	1	1	0.950000	1	1	0.990000
2	f1	5	1	1	0.974359	1	1	0.994872

	Metric	Num folds	Fold 1	Fold 2	Fold 3	Fold 4	Fold 5	Mean score
0	precision	5	0.944444	0.909091	1.000000	1.000000	0.9375	0.958207
1	recall	5	0.944444	0.952381	0.952381	0.950000	0.9375	0.947341
2	f1	5	0.944444	0.930233	0.975610	0.974359	0.9375	0.952429