-
Notifications
You must be signed in to change notification settings - Fork 6
/
Copy pathEvalRSReclist.py
192 lines (159 loc) · 8.4 KB
/
EvalRSReclist.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
from reclist.logs import LOGGER
from reclist.metadata import METADATA_STORE
import pandas as pd
import numpy as np
import os
from reclist.reclist import rec_test
from reclist.reclist import RecList, CHART_TYPE
from random import choice
TOP_K_CHALLENGE = 100
class EvalRSReclist(RecList):
def __init__(
self,
dataset,
predictions,
model_name,
logger: LOGGER,
metadata_store: METADATA_STORE,
**kwargs
):
super().__init__(
model_name,
logger,
metadata_store,
**kwargs
)
self.dataset = dataset
self._x_train = dataset._get_train_set(fold=0)
subset_of_cols_to_return = ['user_id', 'track_id']
self._x_test = dataset._get_test_set(fold=0, subset_of_cols_to_return=subset_of_cols_to_return)[['user_id']]
self._y_test = dataset._get_test_set(fold=0, subset_of_cols_to_return=subset_of_cols_to_return).set_index('user_id')
self._y_preds = predictions
self.similarity_model = kwargs.get("similarity_model", None)
return
def mrr_at_k_slice(self,
y_preds: pd.DataFrame,
y_test: pd.DataFrame,
slice_info: pd.DataFrame,
slice_key: str):
from reclist.metrics.standard_metrics import rr_at_k
# get rr (reciprocal rank) for each prediction made
rr = rr_at_k(y_preds, y_test, k=TOP_K_CHALLENGE)
# convert to DataFrame
rr = pd.DataFrame(rr, columns=['rr'], index=y_test.index)
# grab slice info
rr[slice_key] = slice_info[slice_key].values
# group-by slice and get per-slice mrr
return rr.groupby(slice_key)['rr'].agg('mean').to_json()
def miss_rate_at_k_slice(self,
y_preds: pd.DataFrame,
y_test: pd.DataFrame,
slice_info: pd.DataFrame,
slice_key: str):
from reclist.metrics.standard_metrics import misses_at_k
# get false positives
m = misses_at_k(y_preds, y_test, k=TOP_K_CHALLENGE).min(axis=2)
# convert to dataframe
m = pd.DataFrame(m, columns=['mr'], index=y_test.index)
# grab slice info
m[slice_key] = slice_info[slice_key].values
# group-by slice and get per-slice mrr
return m.groupby(slice_key)['mr'].agg('mean')
def miss_rate_equality_difference(self,
y_preds: pd.DataFrame,
y_test: pd.DataFrame,
slice_info: pd.DataFrame,
slice_key: str):
from reclist.metrics.standard_metrics import misses_at_k
mr_per_slice = self.miss_rate_at_k_slice(y_preds, y_test, slice_info, slice_key)
mr = misses_at_k(y_preds, y_test, k=TOP_K_CHALLENGE).min(axis=2).mean()
# take negation so that higher values => better fairness
mred = -(mr_per_slice-mr).abs().mean()
res = mr_per_slice.to_dict()
return {'mred': mred, 'mr': mr, **res}
def cosine_sim(self, u: np.array, v: np.array) -> np.array:
return np.sum(u * v, axis=-1) / (np.linalg.norm(u, axis=-1) * np.linalg.norm(v, axis=-1))
@rec_test(test_type='stats')
def stats(self):
tracks_per_users = (self._y_test.values!=-1).sum(axis=1)
return {
'num_users': len(self._x_test['user_id'].unique()),
'max_items': int(tracks_per_users.max()),
'min_items': int(tracks_per_users.min())
}
@rec_test(test_type='HIT_RATE')
def hit_rate_at_100(self):
from reclist.metrics.standard_metrics import hit_rate_at_k
hr = hit_rate_at_k(self._y_preds, self._y_test, k=TOP_K_CHALLENGE)
return hr
@rec_test(test_type='MRR')
def mrr_at_100(self):
from reclist.metrics.standard_metrics import mrr_at_k
return mrr_at_k(self._y_preds, self._y_test, k=TOP_K_CHALLENGE)
@rec_test(test_type='MRED_COUNTRY', display_type=CHART_TYPE.BARS)
def mred_country(self):
country_list = ["US", "RU", "DE", "UK", "PL", "BR", "FI", "NL", "ES", "SE", "UA", "CA", "FR", "NaN"]
user_countries = self.dataset.df_users.loc[self._y_test.index, ['country']].fillna('NaN')
valid_country_mask = user_countries['country'].isin(country_list)
y_pred_valid = self._y_preds[valid_country_mask]
y_test_valid = self._y_test[valid_country_mask]
user_countries = user_countries[valid_country_mask]
return self.miss_rate_equality_difference(y_pred_valid, y_test_valid, user_countries, 'country')
@rec_test(test_type='MRED_USER_ACTIVITY', display_type=CHART_TYPE.BARS)
def mred_user_activity(self):
bins = np.array([1, 100, 1000])
user_activity = self._x_train[self._x_train['user_id'].isin(self._y_test.index)]
user_activity = user_activity.groupby('user_id',as_index=True, sort=False)[['user_track_count']].sum()
user_activity = user_activity.loc[self._y_test.index]
user_activity['bin_index'] = np.digitize(user_activity.values.reshape(-1), bins)
user_activity['bins'] = bins[user_activity['bin_index'].values-1]
return self.miss_rate_equality_difference(self._y_preds, self._y_test, user_activity, 'bins')
@rec_test(test_type='MRED_TRACK_POPULARITY', display_type=CHART_TYPE.BARS)
def mred_track_popularity(self):
bins = np.array([1, 10, 100, 1000])
track_id = self._y_test['track_id']
track_activity = self._x_train[self._x_train['track_id'].isin(track_id)]
track_activity = track_activity.groupby('track_id', as_index=True, sort=False)[['user_track_count']].sum()
track_activity = track_activity.loc[track_id]
track_activity['bin_index'] = np.digitize(track_activity.values.reshape(-1), bins)
track_activity['bins'] = bins[track_activity['bin_index'].values - 1]
return self.miss_rate_equality_difference(self._y_preds, self._y_test, track_activity, 'bins')
@rec_test(test_type='MRED_ARTIST_POPULARITY', display_type=CHART_TYPE.BARS)
def mred_artist_popularity(self):
bins = np.array([1, 100, 1000, 10000])
artist_id = self.dataset.df_tracks.loc[self._y_test['track_id'], 'artist_id']
artist_activity = self._x_train[self._x_train['artist_id'].isin(artist_id)]
artist_activity = artist_activity.groupby('artist_id', as_index=True, sort=False)[['user_track_count']].sum()
artist_activity = artist_activity.loc[artist_id]
artist_activity['bin_index'] = np.digitize(artist_activity.values.reshape(-1), bins)
artist_activity['bins'] = bins[artist_activity['bin_index'].values - 1]
return self.miss_rate_equality_difference(self._y_preds, self._y_test, artist_activity, 'bins')
@rec_test('MRED_GENDER', display_type=CHART_TYPE.BARS)
def mred_gender(self):
user_gender = self.dataset.df_users.loc[self._y_test.index, ['gender']]
return self.miss_rate_equality_difference(self._y_preds, self._y_test, user_gender, 'gender')
@rec_test(test_type='BEING_LESS_WRONG')
def being_less_wrong(self):
from reclist.metrics.standard_metrics import hits_at_k
hits = hits_at_k(self._y_preds, self._y_test, k=TOP_K_CHALLENGE).max(axis=2)
misses = (hits == False)
miss_gt_vectors = self.similarity_model[self._y_test.loc[misses, 'track_id'].values.reshape(-1)]
# we calculate the score w.r.t to the first prediction
miss_pred_vectors = self.similarity_model[self._y_preds.loc[misses, '0'].values.reshape(-1)]
return float(self.cosine_sim(miss_gt_vectors, miss_pred_vectors).mean())
class EvalRSSimpleModel:
"""
This is a dummy model that returns random predictions on the EvalRS dataset.
"""
def __init__(self, items: pd.DataFrame, top_k: int=10, **kwargs):
self.items = items
self.top_k = top_k
print("Received additional arguments: {}".format(kwargs))
return
def predict(self, user_ids: pd.DataFrame) -> pd.DataFrame:
k = self.top_k
num_users = len(user_ids)
pred = self.items.sample(n=k*num_users, replace=True).index.values
pred = pred.reshape(num_users, k)
pred = np.concatenate((user_ids[['user_id']].values, pred), axis=1)
return pd.DataFrame(pred, columns=['user_id', *[str(i) for i in range(k)]]).set_index('user_id')