-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathperformance_metrics.py
297 lines (247 loc) · 12 KB
/
performance_metrics.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
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
from collections import defaultdict
import matplotlib
import scipy.spatial.distance
import numpy as np
import pandas as pd
from datetime import datetime
import matplotlib.pyplot as plt
from plotly.offline import iplot
import plotly
import plotly.graph_objs as go
from tensorflow.python.keras.models import load_model
import sklearn.manifold
import sklearn.cluster
from data_pipeline import ExcelReader
from helper_functions import create_html, get_filepath
class evaluatePerformance:
def __init__(self, model, metrics={}, product_file_dir=None, product_file_loc=None, product_used_loc=None):
self.model_loc = model
self.metrics = metrics
try:
self.product_info = self.read_product_data_simulated(product_file_dir, product_file_loc)
self.products_used = set(np.loadtxt(product_used_loc, delimiter=",", dtype=np.int32))
except FileNotFoundError and TypeError:
print("{} ERROR: Unable to load the product_info or/and product_used files".format(
datetime.now().strftime("%H:%M:%S")))
@staticmethod
def read_product_data_simulated(file_dir, file_location):
data_reader = ExcelReader(4)
simulated_data = data_reader.read_split_excel(file_location, file_dir)
return simulated_data
def read_product_data_instacart(self):
raise NotImplementedError
def get_input_weights(self, type_weights='center_embedding'):
model = load_model("Models/" + self.model_loc)
embedding_layer = model.get_layer(type_weights)
weights = np.array(embedding_layer.get_weights()[0])
weights /= np.linalg.norm(weights, axis=1)[:, np.newaxis]
sorted_products = sorted(self.products_used)
p2v_vectors = []
for product in sorted_products:
p2v_vectors.append(weights[product])
return np.array(p2v_vectors), sorted_products
def simulate_random_weights(self, mean, std, num_weights, size_vector):
random_weights = np.random.normal(loc=mean, scale=std, size=(num_weights, size_vector))
return random_weights
@staticmethod
def get_tsne_embedding(weights, official_run=False):
if not official_run:
tsne_output = sklearn.manifold.TSNE(random_state=1,
n_components=2,
n_iter=4000,
perplexity=15,
angle=.5,
verbose=1).fit_transform(weights)
return tsne_output
else:
best_embedding = None
for i in range(50):
if i % 10 == 0:
print('10 tsne mappings done')
best_divergence = float('inf')
tsne_model = sklearn.manifold.TSNE(n_components=2,
n_iter=4000,
perplexity=15,
angle=.5,
verbose=0)
tsne_output = tsne_model.fit_transform(weights)
divergence = tsne_model.kl_divergence_
if best_divergence > divergence:
best_embedding = tsne_output
return best_embedding
def format_data(self, tsne_output):
tsne_data = []
for product in self.products_used:
x = tsne_output[product][0]
y = tsne_output[product][1]
rows = self.product_info.loc[self.product_info['j'] == product].head(1)
category = rows['category'].values[0]
tsne_data.append(([x, y, category, product]))
return pd.DataFrame(tsne_data, columns=['x', 'y', 'c', 'j'])
def create_scatterplot(self, tsne_df, plot_file_name):
# Function taken from: https://github.com/sbstn-gbl to enable comparison of the plots made
plot_data = [go.Scatter(x=tsne_df['x'].values,
y=tsne_df['y'].values,
text=[
'category = %d <br> product = %d' % (x, y)
for (x, y) in zip(tsne_df['c'].values, tsne_df['j'].values)
],
hoverinfo='text',
mode='markers',
marker=dict(
size=14,
color=tsne_df['c'].values,
colorscale='Jet',
showscale=False
)
)
]
legend = go.layout.Legend()
plot_layout = go.Layout(
width=800,
height=600,
margin=go.layout.Margin(l=0, r=0, b=0, t=0, pad=4),
hovermode='closest',
legend=legend,
showlegend=True
)
# plot
fig = go.Figure(data=plot_data, layout=plot_layout)
iplot(fig)
plt.show()
def category_level_similarity(self, num_product_in_category, vectors):
category_to_assign = 0
category_product = defaultdict(list)
p2v_vectors = self.get_input_weights()
acc = 0
category = 0
average_vectors = {}
summed_vec = np.zeros(vectors.shape[1])
for product in self.products_used:
summed_vec += vectors[product]
if acc == num_product_in_category - 1:
acc = 0
average_vectors[category] = summed_vec / num_product_in_category
category += 1
summed_vec = np.zeros(vectors.shape[1])
else:
acc += 1
return average_vectors
@staticmethod
def similarity_matrix(average_vectors, filename):
num_cat = len(average_vectors.keys())
similarity = np.zeros((num_cat, num_cat))
for i in range(num_cat):
for j in range(num_cat):
similarity[i][j] = average_vectors[i] @ average_vectors[j]
create_html(similarity, filename)
f = plt.figure(figsize=(19, 15))
plt.matshow(similarity, fignum=f.number)
plt.xticks(range(similarity.shape[1]), fontsize=14, rotation=90)
plt.yticks(range(similarity.shape[1]), fontsize=14)
cb = plt.colorbar()
cb.set_clim(-1, 1)
cb.ax.tick_params(labelsize=14)
plt.show()
def co_occurence_matrix(self, average_vectors_co, average_vectors_ce, filename):
num_cat = len(average_vectors_ce.keys())
similarity = np.zeros((num_cat, num_cat))
for i in range(num_cat):
for j in range(i, num_cat):
similarity[i][j] = average_vectors_ce[j] @ average_vectors_co[i]
similarity[j][i] = average_vectors_ce[j] @ average_vectors_co[i]
create_html(similarity, filename)
f = plt.figure(figsize=(19, 15))
plt.matshow(similarity, fignum=f.number)
plt.xticks(range(similarity.shape[1]), fontsize=14, rotation=90)
plt.yticks(range(similarity.shape[1]), fontsize=14)
cb = plt.colorbar()
cb.ax.tick_params(labelsize=14)
plt.show()
def bench_marks(self, data):
x_y_df = data.set_index(['j', 'c'])
x_y_array = x_y_df.to_numpy()
true_clusters = data['c'].to_numpy()
clustering = sklearn.cluster.KMeans(n_clusters=len(np.unique(true_clusters)), n_init=30)
kmean_pred = clustering.fit_predict(x_y_df.values)
s_score = sklearn.metrics.silhouette_score(X=x_y_array, labels=true_clusters)
adjusted_mis = sklearn.metrics.adjusted_mutual_info_score(
labels_true=true_clusters,
labels_pred=kmean_pred,
average_method='arithmetic'
)
nn_score = self.get_hitrate(data, 14)
return s_score, adjusted_mis, nn_score
def get_hitrate(self, x, num_nn):
xy = x[['x', 'y']].to_numpy()
j = x['j'].to_numpy()
c = x['c'].to_numpy()
distances = scipy.spatial.distance.cdist(xy, xy)
distance_df = pd.DataFrame({
'j': np.repeat(j, len(j)),
'c': np.repeat(c, len(c)),
'j2': np.tile(j, len(j)),
'c2': np.tile(c, len(c)),
'd': distances.flatten()
})
distance_df = distance_df[distance_df['j'] != distance_df['j2']]
distance_df = distance_df.sort_values('d')
distance_df['rank_d'] = distance_df.groupby('j').cumcount()
nn = distance_df[distance_df['rank_d'] < num_nn]
score = float(sum(nn['c'] == nn['c2'])) / nn.shape[0]
return score
def create_histogram(self):
correlation = self.product_info.corr()
self.product_info.hist(column='j')
plt.show()
def main():
model_used = 'simulated_test.h5'
data_dir = 'large_data'
data_file = 'simulated_data'
product_used = r'large_data\center_products_simulated'
performance_logger = evaluatePerformance(model=model_used, product_file_dir=data_dir,
product_file_loc=data_file,
product_used_loc=product_used)
weights, sorted_products = performance_logger.get_input_weights()
#co_occurence_weights, sorted_products = performance_logger.get_input_weights(type_weights='context_embedding')
random_weights = performance_logger.simulate_random_weights(0, .2, weights.shape[0], weights.shape[1])
print("Starting on center weights")
tsne_ce = performance_logger.get_tsne_embedding(weights, official_run=True)
dank_df = performance_logger.format_data(tsne_ce)
s_score, info_score, nn_score = performance_logger.bench_marks(dank_df)
av_vec_ce = performance_logger.category_level_similarity(15, weights)
performance_logger.similarity_matrix(av_vec_ce, 'similarity_correlation.html')
print("Starting on random weights")
tsne_r = performance_logger.get_tsne_embedding(random_weights)
dank_df_r = performance_logger.format_data(tsne_r)
s_score_r, info_score_r, nn_score_r = performance_logger.bench_marks(dank_df_r)
av_vec = performance_logger.category_level_similarity(15, random_weights)
performance_logger.similarity_matrix(av_vec, 'random_weights.html')
"""
print("Starting on context embeddings")
tsne_co = performance_logger.get_tsne_embedding(co_occurence_weights, official_run=True)
dank_df_c = performance_logger.format_data(tsne_co)
s_score_c, info_score_c, nn_score_c = performance_logger.bench_marks(dank_df_c)
av_vec_co = performance_logger.category_level_similarity(15, co_occurence_weights)
performance_logger.co_occurence_matrix(av_vec_co, av_vec_ce, 'co-occurence_correlation.html')"""
print("Metrics for trained weights, s_score: {}, adj_info: {}, nn_score: {}".format(s_score, info_score,
nn_score))
print("Metrics for random weights, s_score: {}, adj_info: {}, nn_score: {}".format(s_score_r, info_score_r,
nn_score_r))
#print("Metrics for contex weights, s_score: {}, adj_info: {}, nn_score: {}".format(s_score_c, info_score_c, nn_score_c))
performance_logger.create_scatterplot(dank_df, 'test')
#performance_logger.create_scatterplot(dank_df_c, 'test')
def create_correlation_matrix():
PATH = get_filepath('resources', 'correlation_matrix_excel.xlsx')
data = pd.read_excel(PATH, header=None)
correlation_matrix = data.to_numpy()
f = plt.figure(figsize=(19, 15))
plt.matshow(correlation_matrix, fignum=f.number)
plt.xticks(range(correlation_matrix.shape[1]), fontsize=14, rotation=90)
plt.yticks(range(correlation_matrix.shape[1]), fontsize=14)
cb = plt.colorbar()
cb.ax.tick_params(labelsize=14)
plt.show()
create_correlation_matrix()
if __name__ == '__main__':
main()