emotion_analysis/run.py at master · Garfounkel/emotion_analysis · GitHub

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
#!/usr/bin/env python

import pandas as pd
import numpy as np
from keras.utils import to_categorical
from keras.callbacks import ModelCheckpoint, EarlyStopping
from keras.models import load_model
from sklearn.utils import class_weight
from sklearn.linear_model import LogisticRegression
from collections import defaultdict
import matplotlib.pyplot as plt
import pickle

from src.embeddings import get_embedding_dictionnary, get_embeddings_and_word_index, sequences_to_index
from src.dataset import load_dataset, load_datasets_and_vocab_pipeline, train_test_val_split
from src.evaluate import get_metrics, get_predictions, compare_metrics, get_metrics_binary
from src.models import model_mine, model_conv1d, EnsembleModel
from src.submission import generate_predictions
from src.model_saverloader import *
from src.utils import get_wrongs, emotion2label, label2emotion, reindex_sequences

embeddings_path = './data/embeddings/datastories.twitter.300d.txt'

os.makedirs('pickles/models/', exist_ok=True)


# ### Load and preprocess Train and Test
(X_train_txt, y_train), (X_test_txt, y_test), (vocab, max_seq_len) = load_datasets_and_vocab_pipeline()


# ### Compute the classes weights because our dataset is largely unbalanced
cls_weights = class_weight.compute_class_weight('balanced', np.unique(y_train), y_train)


# ### Create an embedding matrix for our vocabulary
max_seq_len = 80
emb_matrix, word_index, unknown_words = get_embeddings_and_word_index(embeddings_path, max_seq_len, vocab)
max_seq_len = word_index['<max_seq_len>']
# Unknown words: 1300

print(f'Unknown words from the vocabulary: {len(unknown_words)} (previously 1300 without spellcheck fixes)')
print(f'{len(unknown_words) / len(vocab) * 100:.2f}% of our vocabulary are unknown words to the embedding matrix')


# ### Transforming our sequences of words to indexes
X_train = sequences_to_index(X_train_txt, word_index)
X_test = sequences_to_index(X_test_txt, word_index)

y_train = to_categorical(y_train, 4) if len(y_train.shape) == 1 else y_train
y_test = to_categorical(y_test, 4) if len(y_test.shape) == 1 else y_test
(x_trn, y_trn), (x_val, y_val) = train_test_val_split(X_train, y_train)


print('training set: ' + str(len(x_trn)) + ' samples')
print('validation set: ' + str(len(x_val)) + ' samples')

print('x_train:', x_trn.shape)
print('y_train:', y_trn.shape)

binarize = np.vectorize(lambda x: 0 if x in [0, 1, 2] else 1)
y_trn_2 = to_categorical(binarize(y_trn.argmax(axis=1)), 2)
y_val_2 = to_categorical(binarize(y_val.argmax(axis=1)), 2)


# ## Training categorical
model = model_mine(emb_matrix, max_seq_len)

callbacks_list = [
        ModelCheckpoint(filepath='pickles/models/best.h5', save_best_only=True, verbose=1),
        EarlyStopping(patience=4, verbose=0)
    ]


history = model.fit(x_trn, y_trn, batch_size=128, validation_data=(x_val, y_val), epochs=15, class_weight=cls_weights, callbacks=callbacks_list)


# ### Evaluating our categorical model
model = load_model('pickles/models/best.h5')
best_metrics = load_best_metrics(mode='categorical')
y_pred_test, proba_preds = get_predictions(model, X_test)
model_metrics = compare_metrics(proba_preds, y_test, best_metrics, mode='categorical')
save_model_full(model, emb_matrix, word_index, model_metrics)


# ## Training Others vs all (binary)
model_bin = model_conv1d(emb_matrix, max_seq_len, class_number=2)

callbacks_list_bin = [
        ModelCheckpoint(filepath='pickles/models/best_bin.h5', save_best_only=True, verbose=1),
        EarlyStopping(patience=4, verbose=0)
    ]


cls_weight_bin = np.array([0.5, 1.5])
cls_weight_bin

history_bin = model_bin.fit(x_trn, y_trn_2, batch_size=128, validation_data=(x_val, y_val_2), epochs=20, class_weight=cls_weight_bin, callbacks=callbacks_list_bin)


# ### Evaluate our binary model
model_bin = load_model('pickles/models/best_bin.h5')
best_bin_metrics = load_best_metrics(mode='binary')
preds_2_tst = model_bin.predict(X_test, batch_size=128)
model_bin_metrics = compare_metrics(preds_2_tst, y_test, best_bin_metrics, mode='binary')
save_model_full(model_bin, emb_matrix, word_index, model_bin_metrics)


# ## Ensemble binary and categorical
model_cat, _, word_index_cat, model_metrics_cat = load_best_model(mode='categorical')
model_bin, _, word_index_bin, model_metrics_bin = load_best_model(mode='binary')
model_ens = EnsembleModel(*load_best_model(mode='ensemble'))

model_1 = model_cat
model_2 = model_bin


# #### Test combi
preds_tst_1 = model_metrics_cat['y_proba']
preds_tst_2 = model_bin_metrics['y_proba']

combi_preds_tst = np.hstack([preds_tst_1, preds_tst_2])


# #### Train combi
reindex_sequences(X_train, word_index, word_index_cat)
trn_preds_1 = model_1.predict(X_train, batch_size=128)
reindex_sequences(X_train, word_index_cat, word_index_bin)
trn_preds_2 = model_2.predict(X_train, batch_size=128)
combi_preds_trn = np.hstack([trn_preds_1, trn_preds_2])

lreg = LogisticRegression(solver='lbfgs', multi_class='auto', max_iter=1000).fit(combi_preds_trn, y_train.argmax(axis=1))


# ### Evaluate our ensembling model
lreg_pred_tst = lreg.predict_proba(combi_preds_tst)
best_metrics_combi = load_best_metrics(mode='ensemble')
model_combi_metrics = compare_metrics(lreg_pred_tst, y_test, best_metrics_combi, mode='ensemble')
save_ensemble_model(lreg, model_combi_metrics, [model_metrics_cat, model_metrics_bin])


# ## Generating a submission file
model = EnsembleModel(*load_best_model(mode='ensemble'))
y_pred_sub, proba_preds = generate_predictions(model, 'data/test.txt', word_index=word_index)