'Perfect Match': 0} for event in set(event_dict.values()): test_rows = np.argwhere(full_event == event) train_rows = np.argwhere(full_event != event) text_test = [full_text[n[0]] for n in test_rows] cat_test = [full_cat[n[0]] for n in test_rows] text_train = [full_text[n[0]] for n in train_rows] cat_train = [full_cat[n[0]] for n in train_rows] cat_test_arr = np.array(cat_test, dtype=np.float64) clas = Classify(text_train, cat_train, 2000, model = LinearSVC(C = c, class_weight = weight, loss = loss, random_state=1)) predict = clas.predict(text_test) simp = clas.simple_evaluation(cat_test, predict) for key in simp: if key in model_res: model_res[key] += simp[key] stats = clas.stats_calc(model_res['True Positive'], model_res['True Negative'], model_res['False Positive'], model_res['False Negative'], model_res['One Label'], model_res['Perfect Match'])
for event in set(event_dict.values()):
test_rows = np.argwhere(full_event == event)
train_rows = np.argwhere(full_event != event)
text_test = [full_text[n[0]] for n in test_rows]
cat_test = [full_cat[n[0]] for n in test_rows]
text_train = [full_text[n[0]] for n in train_rows]
cat_train = [full_cat[n[0]] for n in train_rows]
cat_test_arr = np.array(cat_test, dtype=np.float64)
clas = Classify(text_train,
cat_train,
2000,
model=RandomForestClassifier(random_state=1,
n_jobs=4,
n_estimators=estimators,
class_weight=weight,
criterion=criterion,
bootstrap=bootstrap))
predict = clas.predict(text_test)
simp = clas.evaluate(cat_test, predict)
for key in simp:
if key in model_res:
model_res[key] += simp[key]
stats = clas.stats_calc(model_res['True Positive'],
model_res['True Negative'],
model_res['False Positive'],
full_text = list()
full_cat = list()
for id in (sorted(text_dict.keys())):
full_text.append(text_dict[id])
full_cat.append(cat_dict[id])
text_train, text_test, cat_train, cat_test = train_test_split(full_text,
full_cat,
test_size=.1,
random_state=1)
cat_test_arr = np.array(cat_test, dtype=np.float64)
if pretrained:
clas = Classify(pretrained='pretrained/')
elif classifier == 'rf':
clas = Classify(text_train,
cat_train,
model=RandomForestClassifier(class_weight='balanced',
n_estimators=100))
elif classifier == 'svc':
clas = Classify(text_train, cat_train, model=SVC(class_weight='balanced'))
elif classifier == 'linearsvc':
clas = Classify(text_train,
cat_train,
model=LinearSVC(class_weight='balanced'))
else:
clas = Classify(text_train, cat_train)
predict = clas.predict(text_test)
evals = clas.evaluation_(
}
for event in set(event_dict.values()):
test_rows = np.argwhere(full_event == event)
train_rows = np.argwhere(full_event != event)
text_test = [full_text[n[0]] for n in test_rows]
cat_test = [full_cat[n[0]] for n in test_rows]
text_train = [full_text[n[0]] for n in train_rows]
cat_train = [full_cat[n[0]] for n in train_rows]
cat_test_arr = np.array(cat_test, dtype=np.float64)
clas = Classify(text_train,
cat_train,
1000,
model=KNeighborsClassifier(n_neighbors=k,
weights=weight,
n_jobs=2))
predict = clas.predict(text_test)
simp = clas.simple_evaluation(cat_test, predict)
for key in simp:
if key in model_res:
model_res[key] += simp[key]
stats = clas.stats_calc(model_res['True Positive'],
model_res['True Negative'],
model_res['False Positive'],
model_res['False Negative'],
model_res['One Label'], model_res['Perfect Match'])
'Perfect Match': 0
}
for event in set(event_dict.values()):
test_rows = np.argwhere(full_event == event)
train_rows = np.argwhere(full_event != event)
text_test = [full_text[n[0]] for n in test_rows]
cat_test = [full_cat[n[0]] for n in test_rows]
text_train = [full_text[n[0]] for n in train_rows]
cat_train = [full_cat[n[0]] for n in train_rows]
cat_test_arr = np.array(cat_test, dtype=np.float64)
clas = Classify(text_train,
cat_train,
2000,
model=BernoulliNB(alpha=alpha))
predict = clas.predict(text_test)
simp = clas.evaluate(cat_test, predict)
for key in simp:
if key in model_res:
model_res[key] += simp[key]
stats = clas.stats_calc(model_res['True Positive'],
model_res['True Negative'],
model_res['False Positive'],
model_res['False Negative'],
model_res['One Label'], model_res['Perfect Match'])
'Perfect Match': 0
}
for event in set(event_dict.values()):
test_rows = np.argwhere(full_event == event)
train_rows = np.argwhere(full_event != event)
text_test = [full_text[n[0]] for n in test_rows]
cat_test = [full_cat[n[0]] for n in test_rows]
text_train = [full_text[n[0]] for n in train_rows]
cat_train = [full_cat[n[0]] for n in train_rows]
cat_test_arr = np.array(cat_test, dtype=np.float64)
clas = Classify(text_train,
cat_train,
2000,
model=LogisticRegression(C=c, class_weight=weight))
predict = clas.predict(text_test)
simp = clas.evaluate(cat_test, predict)
for key in simp:
if key in model_res:
model_res[key] += simp[key]
stats = clas.stats_calc(model_res['True Positive'],
model_res['True Negative'],
model_res['False Positive'],
model_res['False Negative'],
model_res['One Label'], model_res['Perfect Match'])
full_cat_confusion = dict()
for event in set(event_dict.values()):
print("Event: ", event)
test_rows = np.argwhere(full_event == event)
train_rows = np.argwhere(full_event != event)
text_test = [full_text[n[0]] for n in test_rows]
cat_test = [full_cat[n[0]] for n in test_rows]
text_train = [full_text[n[0]] for n in train_rows]
cat_train = [full_cat[n[0]] for n in train_rows]
cat_test_arr = np.array(cat_test, dtype=np.float64)
if pretrained:
clas = Classify(pretrained='pretrained/')
elif classifier == 'rf':
clas = Classify(text_train, cat_train, 2000,
model=RandomForestClassifier(class_weight='balanced', n_estimators=1))
elif classifier == 'lsvc':
clas = Classify(text_train, cat_train, 2000,
model=LinearSVC(C=0.01,
class_weight='balanced',
loss='hinge',
random_state=1))
elif classifier == 'log':
clas = Classify(text_train, cat_train, 2000,
model=LogisticRegression(class_weight='balanced', C=0.01))
else:
clas = Classify(text_train, cat_train, 2000)
predict = clas.predict(text_test)
for event in set(event_dict.values()):
print("\n\n")
print("Event: ", event)
test_rows = np.argwhere(full_event == event)
train_rows = np.argwhere(full_event != event)
text_test = [full_text[n[0]] for n in test_rows]
cat_test = [full_cat[n[0]] for n in test_rows]
text_train = [full_text[n[0]] for n in train_rows]
cat_train = [full_cat[n[0]] for n in train_rows]
cat_test_arr = np.array(cat_test, dtype=np.float64)
if pretrained:
clas = Classify(pretrained='pretrained/')
elif classifier == 'rf':
clas = Classify(text_train, cat_train, 2000,
model=RandomForestClassifier(class_weight='balanced',
n_estimators=100))
elif classifier == 'svc':
clas = Classify(text_train, cat_train, 2000,
model=SVC(class_weight='balanced'))
elif classifier == 'linearsvc':
clas = Classify(text_train, cat_train, 2000,
model=LinearSVC(class_weight='balanced'))
elif classifier == 'log':
clas = Classify(text_train, cat_train, 2000,
model=LogisticRegression(class_weight='balanced'))
else:
clas = Classify(text_train, cat_train, 2000)