def crossValidation(folds, classifiers, inf, sup):
acc = [[] for c in classifiers]
print(acc)
for i in range(len(folds)):
print('fold: ', i)
aux = [df for j, df in enumerate(folds) if j != i]
tr = pd.concat(aux, ignore_index=True)
ts = folds[i]
v = TfidfVectorizer(lowercase=True,
min_df=inf,
max_df=sup,
tokenizer=token_stem)
v.encoding = 'ISO-8859-14'
m = v.fit_transform(tr['Text'].values)
train_df = pd.DataFrame(data=m.toarray(),
columns=v.get_feature_names())
train_df = pd.concat([train_df, tr['Class'].reset_index(drop=True)],
axis=1)
test_df = pd.DataFrame(columns=v.get_feature_names(),
data=v.transform(ts['Text'].values).toarray())
test_df = pd.concat([test_df, ts['Class'].reset_index(drop=True)],
axis=1)
ts = test_df.copy()
tr = train_df.copy()
for j, c in enumerate(classifiers):
print('alg: ', j)
try:
c.fit(tr)
res = c.predict(ts.iloc[:, :-1])
cm = confusion_matrix(y_pred=res, y_true=ts['Class'].values)
acc[j].append(Accuracy(y_pred=res, y_true=ts['Class'].values))
#print(cm)
#print("Accuracy: {:.3}".format(acc[-1]))
except Exception as e:
c.fit(tr.iloc[:, :-1], tr.iloc[:, -1])
res = c.predict(ts.iloc[:, :-1].values)
cm = confusion_matrix(y_pred=res, y_true=ts['Class'].values)
acc[j].append(Accuracy(y_pred=res, y_true=ts['Class'].values))
#print(cm)
#print("Accuracy: {:.3}".format(acc[-1]))
acc_mean = [sum(a) / len(a) for a in acc]
return acc, acc_mean
