def run_knn(self, k2, outputfile):
scores = '_knn_scores.txt'
print('Running model...')
forest = KNeighborsClassifier(train=self.x_train,
test=self.x_test,
k=k2,
scores=outputfile + "_" + str(k2) +
scores,
Y_train=self.y_train)
forest.preprocess_input_vectors()
pred, predprob = forest.build_knn_model()
# forest.model_evaluation(prediction, self.y_test)
f1, prec, rec, acc = forest.model_evaluation2(pred, predprob,
self.y_test)
print("f1_binary: %f" % f1)
print("Accuracy: %f" % acc)
print("Precision: %f" % prec)
print("Recall: %f" % rec)
PrecisionGraphics(
str(prunning) + '_' + str(balanced) + '_' + str(imputer) + scores,
str(prunning) + '_' + str(balanced) + '_' + str(imputer) +
"_precisionvsrecall.pdf")
NDCGGraphics(
str(prunning) + '_' + str(balanced) + '_' + str(imputer) + scores,
str(prunning) + '_' + str(balanced) + '_' + str(imputer) +
"_NDCG.pdf")
ids = GetIds("files3_claro_mig_model")
ids.gen_features(
str(prunning) + '_' + str(balanced) + '_' + str(imputer) +
"_ids.txt")
def run_rf(self, prunning=5, balanced=None, imputer=False):
scores = '_rf_scores.txt'
if (imputer):
imp = Imputer(missing_values=-1,
strategy='median').fit(self.x_train)
self.x_train = imp.transform(self.x_train)
self.x_test = imp.transform(self.x_test)
print('Gerando modelo...')
forest = RandomForestClassifier(train=self.x_train,
test=self.x_test,
prunning=prunning,
n_estimators=100,
balanced=balanced,
scores=str(prunning) + '_' +
str(balanced) + '_' + str(imputer) +
scores,
Y_train=self.y_train)
forest.preprocess_input_vectors()
pred, predprob = forest.build_rf_model()
# forest.model_evaluation(prediction, self.y_test)
f1, prec, rec, acc = forest.model_evaluation2(pred, predprob,
self.y_test)
print("f1_binary: %f" % f1)
print("Accuracy: %f" % acc)
print("Precision: %f" % prec)
print("Recall: %f" % rec)
PrecisionGraphics(
str(prunning) + '_' + str(balanced) + '_' + str(imputer) + scores,
str(prunning) + '_' + str(balanced) + '_' + str(imputer) +
"_precisionvsrecall.pdf")
NDCGGraphics(
str(prunning) + '_' + str(balanced) + '_' + str(imputer) + scores,
str(prunning) + '_' + str(balanced) + '_' + str(imputer) +
"_NDCG.pdf")
ids = GetIds("files3_claro_mig_model")
ids.gen_features(
str(prunning) + '_' + str(balanced) + '_' + str(imputer) +
"_ids.txt")
def run_rf(self, prunning=5, balanced='balanced', imputer=False):
scores = '_rf_scores.txt'
if (imputer):
imp = Imputer(missing_values=-1,
strategy='median').fit(self.x_train)
self.x_train = imp.transform(self.x_train)
self.x_test = imp.transform(self.x_test)
print('Running model...')
forest = RandomForestClassifier(train=self.x_train,
test=self.x_test,
prunning=prunning,
n_estimators=100,
balanced=balanced,
scores=str(prunning) + '_' +
str(balanced) + '_' + str(imputer) +
scores,
Y_train=self.y_train)
index_1 = np.where(self.y_train == 1)
data_y_balanced_1 = self.y_train[index_1]
index_0 = np.where(self.y_train == 0)
data_y_balanced_0 = self.y_train[
index_0[0]] # [0:30*len(data_y_balanced_1[:])]]
data_y_balanced = np.array(
np.hstack([data_y_balanced_0, data_y_balanced_1]))
data_x_balanced_1 = self.x_train[index_1]
data_x_balanced_0 = self.x_train[
index_0[0]] # [0:30*len(data_x_balanced_1[:])]]
data_x_balanced = np.array(
np.vstack([data_x_balanced_0, data_x_balanced_1]))
del data_x_balanced_0
clf = MLPClassifier(solver='sgd',
alpha=1e-5,
hidden_layer_sizes=(50, 100, 50),
max_iter=50,
random_state=1,
verbose=1)
filename = 'clf_model_tT4.sav'
try:
clf = joblib.load(filename)
except:
clf.fit(np.nan_to_num(data_x_balanced), data_y_balanced)
# save the model to disk
joblib.dump(clf, filename)
pred = clf.predict(np.nan_to_num(self.x_test))
print(classification_report(self.y_test[:len(pred)], pred))
#predprob = clf.predict_proba(np.nan_to_num(self.x_test))[:, 1]
forest.preprocess_input_vectors()
pred, predprob = forest.build_rf_model()
# forest.model_evaluation(prediction, self.y_test)
f1, prec, rec, acc = forest.model_evaluation2(pred, predprob,
self.y_test)
print("f1_binary: %f" % f1)
print("Accuracy: %f" % acc)
print("Precision: %f" % prec)
print("Recall: %f" % rec)
plt.hist(predprob)
plt.title("Histogram")
plt.xlabel("Value")
plt.ylabel("Frequency")
plt.show()
PrecisionGraphics(
str(prunning) + '_' + str(balanced) + '_' + str(imputer) + scores,
str(prunning) + '_' + str(balanced) + '_' + str(imputer) +
"_precisionvsrecall.pdf")
NDCGGraphics(
str(prunning) + '_' + str(balanced) + '_' + str(imputer) + scores,
str(prunning) + '_' + str(balanced) + '_' + str(imputer) +
"_NDCG.pdf")
ids = GetIds("files3_claro_mig_model_2")
ids.gen_features(
str(prunning) + '_' + str(balanced) + '_' + str(imputer) +
"_ids.txt")