y_pred = baseline_model.predict(x_test)
stop_test = time()
print("Accuracy: %.2f%% Training: %.2fs Testing: %.2fs" %
(f1_score(y_test, y_pred, average='macro') * 100,
stop_train - start_train, stop_test - start_test))
print(
pd.crosstab(y_pred,
y_test,
rownames=['True'],
colnames=['Predicted'],
margins=True))
print('\nClassification Report:\n', classification_report(y_test, y_pred))
plot_confusion_matrix(confusion_matrix(y_test,
y_pred,
labels=['win', 'loss', 'draw']),
target_names=['win', 'loss', 'draw'],
title="Baseline Support-Vector Machine")
tuned_model = svm.SVC(decision_function_shape='ovo',
C=10,
gamma=0.1,
kernel='rbf',
random_state=42)
start_train = time()
tuned_model.fit(x_train, y_train)
stop_train = time()
start_test = time()
y_pred = tuned_model.predict(x_test)
x_train, x_test, y_train, y_test = train_test_split(x_transformed,
y,
test_size=0.20,
random_state=42)
clf = GaussianNB()
start_train = time()
clf.fit(x_train, y_train)
stop_train = time()
start_test = time()
y_pred = clf.predict(x_test)
stop_test = time()
print("Accuracy: %.2f%% Training: %.2fs Testing: %.2fs" %
(f1_score(y_test, y_pred, average='macro') * 100,
stop_train - start_train, stop_test - start_test))
print(
pd.crosstab(y_test,
y_pred,
rownames=['True'],
colnames=['Predicted'],
margins=True))
print('\nClassification Report:\n', classification_report(y_test, y_pred))
plot_confusion_matrix(confusion_matrix(y_test,
y_pred,
labels=['win', 'loss', 'draw']),
target_names=['win', 'loss', 'draw'],
title="Gaussian Naive Bayes")
tm = MultiClassTsetlinMachine(2000, 80, 2.0, weighted_clauses=True)
start_train = time()
tm.fit(x_train, y_train, epochs=1, incremental=True)
stop_train = time()
start_test = time()
y_pred = tm.predict(x_test)
stop_test = time()
print("Accuracy: %.2f%% Training: %.2fs Testing: %.2f" %
(f1_score(y_test, y_pred, average='macro'), stop_train - start_train,
stop_test - start_test))
print(
pd.crosstab(y_test,
y_pred,
rownames=['True'],
colnames=["Predicted"],
margins=True))
print('\nClassification Report:\n', classification_report(y_test, y_pred))
y_test = le.inverse_transform(y_test)
y_pred = le.inverse_transform(y_pred)
plot_confusion_matrix(confusion_matrix(y_test,
y_pred,
labels=['win', 'loss', 'draw']),
target_names=['win', 'loss', 'draw'],
title="Optimized Tsetlin Machine")
stop_test = time()
print("Baseline")
print("Accuracy: %.2f%% Training: %.2fs Testing: %.2fs \n" %
(f1_score(y_pred, y_test, average='macro') * 100,
stop_train - start_train, stop_test - start_test))
print(
pd.crosstab(y_test,
y_pred,
rownames=['True'],
colnames=['Predicted'],
margins=True))
print('\nClassification Report:\n', classification_report(y_test, y_pred))
plot_confusion_matrix(confusion_matrix(y_test,
y_pred,
labels=['win', 'loss', 'draw']),
target_names=['win', 'loss', 'draw'],
title="Baseline KNN")
start_train = time()
tuned_model.fit(x_train, y_train)
stop_train = time()
start_test = time()
y_pred = tuned_model.predict(x_test)
stop_test = time()
print("Tuned")
print("Accuracy: %.2f%% Training: %.2fs Testing: %.2fs \n" %
(f1_score(y_pred, y_test, average='macro') * 100,
stop_train - start_train, stop_test - start_test))