def knn(wine_train, wine_test, heart_train, heart_test):
wine_train_scores = []
heart_train_scores = []
wine_test_scores = []
heart_test_scores = []
for k in tqdm(range(1, 19, 2)):
heart_knn = alg.NearestNeighbors(k)
heart_knn.train(heart_train)
accuracy = heart_knn.score(heart_train)
heart_train_scores.append({'k': k, 'accuracy': accuracy})
accuracy = heart_knn.score(heart_test)
heart_test_scores.append({'k': k, 'accuracy': accuracy})
wine_knn = alg.NearestNeighbors(k)
wine_knn.train(wine_train)
accuracy = wine_knn.score(wine_train)
wine_train_scores.append({'k': k, 'accuracy': accuracy})
accuracy = wine_knn.score(wine_test)
wine_test_scores.append({'k': k, 'accuracy': accuracy})
heart_train_results = pd.DataFrame.from_records(heart_train_scores)
wine_train_results = pd.DataFrame.from_records(wine_train_scores)
heart_test_results = pd.DataFrame.from_records(heart_test_scores)
wine_test_results = pd.DataFrame.from_records(wine_test_scores)
colors = alg.ColorPopper()
ax = heart_train_results.plot.scatter(x='k',
y='accuracy',
color=colors.pop(),
label='train')
heart_test_results.plot.scatter(x='k',
y='accuracy',
ax=ax,
color=colors.pop(),
label='test')
plt.ylim(bottom=0, top=1)
plt.grid()
plt.title('Heart Failure KNN Train/Test Accurarcy')
plt.legend(loc='lower left')
plt.savefig('./knn_heart.png')
colors = alg.ColorPopper()
ax = wine_train_results.plot.scatter(x='k',
y='accuracy',
color=colors.pop(),
label='train')
wine_test_results.plot.scatter(x='k',
y='accuracy',
ax=ax,
color=colors.pop(),
label='test')
plt.ylim(bottom=0, top=1)
plt.grid()
plt.title('Wine Quality KNN Train/Test Accurarcy')
plt.legend(loc='lower left')
plt.savefig('./knn_wine.png')
def boosting(wine_train, wine_test, heart_train, heart_test):
wine_train_scores = []
heart_train_scores = []
wine_test_scores = []
heart_test_scores = []
for alpha in np.linspace(0, 0.05, 20):
heart_dt = alg.Boosting(alpha)
heart_dt.train(heart_train)
accuracy = heart_dt.score(heart_train)
heart_train_scores.append({'alpha': alpha, 'accuracy': accuracy})
accuracy = heart_dt.score(heart_test)
heart_test_scores.append({'alpha': alpha, 'accuracy': accuracy})
wine_dt = alg.Boosting(alpha)
wine_dt.train(wine_train)
accuracy = wine_dt.score(wine_train)
wine_train_scores.append({'alpha': alpha, 'accuracy': accuracy})
accuracy = wine_dt.score(wine_test)
wine_test_scores.append({'alpha': alpha, 'accuracy': accuracy})
heart_train_results = pd.DataFrame.from_records(heart_train_scores)
wine_train_results = pd.DataFrame.from_records(wine_train_scores)
heart_test_results = pd.DataFrame.from_records(heart_test_scores)
wine_test_results = pd.DataFrame.from_records(wine_test_scores)
colors = alg.ColorPopper()
ax = heart_train_results.plot.scatter(x='alpha',
y='accuracy',
color=colors.pop(),
label='train')
heart_test_results.plot.scatter(x='alpha',
y='accuracy',
ax=ax,
color=colors.pop(),
label='test')
plt.ylim(bottom=0, top=1)
plt.grid()
plt.title('Heart Failure Random Forest Train/Test Accurarcy')
plt.legend(loc='lower left')
plt.savefig('./boost_heart.png')
colors = alg.ColorPopper()
ax = wine_train_results.plot.scatter(x='alpha',
y='accuracy',
color=colors.pop(),
label='train')
wine_test_results.plot.scatter(x='alpha',
y='accuracy',
ax=ax,
color=colors.pop(),
label='test')
plt.ylim(bottom=0, top=1)
plt.grid()
plt.title('Wine Quality Random Forest Train/Test Accurarcy')
plt.legend(loc='lower left')
plt.savefig('./boost_wine.png')
def svm(wine_train, wine_test, heart_train, heart_test):
wine_train_scores = []
heart_train_scores = []
wine_test_scores = []
heart_test_scores = []
for kernel in tqdm(['linear', 'poly', 'rbf']):
heart_svm = alg.SupportVectorMachine(kernel)
heart_svm.train(heart_train)
accuracy = heart_svm.score(heart_train)
heart_train_scores.append({'kernel': kernel, 'accuracy': accuracy})
accuracy = heart_svm.score(heart_test)
heart_test_scores.append({'kernel': kernel, 'accuracy': accuracy})
wine_svm = alg.SupportVectorMachine(kernel)
wine_svm.train(wine_train)
accuracy = wine_svm.score(wine_train)
wine_train_scores.append({'kernel': kernel, 'accuracy': accuracy})
accuracy = wine_svm.score(wine_test)
wine_test_scores.append({'kernel': kernel, 'accuracy': accuracy})
heart_train_results = pd.DataFrame.from_records(heart_train_scores)
wine_train_results = pd.DataFrame.from_records(wine_train_scores)
heart_test_results = pd.DataFrame.from_records(heart_test_scores)
wine_test_results = pd.DataFrame.from_records(wine_test_scores)
colors = alg.ColorPopper()
ax = heart_train_results.plot.bar(x='kernel',
y='accuracy',
color=colors.pop(),
label='train')
heart_test_results.plot.bar(x='kernel',
y='accuracy',
ax=ax,
color=colors.pop(),
label='test')
plt.ylim(bottom=0, top=1)
plt.grid()
plt.title('Heart Failure SVM Train/Test Accurarcy')
plt.legend(loc='lower left')
plt.savefig('./svm_heart.png')
colors = alg.ColorPopper()
ax = wine_train_results.plot.bar(x='kernel',
y='accuracy',
color=colors.pop(),
label='train')
wine_test_results.plot.bar(x='kernel',
y='accuracy',
ax=ax,
color=colors.pop(),
label='test')
plt.ylim(bottom=0, top=1)
plt.grid()
plt.title('Wine Quality SVM Train/Test Accurarcy')
plt.legend(loc='lower left')
plt.savefig('./svm_wine.png')
def nn_plot(df):
colors = alg.ColorPopper()
fig, ax = plt.subplots()
for key, grp in df.groupby(['hidden_layers']):
ax = grp.plot.scatter(ax=ax,
x='hidden_size',
y='accuracy',
label=f'{key} hidden layers',
color=colors.pop())
plt.legend(loc='lower left')
plt.show()