def test_learning(method='drop'):
db = get_cleaned_data_from_DB(method)
db_list = []
target = []
test = []
y_test = []
i = 0
for line in db:
line_list = []
for key in line:
if key == 'id':
continue
elif key == 'target':
cls = line[key]
if cls > 0:
cls = 1
if i < 270:
target.append(cls)
else:
y_test.append(cls)
elif key == 'thal':
if line[key] == 3:
line_list += [1.0, 0.0, 0.0]
elif line[key] == 6:
line_list += [0.0, 1.0, 0.0]
else:
line_list += [0.0, 0.0, 1.0]
elif key == 'pain_type':
pain_type = [0.0, 0.0, 0.0, 0.0]
pain_type[int(line[key]) - 1] = 1.0
line_list += pain_type
elif key == 'electrocardiographic':
el = [0.0, 0.0, 0.0, 0.0]
el[int(line[key])] = 1.0
line_list += el
else:
line_list.append(line[key])
if i < 270:
db_list.append(line_list)
else:
test.append(line_list)
i += 1
X = np.array(db_list)
X = normalize(X, axis=0, norm='max')
y = np.array(target)
test = np.array(test)
test = normalize(test, axis=0, norm='max')
clf = StackClassification()
print(X.shape, y.shape)
clf.fit(X, y)
print(clf.predict(test))
for n in y_test:
print(n, end=' ')
def learningCurve(method='drop', model_type='stack'):
db = dbManipulation.get_cleaned_data_from_DB(method)
db_list = []
target = []
for line in db:
line_list = []
for key in line:
if key == 'id':
continue
elif key == 'target':
cls = line[key]
if cls > 0:
cls = 1
target.append(cls)
elif key == 'thal':
if line[key] == 3:
line_list += [1.0, 0.0, 0.0]
elif line[key] == 6:
line_list += [0.0, 1.0, 0.0]
else:
line_list += [0.0, 0.0, 1.0]
elif key == 'pain_type':
pain_type = [0.0, 0.0, 0.0, 0.0]
pain_type[int(line[key]) - 1] = 1.0
line_list += pain_type
elif key == 'electrocardiographic':
el = [0.0, 0.0, 0.0, 0.0]
el[int(line[key])] = 1.0
line_list += el
else:
line_list.append(line[key])
db_list.append(line_list)
X = np.array(db_list)
X = normalize(X, axis=0, norm='max')
y = np.array(target)
print("Start training...")
if model_type == 'stack':
clf = StackClassification()
else:
clf = LogisticRegression(C=0.01, penalty='l2')
train_sizes, train_scores, valid_scores = learning_curve(
clf,
X,
y,
train_sizes=[
20, 40, 60, 80, 100, 120, 140, 160, 180, 200, 220, 240, 260
],
cv=10,
scoring='accuracy')
print("finish training")
return train_sizes, np.mean(train_scores, axis=1), np.mean(valid_scores,
axis=1)
def clustering(cluster_method='kmeans'):
db = get_cleaned_data_from_DB(method='drop')
db_list = []
target = []
for line in db:
line_list = []
for key in line:
if key == 'id':
continue
elif key == 'target':
cls = line[key]
if cls > 0:
cls = 1
target.append(cls)
elif key == 'thal':
if line[key] == 3:
line_list += [1.0, 0.0, 0.0]
elif line[key] == 6:
line_list += [0.0, 1.0, 0.0]
else:
line_list += [0.0, 0.0, 1.0]
elif key == 'pain_type':
pain_type = [0.0, 0.0, 0.0, 0.0]
pain_type[int(line[key]) - 1] = 1.0
line_list += pain_type
elif key == 'electrocardiographic':
el = [0.0, 0.0, 0.0, 0.0]
el[int(line[key])] = 1.0
line_list += el
else:
line_list.append(line[key])
db_list.append(line_list)
X = np.array(db_list)
X = normalize(X, axis=0, norm='max')
y = np.array(target)
pca = PCA(n_components=2)
pca.fit(X.T)
X = pca.components_.T
if cluster_method == 'spectral':
clustering = SpectralClustering(n_clusters=2,
assign_labels="discretize",
random_state=0).fit(X)
else:
clustering = KMeans(n_clusters=2, random_state=0).fit(X)
return X, clustering.labels_, y
def training_model(method='drop', model_type='stack'):
db = dbManipulation.get_cleaned_data_from_DB(method)
db_list = []
target = []
for line in db:
line_list = []
for key in line:
if key == 'id':
continue
elif key == 'target':
cls = line[key]
if cls > 0:
cls = 1
target.append(cls)
elif key == 'thal':
if line[key] == 3:
line_list += [1.0, 0.0, 0.0]
elif line[key] == 6:
line_list += [0.0, 1.0, 0.0]
else:
line_list += [0.0, 0.0, 1.0]
elif key == 'pain_type':
pain_type = [0.0, 0.0, 0.0, 0.0]
pain_type[int(line[key]) - 1] = 1.0
line_list += pain_type
elif key == 'electrocardiographic':
el = [0.0, 0.0, 0.0, 0.0]
el[int(line[key])] = 1.0
line_list += el
else:
line_list.append(line[key])
db_list.append(line_list)
X = np.array(db_list)
X = normalize(X, axis=0, norm='max')
y = np.array(target)
if model_type == 'stack':
clf = StackClassification()
else:
clf = LogisticRegression(C=0.01, penalty='l2')
clf.fit(X, y)
return clf
def cleanInput(inputX):
cleanData = get_cleaned_data_from_DB(method='drop')
cleaned_List = []
for data in cleanData:
line_list = []
for key in data:
if key == 'target' or key == 'id':
continue
else:
line_list.append(data[key])
cleaned_List.append(line_list)
cleaned_List = np.array(cleaned_List)
mean_vals = np.mean(cleaned_List, axis=0)
key_map = feature_map()
input_list = []
for i in range(1, 14):
if i == 10 and (inputX[key_map[i]] == 0 or inputX[key_map[i]] == '?'):
il = 1.0
elif inputX[key_map[i]] != '?':
il = float(inputX[key_map[i]])
else:
il = inputX[key_map[i]]
input_list.append(il)
missing_index = []
for i in range(len(input_list)):
if input_list[i] == '?':
input_list[i] = mean_vals[i]
missing_index.append(i)
for i in missing_index:
X = np.concatenate((cleaned_List[:, :i], cleaned_List[:, i + 1:]),
axis=1)
y = cleaned_List[:, i:i + 1].reshape(1, -1)[0]
test = np.array(input_list[:i] + input_list[i + 1:]).reshape(1, -1)
input_list[i] = arrayKNN(X, y, test)[0]
cleandict = {}
for i in range(1, 14):
cleandict[key_map[i]] = input_list[i - 1]
line_list = []
for key in cleandict:
if key == 'thal':
if cleandict[key] == 3:
line_list += [1.0, 0.0, 0.0]
elif cleandict[key] == 6:
line_list += [0.0, 1.0, 0.0]
else:
line_list += [0.0, 0.0, 1.0]
elif key == 'pain_type':
pain_type = [0.0, 0.0, 0.0, 0.0]
pain_type[int(cleandict[key]) - 1] = 1.0
line_list += pain_type
elif key == 'electrocardiographic':
el = [0.0, 0.0, 0.0, 0.0]
el[int(cleandict[key])] = 1.0
line_list += el
else:
line_list.append(cleandict[key])
InputX = np.array(line_list)
db = get_cleaned_data_from_DB(method='drop')
db_list = []
target = []
for line in db:
line_list = []
for key in line:
if key == 'id':
continue
elif key == 'target':
cls = line[key]
if cls > 0:
cls = 1
target.append(cls)
elif key == 'thal':
if line[key] == 3:
line_list += [1.0, 0.0, 0.0]
elif line[key] == 6:
line_list += [0.0, 1.0, 0.0]
else:
line_list += [0.0, 0.0, 1.0]
elif key == 'pain_type':
pain_type = [0.0, 0.0, 0.0, 0.0]
pain_type[int(line[key]) - 1] = 1.0
line_list += pain_type
elif key == 'electrocardiographic':
el = [0.0, 0.0, 0.0, 0.0]
el[int(line[key])] = 1.0
line_list += el
else:
line_list.append(line[key])
db_list.append(line_list)
X = np.array(db_list)
concat = np.concatenate((X, InputX.reshape(1, -1)), axis=0)
concat = normalize(concat, axis=0, norm='max')
return concat[-1, :].reshape(1, -1)