def test():
df = pd.read_csv('sample/boston.csv')
y = df['Price']
X = df.drop(['Price'], 1)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3)
model = Model()
model.add_layer(DataFrameInputs())
model.add_layer(PolynomialLayer())
model.add_layer(StandardScalerLayer())
model.add_layer(EnsembleGBRLayer(n_models=10, row_ratio=0.5,
col_ratio=0.5, scale=True))
model.add_layer(EnsembleRidgeLayer(n_models=30, row_ratio=0.6,
col_ratio=0.6, scale=True))
model.add_layer(AverageLayer())
model.train(X_train, y_train)
model.summary()
model.save('sample/test.pkl')
with open('sample/test.pkl', 'rb') as f:
model = pickle.load(f)
X_pred = model.predict(X_test)
print("Predict shape", X_pred.shape)
X_pred = model.predict_proba(X_test)
print("Predict_proba shape", X_pred[1].shape)
score = model.score(X_test, y_test)
print(score)
def test_8():
print("test7")
df = pd.read_csv('sample/boston2.csv')
y = df['Price']
X = df.drop(['Price'], 1)
model = Model()
model.add_layer(DataFrameInputs())
model.add_layer(DuplicateLayer(n=1))
model.add_layer(MultiheadRidgeLayer(n_models=300, row_ratio=1.,
col_ratio=0.1, scale=True))
model.add_layer(MultiheadAggregateLayer())
model.add_layer(EnsembleRidgeLayer(n_models=100, row_ratio=1.,
col_ratio=0.3, scale=True))
model.add_layer(AverageLayer())
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3)
pipeline = ClusterDataset(model, n_poly=2, n_cluster=3)
pipeline.train(X_train, y_train)
pred, proba = pipeline.predict_proba(X_test)
plt.scatter(y_test.values, pred)
plt.show()
def test_7():
print("test7")
df = pd.read_csv('sample/boston2.csv')
y = df['Price']
X = df.drop(['Price'], 1)
model = Model()
model.add_layer(DataFrameInputs())
model.add_layer(DuplicateLayer(n=1))
model.add_layer(MultiheadSVRLayer(n_models=10, row_ratio=0.8,
col_ratio=0.6, scale=True))
model.add_layer(MultiheadRidgeLayer(n_models=3, row_ratio=0.8,
col_ratio=0.8, scale=True))
model.add_layer(MultiheadAggregateLayer())
model.add_layer(EnsembleRidgeLayer(n_models=1, row_ratio=1.,
col_ratio=1., scale=True))
model.add_layer(AverageLayer())
pipeline = TableDataset(model, n_poly=2, prescale=False)
pipeline.train(X, y)
X_sample = X.iloc[0, :]
pred, proba = pipeline.predict_proba(X_sample)
print(pred, proba)
pipeline.valid(X, y)
def test_5():
"""
基本形のコード
"""
df = pd.read_csv('sample/boston.csv')
y = df['Price']
X = df.drop(['Price'], 1)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3)
model = Model()
model.add_layer(DataFrameInputs())
model.add_layer(PolynomialLayer())
model.add_layer(StandardScalerLayer())
model.add_layer(DuplicateLayer(n=1))
model.add_layer(MultiheadGBRLayer(n_models=50, row_ratio=0.8,
col_ratio=0.8, scale=True))
model.add_layer(MultiheadRidgeLayer(n_models=5, row_ratio=0.8,
col_ratio=0.8, scale=True))
model.add_layer(MultiheadAggregateLayer())
model.add_layer(EnsembleRidgeLayer(n_models=1, row_ratio=1.,
col_ratio=1., scale=True))
model.add_layer(AverageLayer())
model.valid(X, y, n=3, cv='KFold')
def test_4():
df = pd.read_csv('sample/boston.csv')
y = df['Price']
X = df.drop(['Price'], 1)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3)
model = Model()
model.add_layer(DataFrameInputs())
model.add_layer(PolynomialLayer())
model.add_layer(StandardScalerLayer())
model.add_layer(DuplicateLayer(n=1))
model.add_layer(MultiheadGBRLayer(n_models=50, row_ratio=0.8,
col_ratio=0.8, scale=True))
model.add_layer(MultiheadRidgeLayer(n_models=10, row_ratio=0.8,
col_ratio=0.8, scale=True))
model.add_layer(MultiheadAggregateLayer())
model.add_layer(EnsembleRidgeLayer(n_models=1, row_ratio=1.,
col_ratio=1., scale=True))
model.add_layer(AverageLayer())
model.train(X_train, y_train)
X_pred = model.predict(X_test)
print("Predict shape", X_pred.shape)
model.summary()
score = model.score(X_test, y_test)
print(score)
def test_10():
df = pd.read_csv('sample/smi.csv')
y = df['ESOL']
X = df['smiles']
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3)
model = Model()
model.add_layer(SmilesInputs())
model.add_layer(MultiheadGBRLayer(n_models=3, row_ratio=0.8,
col_ratio=0.8, scale=True))
model.add_layer(MultiheadAggregateLayer())
model.add_layer(EnsembleRidgeLayer(n_models=1, row_ratio=1.,
col_ratio=1., scale=True))
model.add_layer(AverageLayer())
model.valid(X, y, n=3, cv='KFold')
def test_3():
df = pd.read_csv('sample/smi.csv')
y = df['ESOL']
X = df['smiles']
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3)
model = Model()
model.add_layer(SmilesInputs())
model.add_layer(MultiheadGBRLayer(n_models=50, row_ratio=0.6,
col_ratio=0.6, scale=True))
model.add_layer(MultiheadRidgeLayer(n_models=30, row_ratio=0.9,
col_ratio=0.9, scale=True))
model.add_layer(MultiheadAggregateLayer())
model.add_layer(EnsembleRidgeLayer(n_models=1, row_ratio=1.,
col_ratio=1., scale=True))
model.add_layer(AverageLayer())
model.train(X_train, y_train)
model.summary()
score = model.score(X_test, y_test)
print(score)