def train(params=[10**-6, 10**-6, 10**-6, 10**-6]):
global data_products, model_products
alpha1, alpha2, lambda1, lambda2 = params
for product_id in data_products:
data = data_products[product_id].dropna()
if len(data.index) <= 0:
return
X = data[[
'amount_of_all_competitors', 'average_price_on_market',
'distance_to_cheapest_competitor', 'price_rank', 'quality_rank'
]]
y = data['sold'].copy()
y[y > 1] = 1
model = BayesianRidge(n_iter=1000, tol=0.0001, normalize=True)
model.set_params(alpha_1=alpha1,
alpha_2=alpha2,
lambda_1=lambda1,
lambda_2=lambda2)
model.fit(X, y)
model_products[product_id] = model
def bayes_regr(X, Y):
regr = BayesianRidge(compute_score=True)
regr.set_params(alpha_1=10, lambda_1=1e-3)
regr.fit(X, Y)
w_hat = regr.coef_
# Y_pred = regr.predict(X)
return w_hat
def plot_bayesian_ridge_curvefit():
def func(x):
return np.sin(2 * np.pi * x)
# #############################################################################
# Generate sinusoidal data with noise
size = 25
rng = np.random.RandomState(1234)
x_train = rng.uniform(0., 1., size)
y_train = func(x_train) + rng.normal(scale=0.1, size=size)
x_test = np.linspace(0., 1., 100)
# #############################################################################
# Fit by cubic polynomial
n_order = 3
X_train = np.vander(x_train, n_order + 1, increasing=True)
X_test = np.vander(x_test, n_order + 1, increasing=True)
# #############################################################################
# Plot the true and predicted curves with log marginal likelihood (L)
reg = BayesianRidge(tol=1e-6, fit_intercept=False, compute_score=True)
fig, axes = plt.subplots(1, 2, figsize=(8, 4))
for i, ax in enumerate(axes):
# Bayesian ridge regression with different initial value pairs
if i == 0:
init = [1 / np.var(y_train), 1.] # Default values
elif i == 1:
init = [1., 1e-3]
reg.set_params(alpha_init=init[0], lambda_init=init[1])
reg.fit(X_train, y_train)
ymean, ystd = reg.predict(X_test, return_std=True)
ax.plot(x_test, func(x_test), color="blue", label="sin($2\\pi x$)")
ax.scatter(x_train, y_train, s=50, alpha=0.5, label="observation")
ax.plot(x_test, ymean, color="red", label="predict mean")
ax.fill_between(x_test,
ymean - ystd,
ymean + ystd,
color="pink",
alpha=0.5,
label="predict std")
ax.set_ylim(-1.3, 1.3)
ax.legend()
title = "$\\alpha$_init$={:.2f},\\ \\lambda$_init$={}$".format(
init[0], init[1])
if i == 0:
title += " (Default)"
ax.set_title(title, fontsize=12)
text = "$\\alpha={:.1f}$\n$\\lambda={:.3f}$\n$L={:.1f}$".format(
reg.alpha_, reg.lambda_, reg.scores_[-1])
ax.text(0.05, -1.0, text, fontsize=12)
plt.tight_layout()
plt.show()
class BayesianLinearRegression:
def __init__(self, alphaInit=1., lambdaInit=0.2):
self.alphaInit = alphaInit
self.lambdaInit = lambdaInit
self.clf = BayesianRidge(fit_intercept=False)
self.clf.set_params(alpha_init=self.alphaInit,
lambda_init=self.lambdaInit)
def fit(self, X_train, y_train):
self.clf.fit(X_train, y_train)
def predict(self, X_test):
return self.clf.predict(X_test.astype(np.float32), return_std=True)
def find_best_bayesian_ridge(X_train, y_train):
reg = BayesianRidge(compute_score=True, tol=1e-5)
parameters = {
'alpha_init': (0.2, 0.5, 1, 1.5),
'lambda_init': [1e-3, 1e-4, 1e-5, 1e-6]
}
srch = GridSearchCV(reg, parameters)
srch.fit(X_train, y_train)
params = srch.get_params()
reg.set_params(alpha_init=params["estimator__alpha_init"],
lambda_init=params["estimator__lambda_init"])
reg.fit(X_train, y_train)
return reg, params
def trainModels(self, dates:np.array, scaler:MinMaxScaler):
# First, splitting data
x_train, x_test, y_train, y_test = train_test_split(scaler.transform(dates.reshape(-1, 1)), self.data_scaler.transform(self.tendency.values.reshape(-1 , 1)), test_size=0.2)
x_train = np.vander(x_train.reshape(-1), self.n + 1, increasing=True)
x_test = np.vander(x_test.reshape(-1), self.n + 1, increasing=True)
# Bayesian Ridge
bayRidge = BayesianRidge(tol=1e-6, compute_score=True)
bayRidge.set_params(alpha_init=1, lambda_init=0.001)
bayRidge.fit(x_train, y_train.reshape(-1))
bayScore = bayRidge.score(x_test, y_test.reshape(-1))
# SVR
svr_rbf = SVR(kernel='rbf', C=1, gamma=0.25)
svr_rbf.fit(x_train, y_train.reshape(-1))
svr_score = svr_rbf.score(x_test, y_test.reshape(-1))
return bayRidge, svr_rbf, np.array([bayScore, svr_score])
def cross_validate(params):
global test_data_products, model_products
alpha1, alpha2, lambda1, lambda2 = params
data = test_data_products[1].dropna()
if len(data.index) <= 0:
return 0
X = data[[
'amount_of_all_competitors', 'average_price_on_market',
'distance_to_cheapest_competitor', 'price_rank', 'quality_rank'
]]
y = data['sold'].copy()
y[y > 1] = 1
model = BayesianRidge(n_iter=1000, tol=0.0001, normalize=True)
model.set_params(alpha_1=alpha1,
alpha_2=alpha2,
lambda_1=lambda1,
lambda_2=lambda2)
score = -np.mean(cross_val_score(model, X, y, cv=3, scoring='r2'))
return score
X_test = np.vander(x_test, n_order + 1, increasing=True)
reg = BayesianRidge(tol=1e-6, fit_intercept=False, compute_score=True)
# %%
# Plot the true and predicted curves with log marginal likelihood (L)
# -------------------------------------------------------------------
import matplotlib.pyplot as plt
fig, axes = plt.subplots(1, 2, figsize=(8, 4))
for i, ax in enumerate(axes):
# Bayesian ridge regression with different initial value pairs
if i == 0:
init = [1 / np.var(y_train), 1.0] # Default values
elif i == 1:
init = [1.0, 1e-3]
reg.set_params(alpha_init=init[0], lambda_init=init[1])
reg.fit(X_train, y_train)
ymean, ystd = reg.predict(X_test, return_std=True)
ax.plot(x_test, func(x_test), color="blue", label="sin($2\\pi x$)")
ax.scatter(x_train, y_train, s=50, alpha=0.5, label="observation")
ax.plot(x_test, ymean, color="red", label="predict mean")
ax.fill_between(x_test,
ymean - ystd,
ymean + ystd,
color="pink",
alpha=0.5,
label="predict std")
ax.set_ylim(-1.3, 1.3)
ax.legend()
title = "$\\alpha$_init$={:.2f},\\ \\lambda$_init$={}$".format(
break
X_train.shape, X_test_scaled.shape
y_mlp = best_model.predict(X_test_scaled)
reg = BayesianRidge(compute_score=True, tol=1e-5)
parameters = {'alpha_init':(0.2, 0.5, 1, 1.5), 'lambda_init':[1e-3, 1e-4, 1e-5,1e-6]}
srch = GridSearchCV(reg, parameters)
srch.fit(X_train, y_train)
params = srch.get_params()
reg.set_params(alpha_init=params["estimator__alpha_init"], lambda_init=params["estimator__lambda_init"])
reg.fit(X_train, y_train)
ymean, ystd = reg.predict(X_test, return_std=True)
# exit()
folder = "./models/seirhcd/{}".format(current_dataset_date)
os.makedirs(folder, exist_ok=True)
joblib.dump(best_model, '{}/mlp.save'.format(folder))
joblib.dump(scaler, "{}/scaler.save".format(folder))
with open('{}/metrics.json'.format(folder), 'w') as fp:
json.dump({"perf":reports,"std_test":list(ystd.values), "columns":columns, "countries":list(all_countries)}, fp)
merged.to_csv('{}/features.csv'.format(folder))
# Fit by cubic polynomial
n_order = 3
X_train = np.vander(x_train, n_order + 1, increasing=True)
X_test = np.vander(x_test, n_order + 1, increasing=True)
# #############################################################################
# Plot the true and predicted curves with log marginal likelihood (L)
reg = BayesianRidge(tol=1e-6, fit_intercept=False, compute_score=True)
fig, axes = plt.subplots(1, 2, figsize=(8, 4))
for i, ax in enumerate(axes):
# Bayesian ridge regression with different initial value pairs
if i == 0:
init = [1 / np.var(y_train), 1.] # Default values
elif i == 1:
init = [1., 1e-3]
reg.set_params(alpha_init=init[0], lambda_init=init[1])
reg.fit(X_train, y_train)
ymean, ystd = reg.predict(X_test, return_std=True)
ax.plot(x_test, func(x_test), color="blue", label="sin($2\\pi x$)")
ax.scatter(x_train, y_train, s=50, alpha=0.5, label="observation")
ax.plot(x_test, ymean, color="red", label="predict mean")
ax.fill_between(x_test, ymean-ystd, ymean+ystd,
color="pink", alpha=0.5, label="predict std")
ax.set_ylim(-1.3, 1.3)
ax.legend()
title = "$\\alpha$_init$={:.2f},\\ \\lambda$_init$={}$".format(
init[0], init[1])
if i == 0:
title += " (Default)"
ax.set_title(title, fontsize=12)
def bayes_regression(params):
clf = BayesianRidge()
clf.set_params(**params)
return clf
