def fit(self, deathtraining, casetraining, K1, K2, TRAINERROR):
ts_1 = deathtraining
ts_2 = casetraining
T = len(ts_1)
y = ts_1[K1 - 1:T] # size = T-K
n = len(y)
y = y.reshape(n, 1)
X = np.ones((n, 1))
for i in range(K1 - 1):
x_j = ts_1[i:n + i].reshape(n, 1)
X = np.concatenate((X, x_j), axis=1)
for i in range(K2 - 1):
x_j2 = ts_2[i:n + i].reshape(n, 1)
X = np.concatenate((X, x_j2), axis=1)
#print(X.shape)
# train precess
model = linear_model.LeastSquares()
model.fit(X, y)
self.model = model
#print(model.w)
if TRAINERROR:
y_hat = model.predict(X)
#print(y_hat, y)
#plt.plot(y_hat)
tr_error = np.mean((y - y_hat)**2)
print("Training error = %.3f" % tr_error)
def fit(self, t, T, K):
ts = self.ts
y = ts[K - 1:T] # size = T-K
n = len(y)
y = y.reshape(n, 1)
X = np.ones((n, 1))
for i in range(K - 1):
x_j = ts[i:n + i].reshape(n, 1)
X = np.concatenate((X, x_j), axis=1)
# train
model = linear_model.LeastSquares()
model.fit(X, y)
yhat = model.predict(X)
trainError = np.mean((yhat - y)**2)
# print("Training error = %.3f" % trainError)
# T+1
X_test = np.concatenate(([1], ts[T - K + 1:T]), axis=0).reshape(1, K)
y_test = np.array(ts[T]).reshape(1, 1)
y_test_overall = y_test
y_pred = model.predict(X_test).reshape(1, 1)
# print("predicted case number: %.3f, real case number: %.3f" % (y_pred, y_test))
y_pred_overall = y_pred
# predict
for i in range(1, t):
# print("day#: %d" % (i + 1))
a = X_test[:, 2:]
X_test = np.concatenate(([[1]], a, y_pred), axis=1)
y_test_i = np.array(ts[T + i]).reshape(1, 1)
y_test_overall = np.concatenate((y_test_overall, y_test_i), axis=0)
y_pred = model.predict(X_test).reshape(1, 1)
# print("predicted case number: %.3f, real case number: %.3f" % (y_pred, y_test_i))
y_pred_overall = np.concatenate((y_pred_overall, y_pred), axis=0)
testError = np.mean((y_pred_overall - y_test_overall)**2)
# print("Test error = %.3f" % testError)
return [trainError, testError, model, y_test_overall, y]
plt.savefig(filename)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument('-q', '--question', required=True)
io_args = parser.parse_args()
question = io_args.question
if question == "2":
data = load_dataset("outliersData.pkl")
X = data['X']
y = data['y']
# Fit least-squares estimator
model = linear_model.LeastSquares()
model.fit(X, y)
print(model.w)
test_and_plot(model,
X,
y,
title="Least Squares",
filename="least_squares_outliers.pdf")
elif question == "2.1":
data = load_dataset("outliersData.pkl")
X = data['X']
y = data['y']
''' YOUR CODE HERE '''
model = linear_model.WeightedLeastSquares()
def fit_next_case_num(raw, feature1, feature2, country, t, T, K):
print(" ")
# print("predicting for country:%s" % (country))
ca = raw[raw["country_id"] == country]
#print(ca)
ca_case_ts = ca[feature1].values
ca_death_ts = ca[feature2].values # 280*1
#print(ca_case_ts)
ca_death_ts = np.log(ca_death_ts + 1)
y = ca_case_ts[K - 1:T] # size = T-K
n = len(y)
y = y.reshape(n, 1)
X = np.ones((n, 1))
for i in range(K - 1):
x_j = ca_case_ts[i:n + i].reshape(n, 1)
X = np.concatenate((X, x_j), axis=1)
for i in range(K - 1):
x_j2 = ca_death_ts[i:n + i].reshape(n, 1)
X = np.concatenate((X, x_j2), axis=1)
# train
model = linear_model.LeastSquares()
model.fit(X, y)
# print(model.w)
yhat = model.predict(X)
"""
for i in range(len(yhat)):
if ( yhat[i] - y[i] )>= 10:
print(i)
#plt.plot(yhat-y)
#plt.plot(y)
filename = os.path.join("..", "figs", "plotofyhat&y.png")
print("Saving", filename)
plt.savefig(filename)
"""
trainError = np.mean((yhat - y)**2)
print("Training error = %.3f" % trainError)
# T+1
X_test = np.concatenate(
([1], ca_case_ts[T - K + 1:T], ca_death_ts[T - K + 1:T]),
axis=0) #.reshape(1, 2*K-1)
print(X_test.shape)
y_test = np.array(ca_case_ts[T]).reshape(1, 1)
y_test_overall = y_test
y_pred = model.predict(X_test).reshape(1, 1)
# print("predicted case number: %.3f, real case number: %.3f" % (y_pred_curr, y_test))
y_pred_overall = y_pred
# predict
for i in range(1, t):
# print("day#: %d" % (i + 1))
# new X_(T+1)
#a=X_test[:,2:]
#print([[1]].shape)
X_test = np.concatenate(([1], ca_case_ts[T - K + i:T + i - 1],
ca_death_ts[T - K + i:T + i - 1]),
axis=0)
print(X_test.shape)
y_test_i = np.array(ca_case_ts[T + i]).reshape(1, 1)
y_test_overall = np.concatenate((y_test_overall, y_test_i), axis=0)
y_pred = model.predict(X_test).reshape(1, 1)
print("predicted case number: %.3f, real case number: %.3f" %
(y_pred, y_test_i))
y_pred_overall = np.concatenate((y_pred_overall, y_pred), axis=0)
# X_test = np.concatenate((X_test, X_test_i), axis=0)
# y_pred = model.predict(X_test)
testError = np.mean((y_pred_overall - y_test_overall)**2)
print("Test error = %.3f" % testError)
return [trainError, testError]
