def predict(self, t_x):
all_t = t_x[0:, 0]
x1 = t_x[0:, 1:]
x1_pred = _res_func_list.res_funcs[self.mdl_name].compute(
self.params, all_t, x1)
x_pred = Accumulation.agom(x1_pred, self.x_orig[0][0], False)
return x_pred
def fit(self, x, y):
x1 = Accumulation.agom(y, None, True)
x1_0 = x1[0:, 0]
z1 = ModelMethods.get_backvalue(x1_0)
n_array = x1[1:, 1:]
B = ModelMethods.construct_matrix(z1, n_array)
self.x_orig = y
self.params = ModelMethods.get_params(B, np.array(y)[0:, 0])
self.x1 = x1
return x1
def fit(self, x, y):
x1 = Accumulation.agom(y, None, True)
ones_array = np.diff(x).astype(np.float64)
ones_array = ones_array.reshape([-1, 1])
x1_0 = x1[0:-1, 0]
x1_0 = x1_0.reshape([-1, 1])
x1_n = x1[1:, 1:]
B_x = ModelMethods.construct_matrix(-x1_0, x1_n)
B = ModelMethods.construct_matrix(-B_x, ones_array)
self.x_orig = y
self.params = ModelMethods.get_params(B, np.array(y)[0:, 0])
self.x1 = x1
return x1
def fit(self, y, x):
x_train = x[0:, 1:]
y_conect = np.concatenate((y, x_train), axis=1)
x1 = Accumulation.agom(y_conect, None, True)
x1_0 = x1[0:, 0]
z1 = ModelMethods.get_backvalue(x1_0)
n_array = x1[1:, 1:]
self.B = ModelMethods.construct_matrix(z1, n_array)
self.x_orig = y
self.params = ModelMethods.get_params(self.B,
np.array(y_conect)[0:, 0])
self.x1 = x1
self.is_fitted = True
return self
def fit(self, y, x):
x_train = x[0:, 1:]
t = x[0:, 0]
y_conect = np.concatenate((y, x_train), axis=1)
x1 = Accumulation.agom(y_conect, None, True)
ones_array = np.diff(t).astype(np.float64)
ones_array = ones_array.reshape([-1, 1])
x1_0 = x1[0:-1, 0]
x1_0 = x1_0.reshape([-1, 1])
x1_n = x1[1:, 1:]
B_x = ModelMethods.construct_matrix(-x1_0, x1_n)
self.B = ModelMethods.construct_matrix(-B_x, ones_array)
self.x_orig = y
self.params = ModelMethods.get_params(self.B, np.array(y)[0:, 0])
self.x1 = x1
self.is_fitted = True
return self
def generate_report_gmn(self, mdl, t_train, x0_train, t_test, x0_test):
if (mdl.is_fitted == True):
t_all = np.concatenate((t_train, t_test), axis=0)
x_all = np.concatenate((x0_train, x0_test), axis=0)
t_x = np.concatenate((t_all, mdl.x1), axis=1)
x_pred = mdl.predict(t_x)
x_orig = mdl.x_orig[0:, 1]
lens = len(x_orig)
err = x_pred[:lens] - x_orig
m = len(t_train)
x0_train = x_all[:m]
x0_test = x_all[m - 1:]
t_train = t_all[:m]
t_test = t_all[m - 1:]
rmse = np.sqrt(sum((err)**2 / lens))
rmse = np.round(rmse, 4)
t3 = np.arange(lens) + 1
p = plt.figure()
plt.subplot(2, 1, 1)
plt.plot(t_train, x0_train)
plt.plot(t_train, x_pred[:m])
plt.plot(t_test, x0_test)
plt.plot(t_test, x_pred[m - 1:lens])
plt.axvline(x=m, ls="--", lw=1, c='k')
plt.legend([
'x_orig_models', 'x_pred_models', 'x_orig_detected',
'x_pred_detected'
])
plt.axvline(x=m, ls="--", lw=1, c='k')
plt.title('Model and Test')
plt.subplot(2, 1, 2)
plt.stem(t3, err, linefmt="c:", markerfmt="o", basefmt="r-")
plt.title('RMSE = {}'.format(rmse))
plt.axvline(x=m, ls="--", lw=1, c='k')
# 打印报告
x1 = Accumulation.agom(mdl.x_orig, None, True)
x1 = x1[0:, 1]
x1_pred = Accumulation.agom(x_pred, None, True)
model_report = pd.DataFrame({
'x_orig': x_orig,
'x1': x1,
'x1_pred': x1_pred,
'x_pred': x_pred,
'error': err,
})
# 公式表达
eq = generate_report.formula_expression_gmn(None, mdl)
c1 = [p, model_report, eq]
return c1
else:
t = np.concatenate((t_train, t_test), axis=1)
x0 = np.concatenate((x0_train, x0_test), axis=1)
mdl.fit(t, x0)
generate_report.generate_report_gmn(mdl, t_train, x0_train, t_test,
x0_test)
