如何使用Symfit或curve_fit对分段模型执行交叉验证？

import symfit as sf
import numpy as np
from scipy.optimize import curve_fit
from sklearn.model_selection import KFold

x, y = sf.variables('x y')
a, b, c = sf.parameters('a b c')

model = {y: a * x ** 2 + b * x + c}

def segmented_model(x, a1, b1, c1, a2, b2, c2):
    return np.piecewise(x,
                        [x < 0, x >= 0],
                        [lambda x: a1 * x ** 2 + b1 * x + c1,
                         lambda x: a2 * x ** 2 + b2 * x + c2])

x_data, y_data = np.array([...]), np.array([...])  # 填入实际数据

x_data_sf, y_data_sf = sf.Tuple([x, y], [x_data, y_data])

a1, b1, c1, a2, b2, c2 = sf.parameters('a1 b1 c1 a2 b2 c2')

model_sf = {y: segmented_model(x, a1, b1, c1, a2, b2, c2)}

def error_func(params, x, y):
    return np.sum((segmented_model(x, *params) - y) ** 2)

k = 5  # 交叉验证的折数
kf = KFold(n_splits=k)

errors = []
for train_index, test_index in kf.split(x_data):
    x_train, x_test = x_data[train_index], x_data[test_index]
    y_train, y_test = y_data[train_index], y_data[test_index]

    params, _ = curve_fit(segmented_model, x_train, y_train)
    error = error_func(params, x_test, y_test)
    errors.append(error)

average_error = np.mean(errors)