Python 曲线拟合的最小二乘法

模块导入

import numpy as np
import gaosi as gs

代码

"""
本函数通过创建增广矩阵，并调用高斯列主元消去法模块进行求解。

"""
import numpy as np
import gaosi as gs

shape = int(input('请输入拟合函数的次数:'))

x = np.array([0.6,1.3,1.64,1.8,2.1,2.3,2.44])
y = np.array([7.05,12.2,14.4,15.2,17.4,19.6,20.2])
data = []
for i in range(shape*2+1):
    if i != 0:
        data.append(np.sum(x**i))
    else:
        data.append(len(x))
b = []
for i in range(shape+1):
    if i != 0:
        b.append(np.sum(y*x**i))
    else:
        b.append(np.sum(y))
b = np.array(b).reshape(shape+1,1)
n = np.zeros([shape+1,shape+1])
for i in range(shape+1):
    for j in range(shape+1):
        n[i][j] = data[i+j]
result = gs.Handle(n,b)
if not result:
    print('增广矩阵求解失败！')
    exit()
fun='f(x) = '
for i in range(len(result)):
    if type(result[i]) == type(''):
        print('存在自由变量！')
        fun = fun + str(result[i])
    elif i == 0:
        fun = fun + '{:.3f}'.format(result[i])
    else:
        fun = fun + '+{0:.3f}*x^{1}'.format(result[i],i)
print('求得{0}次拟合函数为：'.format(shape))
print(fun)

高斯模块

# 导入 numpy 模块
import numpy as np


# 行交换
def swap_row(matrix, i, j):
    m, n = matrix.shape
    if i >= m or j >= m:
        print('错误! : 行交换超出范围 ...')
    else:
        matrix[i],matrix[j] = matrix[j].copy(),matrix[i].copy()
    return matrix


# 变成阶梯矩阵
def matrix_change(matrix):
    m, n = matrix.shape
    main_factor = []
    main_col = main_row = 0
    while main_row < m and main_col < n:
        # 选择进行下一次主元查找的列
        main_row = len(main_factor)
        # 寻找列中非零的元素
        not_zeros = np.where(abs(matrix[main_row:,main_col]) > 0)[0]
        # 如果该列向下全部数据为零，则直接跳过列
        if len(not_zeros) == 0:
            main_col += 1
            continue
        else:
            # 将主元列号保存在列表中
            main_factor.append(main_col)
            # 将第一个非零行交换至最前
            if not_zeros[0] != [0]:
                matrix = swap_row(matrix,main_row,main_row+not_zeros[0])
            # 将该列主元下方所有元素变为零
            if main_row < m-1:
                for k in range(main_row+1,m):
                    a = float(matrix[k, main_col] / matrix[main_row, main_col])
                    matrix[k] = matrix[k] - matrix[main_row] * matrix[k, main_col] / matrix[main_row, main_col]
            main_col += 1
    return matrix,main_factor


# 回代求解
def back_solve(matrix, main_factor):
    # 判断是否有解
    if len(main_factor) == 0:
        print('主元错误,无主元！ ...')
        return None
    m, n = matrix.shape
    if main_factor[-1] == n - 1:
        print('无解！ ...')
        return None
    # 把所有的主元元素上方的元素变成0
    for i in range(len(main_factor) - 1, -1, -1):
        factor = matrix[i, main_factor[i]]
        matrix[i] = matrix[i] / float(factor)
        for j in range(i):
            times = matrix[j, main_factor[i]]
            matrix[j] = matrix[j] - float(times) * matrix[i]
    # 先看看结果对不对
    return matrix


# 结果打印
def print_result(matrix, main_factor):
    if matrix is None:
        print('阶梯矩阵为空！ ...')
        return None
    m, n = matrix.shape
    result = [''] * (n - 1)
    main_factor = list(main_factor)
    for i in range(n - 1):
        # 如果不是主元列，则为自由变量
        if i not in main_factor:
            result[i] = '(free var)'
        # 否则是主元变量，从对应的行，将主元变量表示成非主元变量的线性组合
        else:
            # row_of_main表示该主元所在的行
            row_of_main = main_factor.index(i)
            result[i] = matrix[row_of_main, -1]
    return result


# 得到简化的阶梯矩阵和主元列
def Handle(matrix_a, matrix_b):
    # 拼接成增广矩阵
    matrix_01 = np.hstack([matrix_a, matrix_b])
    matrix_01, main_factor = matrix_change(matrix_01)
    matrix_01 = back_solve(matrix_01, main_factor)
    result = print_result(matrix_01, main_factor)
    return result


if __name__ == '__main__':
    a = np.array([[2, 1, 1], [3, 1, 2], [1, 2, 2]],dtype=float)
    b = np.array([[4],[6],[5]],dtype=float)
    a = Handle(a, b)