Python数学建模 • Orac1e の blog

python 科学计算常用库#

数值计算-Numpy#

常用操作

1、创建数组

arr = np.array([1, 2, 3, 4, 5])

print(type(arr)) # <class 'numpy.ndarray'>

python

多维

arr = np.array([[1, 2, 3, 4, 5],
                [2, 3, 4, 5, 6]])

print(arr.shape) # 属性：(2, 5)

python

2、索引和切片

arr1 = np.array([1, 2, 3, 4, 5])
arr2 = np.array([[1, 2, 3],
                 [4, 5, 6],
                 [7, 8, 9]])

print(arr1[0])	# 1
print(arr1[0: 3])	# [1 2 3]
print(arr2[0])	# [1 2 3]
print(arr2[0: 2])
""""
[[1 2 3]
 [4 5 6]]
""""

python

3、运算

python中列表加法为拼接，np中为向量加法

arr1 = np.array([1, 2, 3])
arr2 = np.array([4, 5, 6])

print(arr1 + arr2)	# [5 7 9]
print(arr1 * arr2)	# [ 4 10 18]

python

点乘

arr1 = np.array([1, 2, 3])
arr2 = np.array([4, 5, 6])
print(np.dot(arr1, arr2))	# 32

python

A = np.array([[1, 2], [3, 4]])
B = np.array([[5, 6], [7, 8]])
C = np.dot(A, B)

print(C)
'''
[[19 22]
 [43 50]]
'''

python

均值、标准差、求和、最大值、最小值

arr1 = np.array([1, 2, 3])
print(arr1.mean())	# 2.0
print(arr1.std())	# 0.816496580927726
print(arr1.sum())
print(arr1.max())	# 3
print(arr1.min())	# 1

python

排序

np.sort(arr)

python

4、形状操作

arr = np.array([[1, 2, 3],
                [4, 5, 6],
                [7, 8, 9],
                [10, 11, 12]])

print(arr.shape)
new_arr = arr.reshape(2, 6)
print(new_arr)
arr.reshape(-1) # 变为一维行向量

python

转置

print(new_arr)
print(new_arr.transpose())

'''
[[ 1  2  3  4  5  6]
 [ 7  8  9 10 11 12]]
 
 [[ 1  7]
 [ 2  8]
 [ 3  9]
 [ 4 10]
 [ 5 11]
 [ 6 12]]
'''

python

5、元素筛选

arr1 = np.array([[1, 4, 3],
                [2, 5, 6]])

print(arr1 > 3)
print(arr1[arr1 > 3])
'''
[[False  True False]
 [False  True  True]]
[4 5 6]
'''

python

6、导出和导入

A = np.array([[1, 2], [3, 4]])
B = np.array([[5, 6], [7, 8]])
C = np.dot(A, B)

np.save("arr", C) # 导出arr.npy

python

arr = np.load("arr.npy") # 导入
print(arr)

python

数据处理-Pandas#

帮助文档 ↗

1、读取excel(其他数据格式可查看文档)

import pandas as pd

df = pd.read_excel("1.xlsx", "Sheet1", engine="openpyxl")
print(df.head(5))
print(type(df))	# <class 'pandas.core.frame.DataFrame'>

python

将数据转换为pandas格式

data = {'姓名': [1, 2, 3], '成绩': [1, 2, 3]}
data_df = pd.DataFrame(data)
print(data_df)

python

2、查看信息(样本数，数据类型等)

df = pd.read_excel("1.xlsx", "Sheet1", engine="openpyxl")
print(df.info())

python

3、处理数据

缺失值

df = pd.read_excel("1.xlsx", "Sheet1", engine="openpyxl")
df = df.dropna()

python

类型转换

df = pd.read_excel("1.xlsx", "Sheet1", engine="openpyxl")
df['成绩'] = df['成绩'].astype(float)
print(df.info())

python

4、数据选择和过滤

df = pd.read_excel("1.xlsx", "Sheet1", engine="openpyxl")
avg = df['成绩'].mean()
print(df[df['成绩'] >= avg])

python

可按照 $3\sigma$ 原则筛选异常值

可视化-Matplotlib#

帮助文档 ↗

折线图

import numpy as np
import matplotlib.pyplot as plt

x = np.linspace(0, 10, 100)  # 0-10均匀分布
y = np.sin(x)
plt.plot(x, y)
plt.title("y = sin(x)")
plt.xlabel("x")
plt.ylabel("y")
plt.show()

python

散点图

x = np.linspace(0, 10, 10)  # 0-10均匀分布
y = np.sin(x)
plt.scatter(x, y)
plt.title("y = sin(x)")
plt.xlabel("x")
plt.ylabel("y")
plt.show()

python

结合（拟合图像绘制）

import numpy as np
import matplotlib.pyplot as plt
from matplotlib.pylab import mpl # 一下三行代码为显示中文legend
mpl.rcParams['font.sans-serif'] = ['SimHei']
mpl.rcParams['axes.unicode_minus'] = False

x = np.linspace(0, 10, 10)  # 0-10均匀分布
y = np.sin(x)
x2 = np.linspace(0, 10, 100)
y2 = np.sin(x2)
plt.scatter(x, y, marker='*', c='r', label="数据点") # 散点
plt.plot(x2, y2, linestyle='--', label="折线") # 拟合曲线
plt.legend()
plt.title("y = sin(x)")
plt.xlabel("x")
plt.ylabel("y")
plt.show()

python

多图绘制

x = np.linspace(0, 10, 10)  # 0-10均匀分布
y = np.sin(x)
x2 = np.linspace(0, 10, 100)
y2 = np.sin(x2)

fig, axes = plt.subplots(1, 2)  # 一行两列
axes[0].scatter(x, y, marker='*', c='r', label="数据点")  # 散点
axes[0].set_title("数据点")
axes[1].plot(x2, y2, linestyle='--', label="折线")  # 拟合曲线
axes[1].set_title("拟合曲线")
fig.legend()
plt.show()

python

直方图

x = [1, 2, 3]
y = [2, 4, 10]
plt.bar(x, y)
plt.show()

python

常见模型#

评价决策类#

归一化->设置权重->加权评价

层次分析#

建立层次结构模型

自顶向下分为目标层（分析问题的预定目标或理想结果）、准则层（所需考虑的准则）、方案层（可供选择的决策方案）

构造各层次中判断矩阵（ $a_{ij}$ 表示第i个指标对j个指标的重要程度）
一致性检验

import numpy as np 
# 判断矩阵A(主观决定)
A = np.array([[1, 2, 3, 5], 
             [1/2, 1, 1/2, 2],
             [1/3, 2, 1, 2],
             [1/5, 1/2, 1/2, 1]])
# 获取A的行数，索引1为列
n = A.shape[0]
# 求特征值eig_val和特征向量eig_vec
eig_val, eig_vec = np.linalg.eig(A)
# 求特征值最大值
Max_eig = max(eig_val)

CI = (Max_eig - n) / (n - 1)
# RI表
RI = [0, 0.0001, 0.52, 0.89, 1.12, 1.26, 1.36, 1.41, 1.46, 1.49, 1.52, 1.54, 1.56, 1.58, 1.59]

CR = CI / RI[n-1]
if CR < 0.10:
    print("一致性比例为", CR, ',一致性可以接受')
else:
    print("一致性比例为", CR, ',需要进行修改')

python

求权重后进行评价

算术平均法

权重向量为 $W_i = \frac{1}{n} \sum_{j=1}^{n} \frac{a_{ij}}{\sum_{k=1}^{n} a_{kj}} \quad (i = 1, 2, 3, \ldots, n)$

import numpy as np 
# 判断矩阵A(主观决定)
A = np.array([[1, 2, 3, 5], 
             [1/2, 1, 1/2, 2],
             [1/3, 2, 1, 2],
             [1/5, 1/2, 1/2, 1]])
# 计算每列的和(1*n向量)
ASum = np.sum(A, axis=0)
# 获取A行数
n = A.shape[0]
# 归一化，利用广播机制，ASum自动扩展
Stand_A = A / ASum
# 按行求和(n*1向量)
ASumr = np.sum(Stand_A, axis=1)
# 计算权重向量
weights = ASumr / n
print(weights)

python

几何平均法

$w_i = \frac{\left(\prod_{j=1}^{n} a_{ij}\right)^{\frac{1}{n}}}{\sum_{k=1}^{n}\left(\prod_{j=1}^{n} a_{kj}\right)^{\frac{1}{n}}} \quad (i = 1, 2, \ldots, n)$

import numpy as np 
# 判断矩阵A(主观决定)
A = np.array([[1, 2, 3, 5], 
             [1/2, 1, 1/2, 2],
             [1/3, 2, 1, 2],
             [1/5, 1/2, 1/2, 1]])
# 按行求连乘积(n*1向量)
prod_A = np.prod(A, axis=1)
# 获取A行数
n = A.shape[0]
# 将向量每个分量开n次方
prod_n_A = np.power(prod_A, 1/n)
#归一化
re_prod_A = prod_n_A / np.sum(prod_n_A)
print(re_prod_A)

python

特征值法

import numpy as np 
# 判断矩阵A(主观决定)
A = np.array([[1, 2, 3, 5], 
             [1/2, 1, 1/2, 2],
             [1/3, 2, 1, 2],
             [1/5, 1/2, 1/2, 1]])
# 获取A行数
n = A.shape[0]
# 求特征值eig_val和特征向量eig_vec
eig_val, eig_vec = np.linalg.eig(A)
# 求最大特征值的索引
max_index = np.argmax(eig_val)
# 求对应的特征向量
max_vector = eig_vec[:, max_index]
# 归一化
weights = max_vector / np.sum(max_vector)
print(weights)

python

TOPSIS法#

通过最接近理想解且最远离负理想解确定最优选择

import numpy as np

n = int(input("输入参评数目："))
m = int(input("输入指标数目："))

kind = input("输入类型矩阵：1、极大型 2、极小型 3、中间型 4、区间型").split(" ")
# 以下输入参数矩阵
A = np.zeros(shape=(n, m))
for i in range(n):
    A[i] = input(f"输入第{i+1}人各指标").split(" ")
    A[i] = list(map(float, A[i]))
print(f"参数矩阵为：\n{A}")

python

将原始矩阵正向化（越大越好）

# 极小型转极大型
def minTomax(max_x, x):
    x = list(x)
    ans = [[(max_x - e)] for e in x]
    return np.array(ans)

python

# 中间值型转极大型
def midTomax(best_x, x):
    x = list(x)
    h = [abs(e - best_x) for e in x]
    M = max(h)
    if M == 0:
        M = 1
    ans = [[(1-e/M)] for e in h]
    return np.array(ans)

python

def regTomax(low_x, high_x, x):
    x = list(x)
    M = max(low_x-min(x), max(x)-high_x)
    if M == 0:
        M = 1
    ans = []
    for i in range(len(x)):
        if x[i]<low_x:
            ans.append([(1-(low_x-x[i])/M)])
        elif x[i]>high_x:
            ans.append([(1-(x[i]-high_x)/M)])
        else:
            ans.append([1])
    return np.array(ans)

python

for i in range(m):
    if kind[i]=="1":
        v = np.array(A[:, i])
    elif kind[i] == "2":
        maxA = max(A[:, i])
        v = minTomax(maxA, A[:, i])
    elif kind[i] == "3":
        bestA = eval(input("输入最优值："))
        v = midTomax(bestA, A[:, i])
    elif kind[i] == "4":
        lowA = eval(input("输入区间型下界："))
        highA = eval(input("输入区间型上界："))
        v = regTomax(lowA, highA, A[:, i])
    if i == 0:
        X = v.reshape(-1, 1)
    else:
        X = np.hstack([X, v.reshape(-1, 1)])

print(f"正向化后矩阵为：\n{X}")

python

正向矩阵标准化（标准化+权重）

X = X.astype('float')
for j in range(m):
    X[:, j] = X[:, j]/np.sqrt(sum(X[:, j]**2))
print(f"标准化矩阵为：\n{X}")

python

计算得分归一化

x_max = np.max(X, axis=0) # 每列最大值
x_min = np.min(X, axis=0)
x_max_expanded = np.tile(x_max, (n, 1)) # 按行扩展n次
x_min_expanded = np.tile(x_min, (n, 1)) # 按行扩展n次
d_z = np.sqrt(np.sum(np.square(X - x_max_expanded), axis=1)) 
d_f = np.sqrt(np.sum(np.square(X - x_min_expanded), axis=1))
print("每个指标最大值：", x_max)
print("每个指标最小值：", x_min)
print("d+向量：", d_z)
print("d-向量：", d_f)
s = d_f/(d_z + d_f)
Score = 100*s/sum(s)
print(Score)

python