Scipy.optimize：最小化函数产生错误

Question

我正在尝试使用 优化预测参数scipy.optimize。我按照教程进行操作，还在 stackoverflow 上找到了一些很好的示例，但我遇到了一个我无法解决的问题。我开始怀疑使用 pandas 是否是 scipy 的糟糕选择？我已将我的代码设置如下：import simpyimport numpy as npimport matplotlib.pyplot as pltfrom scipy.stats import normimport pandas as pdimport statistics as statimport math as m#from sklearn.grid_search import ParameterGridfrom scipy.optimize import minimize###dataframe for the simulationdf = pd.read_csv('simulation_df_data_2018_2.csv')with pd.option_context("max_rows", None,"max_columns", None):    print(df.head())for i in df.index:    alpha = 0.2    beta = 0.3        x = np.array([alpha, beta])     def holts(x):        LO = np.int(df['average_demand'].loc[i])        print(type(LO))        TO = ((df['m2'].loc[i] - df['m3'].loc[i]) + (df['m1'].loc[i] - df['m2'].loc[i])) / 2        L1 = round(x[0] * df['m3'].loc[i] + (1 - x[0]) * (                LO + TO))        T1 = x[1] * (L1 - LO) + (1 - x[1]) * TO        L2 = round(x[0] * df['m2'].loc[i] + (1 - x[0]) * (                L1 + T1))        T2 = x[1] * (L2 - L1) + (1 - x[1]) * T1        L3 = round(x[0] * df['m1'].loc[i] + (1 - x[0]) * (                L2 + T2))        T3 = beta * (L3 - L2) + (1 - beta) * T2        LT1 = round(L3 + 1 * T3)        MSE = ((df['m3'].loc[i] - L1) + (df['m2'].loc[i] - L2) + (                df['m2'].loc[i] - L3)) ** 2 / 3        return MSE    #print(holts(x))    x0 = [0.1,0.1]    result = minimize(holts, x0, bounds=[(0,1),(0,1)], method="SLSQP")    print(result)    print(x)

Answer 1

因为我不知道你的代码到底在做什么，所以我无法判断下面的代码是否正确。但只是为了指导如何使用scipy.optimize.minimize函数，它应该类似于以下内容：

import numpy as np

import pandas as pd

from scipy.optimize import minimize

df = pd.read_csv('simulation_df_data_2018_2.csv')

for i in df.index:

    # Since you said "alpha" and "beta" are the variable of "holts" function

    # I changed all "alpha" and "beta" to "x[0]" and "x[1]" respectively.

    def holts(x):

        LO = np.int(df['average_demand'].loc[i])

        TO = ((df['m2'].loc[i] - df['m3'].loc[i]) + (df['m1'].loc[i] - df['m2'].loc[i])) / 2

        L1 = x[0] * df['m3'].loc[i] + (1 - x[0]) * (LO + TO)

        T1 = x[1] * (L1 - LO) + (1 - x[1]) * TO

        L2 = x[0] * df['m2'].loc[i] + (1 - x[0]) * (L1 + T1)

        T2 = x[1] * (L2 - L1) + (1 - x[1]) * T1

        L3 = x[0] * df['m1'].loc[i] + (1 - x[0]) * (L2 + T2)

        MSE = ((df['m3'].loc[i] - L1) + (df['m2'].loc[i] - L2) + (df['m2'].loc[i] - L3)) ** 2 / 3

        return MSE

    x0 = np.array([0.1, 0.1])  # initial guess for the points "x" which "holts(x)" attains its minimum.

    result = minimize(holts, x0, bounds=[(0, 1), (0, 1)], method="SLSQP", options={'disp': True})

    print(result)

    print()

哪个打印

Optimization terminated successfully    (Exit mode 0)

            Current function value: 16.921875002444228

            Iterations: 5

            Function evaluations: 15

            Gradient evaluations: 5

     fun: 16.921875002444228

     jac: array([ 8.06808472e-04, -6.23427415e+00])

 message: 'Optimization terminated successfully'

    nfev: 15

     nit: 5

    njev: 5

  status: 0

 success: True

       x: array([0.75000606, 1.        ])

Optimization terminated successfully    (Exit mode 0)

            Current function value: 3.0792324286388828e-09

            Iterations: 7

            Function evaluations: 24

            Gradient evaluations: 7

     fun: 3.0792324286388828e-09

     jac: array([-0.00833083, -0.00081578])

 message: 'Optimization terminated successfully'

    nfev: 24

     nit: 7

    njev: 7

  status: 0

 success: True

       x: array([0.1495232 , 0.25665903])

Optimization terminated successfully    (Exit mode 0)

            Current function value: 1.9951658230345873e-10

            Iterations: 9

            Function evaluations: 32

            Gradient evaluations: 9

     fun: 1.9951658230345873e-10

     jac: array([0.03791677, 0.00858221])

 message: 'Optimization terminated successfully'

    nfev: 32

     nit: 9

    njev: 9

  status: 0

 success: True

       x: array([0.37155192, 0.19218713])

Optimization terminated successfully    (Exit mode 0)

            Current function value: 114.53090716252409

            Iterations: 11

            Function evaluations: 37

            Gradient evaluations: 11

     fun: 114.53090716252409

     jac: array([ 3.06129456e-04, -3.04457455e+01])

 message: 'Optimization terminated successfully'

    nfev: 37

     nit: 11

    njev: 11

  status: 0

 success: True

       x: array([0.7171298, 1.       ])

Optimization terminated successfully    (Exit mode 0)

            Current function value: 33.333333333333336

            Iterations: 2

            Function evaluations: 6

            Gradient evaluations: 2

     fun: 33.333333333333336

     jac: array([-403.33331966,   -0.        ])

 message: 'Optimization terminated successfully'

    nfev: 6

     nit: 2

    njev: 2

  status: 0

 success: True

       x: array([1., 1.])

更新：似乎函数是函数不起作用的round原因，因为函数会在函数图上创建许多高原。我删除了功能并更新了打印的结果。minimizeroundholtsround