如何在 PyPlot 中的分类箱线图上绘制拟合曲线？ 为什么结果与 Google Sheets 中的同一个图不同？

Question

我有以下 csv 数据：

Dataset Size,MAPE,MAE,STD MAPE,STD MAE
35000,0.0715392337,23.38300578,0.9078698348,2.80407539
26250,0.06893431034,22.34732326,0.9833948236,1.926517044
17500,0.0756695622,26.0900766,0.6055443674,8.842862631
8750,0.07176532526,23.02646184,0.8284005282,2.190506033
4200,0.08661127364,29.89234607,0.9395831421,7.587818412
2100,0.08072315267,27.20110884,0.03956974712,4.948606892
1050,0.07505202908,27.04025924,0.841966778,4.550482956
700,0.07703248113,26.17923045,0.4468447145,1.523638508
350,0.08695408769,32.35331585,0.7891190087,4.18648457
200,0.09770903032,30.96197823,0.04648972591,3.892800694
170,0.1202382169,41.87828814,0.7257680584,6.70453713
150,0.1960949784,77.20321559,0.5661066006,21.57418682

根据上述数据，我想使用 matplotlib 或类似（seaborn、pandas 等）生成以下图：

from pathlib import Path
from matplotlib import animation
import seaborn as sns
import matplotlib.pyplot as plt
import pandas as pd
import numpy as np
from scipy.optimize import curve_fit

nr_datapoints = 10
def exponenial_func(x, a, b, c):
    return a*np.exp(-b*x)+c
def myplot(data_file):
    df = pd.read_csv(data_file)
    print(df.head())

    fig, ax = plt.subplots()

    # Exponential line fit
    popt, pcov = curve_fit(exponenial_func, np.array([float(i) for i in range(len(df['Dataset Size']))]), df['MAPE'], p0=(0, 0.0145, 0.0823))
    xp = np.linspace(0,len(df['Dataset Size']), 100)  
    plt.plot(xp, exponenial_func(xp, *popt), color = 'g')
    # barplote with error bars
    ax.bar([str(s) for s in df['Dataset Size']], df['MAPE'], yerr=df['STD MAPE'])
    plt.title('Accuracy of Model vs. Dataset Size')
    plt.xlabel('Dataset Size')
    plt.ylabel('Mean Absolute Percentage Error')
    fig.tight_layout()
    plt.show()

我得到的情节如下：

尽管对数据拟合了指数函数，为什么我的代码最终得到一条线而不是一条曲线？ （鉴于谷歌表格图做同样的事情，例如拟合数据的指数曲线）

Answer 1

尝试了一些函数，我想我可以肯定地说，Google Sheets 指数函数的形式与此接近：

def sheetey_exponential_function(x, a, b, c):
    return a * b ** (x + c)

Answer 2

问题是水平轴不是线性的。 实际上它是逆线性的。 所以，如果你想你适合看起来像一个指数函数，你需要更换x与1/x ：

def exponenial_func(x, a, b, c):
    return a*np.exp(-b/x)+c

结果如下：