如何将adfuller_test遍历DataFrame列的打印输出保存为PDF或TXT文件

方案1：直接保存为TXT文件（改动最少） #

无需修改原有adfuller_test函数，直接捕获终端打印内容写入TXT，输出和终端显示完全一致：

from contextlib import redirect_stdout

# 所有打印输出会直接写入ADF检验结果.txt
with open('ADF检验结果.txt', 'w', encoding='utf-8') as f:
    with redirect_stdout(f):
        for name, column in train_set.iteritems():
            adfuller_test(column, name=column.name)

方案2：生成PDF文件 #

如果需要PDF格式，先把所有输出捕获到内存，再逐行写入PDF即可：

1. 先安装依赖：pip install fpdf2
2. 运行如下代码：

from contextlib import redirect_stdout
import io
from fpdf import FPDF

# 捕获所有打印输出到内存字符串
output_buffer = io.StringIO()
with redirect_stdout(output_buffer):
    for name, column in train_set.iteritems():
        adfuller_test(column, name=column.name)
all_content = output_buffer.getvalue()
content_lines = all_content.split('\n')

# 生成PDF文件
pdf = FPDF()
pdf.add_page()
pdf.set_font("Arial", size=12)
# 逐行写入内容，行高设为7适配字体大小
for line in content_lines:
    pdf.cell(200, 7, txt=line if line.strip() else ' ', ln=1, align='L')

pdf.output("Augmented Dickey Fuller Test.pdf")

可选优化：修改adfuller_test返回结果字符串 #

如果不想用标准输出捕获的方式，可以修改原函数，直接返回检验报告的字符串，后续处理更灵活：

def adfuller_test(series, signif=0.05, name='', verbose=False):
    """Perform ADFuller to test for Stationarity of given series and return report string"""
    r = adfuller(series, autolag='AIC')
    output = {'test_statistic':round(r[0], 4), 'pvalue':round(r[1], 4), 'n_lags':round(r[2], 4), 'n_obs':r[3]}
    p_value = output['pvalue'] 
    def adjust(val, length= 6): return str(val).ljust(length)
    
    res = []
    res.append(f'    Augmented Dickey-Fuller Test on "{name}"' + "\n   " + '-'*47)
    res.append(f' Null Hypothesis: Data has unit root. Non-Stationary.')
    res.append(f' Significance Level    = {signif}')
    res.append(f' Test Statistic        = {output["test_statistic"]}')
    res.append(f' No. Lags Chosen       = {output["n_lags"]}')
    for key,val in r[4].items():
        res.append(f' Critical value {adjust(key)} = {round(val, 3)}')
    if p_value <= signif:
        res.append(f" => P-Value = {p_value}. Rejecting Null Hypothesis.")
        res.append(f" => Series is Stationary.")
    else:
        res.append(f" => P-Value = {p_value}. Weak evidence to reject the Null Hypothesis.")
        res.append(f" => Series is Non-Stationary.")
    
    return '\n'.join(res)

修改后可以直接在循环中收集所有列的检验结果，再写入TXT或PDF即可。

内容的提问来源于stack exchange，提问作者Lachlan Sweet