pd.to_numeric转换报错ValueError：无法解析‘notpresent’求解决方案

Hey there, let's break down what's going wrong and fix that ValueError you're seeing.

What's Causing the Error? #

The problem is straightforward: your bgr column contains non-numeric strings like "notpresent" (complete with extra double quotes!), and pd.to_numeric() can't parse these into floating-point numbers by default. That's exactly why you're getting the Unable to parse string error.

Step-by-Step Fix #

Let's walk through the fixes one by one:

1. Strip Unwanted Double Quotes
   First, we need to remove those extra double quotes from all string columns—they're blocking proper value conversion:

# Remove double quotes from any object-type columns
df = df.apply(lambda col: col.str.replace('"', '') if col.dtype == 'object' else col)

2. Convert Columns Gracefully
   Use the errors='coerce' argument with pd.to_numeric()—this will turn any unparseable strings into NaN instead of crashing your code:

# Convert columns to float, turning bad values into NaN
df['bgr'] = pd.to_numeric(df['bgr'], errors='coerce', downcast='float')
df['wc'] = pd.to_numeric(df['wc'], errors='coerce', downcast='float')
df['rc'] = pd.to_numeric(df['rc'], errors='coerce', downcast='float')

3. Handle Missing Values (Optional but Important)
   Now you'll have NaN values in your dataframe. You can either drop rows with missing values or fill them with a statistic like the mean/median (filling is usually better for retaining data):

# Option 1: Drop rows with any NaN values
df.dropna(axis=0, how='any', inplace=True)

# Option 2: Fill NaNs with column means
# df.fillna(df.mean(numeric_only=True), inplace=True)

Full Fixed Code #

Here's your complete code with all the fixes applied:

import math
import pandas as pd
import matplotlib.pyplot as plt
import matplotlib
from sklearn import preprocessing

# Load the dataset
df = pd.read_html('https://github.com/authman/DAT210x/blob/master/Module4/Datasets/kidney_disease.csv')[0]

# Drop the id column
df.drop('id', axis=1, inplace=True)

# Keep only the specified columns
df = df[['bgr', 'wc', 'rc']]

# Clean up double quotes from string columns
df = df.apply(lambda col: col.str.replace('"', '') if col.dtype == 'object' else col)

# Convert columns to numeric, coercing bad values to NaN
df['bgr'] = pd.to_numeric(df['bgr'], errors='coerce', downcast='float')
df['wc'] = pd.to_numeric(df['wc'], errors='coerce', downcast='float')
df['rc'] = pd.to_numeric(df['rc'], errors='coerce', downcast='float')

# Handle missing values (choose one option)
df.dropna(axis=0, how='any', inplace=True)
# df.fillna(df.mean(numeric_only=True), inplace=True)

# Verify the results
print(df.dtypes)
print(df.head())

Quick Notes #

• errors='coerce' is the game-changer here—it prevents the function from throwing errors when it hits non-numeric strings.
• Cleaning the quotes first is critical because "notpresent" (with quotes) is treated as a different string than notpresent (without), and the former won't be recognized as a missing value indicator otherwise.
• Choosing to fill missing values instead of dropping rows will help you keep more data for analysis—just make sure the fill method makes sense for your specific dataset!

内容的提问来源于stack exchange，提问作者Hussain