对Numpy ndarray执行归一化操作触发TypeError错误的技术问询

Hey there, let's break down why you're running into this error and fix it step by step!

The Root Cause #

When you read your CSV data using csv.reader and convert it to a NumPy array, all the values are stored as string types (NumPy calls this a "flexible type" like dtype='<U10'). Numerical operations like calculating means, minima, or peak-to-peak values (ptp()) require numeric data types (int/float), so trying to run these operations on string arrays triggers that TypeError.

Solution 1: Convert to Numeric Type When Creating the Array #

The simplest fix is to tell NumPy to parse the data as numeric values right when you create the array. Replace your line:

data = np.array(x)

with:

data = np.array(x, dtype=np.float64)

This forces all elements to be 64-bit floats, which supports all the math operations you need.

Solution 2: Convert the Feature Array Later #

If you need to keep the original array (e.g., the first column is a classification label you want as a string), you can just convert the feature subset to numeric types:

X_raw = data[:,1:13].astype(np.float64)

Full Corrected Code (With Both Normalization Options) #

Here's your code updated to work, including the standard (X-μ)/σ normalization you originally wanted:

import csv
import numpy as np
import urllib.request

# Download the dataset
url = 'http://archive.ics.uci.edu/ml/machine-learning-databases/wine/wine.data'
urllib.request.urlretrieve(url,'F:/Python/Wine Dataset/wine_data')

# Read and parse the data
filename = 'F:/Python/Wine Dataset/wine_data'
raw_data = open(filename,'rt')
reader = csv.reader(raw_data)
x = list(reader)
# Convert to numeric array immediately
data = np.array(x, dtype=np.float64)

# Split labels and features
y = data[:,0]
X_raw = data[:,1:13]

# Option 1: Min-max normalization (your original simplified attempt)
X_minmax = (X_raw - X_raw.min(0)) / X_raw.ptp(0)
print("Min-max normalized data:\n", X_minmax[:5])  # Print first 5 rows to check

# Option 2: Standardization (X-μ)/σ (your original desired method)
mu = X_raw.mean(0)
sigma = X_raw.std(0)
X_standardized = (X_raw - mu) / sigma
print("\nStandardized data:\n", X_standardized[:5])

Bonus: Use NumPy's Built-in CSV Reader #

For even cleaner code, you can skip csv.reader entirely and use np.genfromtxt, which automatically handles type conversion:

data = np.genfromtxt(filename, delimiter=',')

This reads the CSV directly into a numeric NumPy array, no extra steps needed!

内容的提问来源于stack exchange，提问作者W. Roberts