如何读取data.txt为numpy.ndarray或转换字符串数组用于降维分析

I'll split your problem into two key tasks and walk you through each with straightforward, efficient solutions tailored for your dimensionality reduction workflow.

1. Read data.txt into a numpy.ndarray #

Since your data is structured with each line as a matrix row and space-separated values, numpy.loadtxt() is the perfect tool here—it's fast, simple, and directly returns the array format you need.

Example code:

import numpy as np

# Load the data directly into a numpy array
data = np.loadtxt('data.txt')

# Verify the shape and type (optional but recommended)
print(f"Data shape: {data.shape}")
print(f"Data dtype: {data.dtype}")

If your file has a header row (string labels you want to skip), just add the skiprows parameter:

data = np.loadtxt('data.txt', skiprows=1)

For larger datasets or files with potential missing values, you can use np.genfromtxt() instead (it handles missing values by default as np.nan):

data = np.genfromtxt('data.txt', dtype=np.float64)

2. Convert String-type numpy.array to Numeric Format #

To quickly convert a string array to a numeric type (like float or int), use the astype() method—it's the most efficient way when you're certain all strings can be safely converted to your target type.

Example code for float conversion:

# Sample string array
str_array = np.array(['1.5', '3.2', '4.7', '6.1'])

# Convert to float64 (standard for most DR methods)
num_array = str_array.astype(np.float64)

If your string array has whitespace or minor formatting quirks, astype() can still handle it in most cases:

str_array_with_spaces = np.array([' 2.3 ', '5.8', ' 7.0 '])
num_array = str_array_with_spaces.astype(np.float64)

Handling Invalid Values #

If there's a chance some strings can't be converted (e.g., non-numeric characters), use np.genfromtxt() to gracefully replace invalid entries with np.nan:

str_array_with_invalid = np.array(['1.2', 'abc', '3.4'])
num_array = np.genfromtxt(str_array_with_invalid, dtype=np.float64)
# Result will be [1.2, nan, 3.4]

Once your data is in a numeric numpy.ndarray, you're ready to feed it into PCA, ICA, t-SNE, LEM, or any other dimensionality reduction library (like scikit-learn) for your comparison experiments.

内容的提问来源于stack exchange，提问作者Fei