如何在循环中解析鸢尾花txt文件创建指定列的Numpy数组？

Hey there! Let's work through why your current code isn't working and fix it up properly.

First, let's break down the problems in your original code: #

• You're overwriting feature_table every loop: Instead of accumulating rows, you create a new array each time, so you'll only ever end up with the last line's data.
• Incorrect numpy array initialization: np.array([currentline[0]],[currentline[3]],[currentline[4]]) is not valid syntax—numpy expects a single nested structure for rows/columns, not separate lists.
• No type conversion: The values from split(",") are strings, but your first two features should be numerical (float/int) for any future analysis.
• Unsafe file handling: You're opening the file but not explicitly closing it—using a with statement is cleaner and safer.
• Inconsistent module naming: You imported numpy as np but tried to use numpy.array in commented lines, which would throw a NameError.

Here's the corrected code using np.append as you requested: #

import numpy as np

# Initialize an empty 2D array to hold our data (dtype=object lets us mix numbers and strings)
feature_table = np.empty((0, 3), dtype=object)

# Use 'with' to safely open/close the file
with open("iris_data.txt", "r") as iris_data:
    for line in iris_data:
        # Strip newline characters and split the line by commas
        currentline = line.strip().split(",")
        
        # Skip any empty lines in the file
        if not currentline:
            continue
            
        # Convert numerical features to float, clean up the category name
        sepal_length = float(currentline[0])
        petal_width = float(currentline[3])
        iris_name = currentline[4].strip()  # Remove any leftover whitespace/newlines
        
        # Create a 2D row array (matches the shape of feature_table for appending)
        current_row = np.array([[sepal_length, petal_width, iris_name]])
        
        # Append the row to our feature table along the row axis (axis=0)
        feature_table = np.append(feature_table, current_row, axis=0)

# Check the result
print(feature_table)
print(feature_table.shape)

A more efficient alternative (avoiding looped np.append): #

Looping with np.append can be slow for large datasets because numpy reallocates memory every time you append. A better approach is to collect all rows in a Python list first, then convert to a numpy array once:

import numpy as np

# Start with an empty Python list to collect rows
data_rows = []

with open("iris_data.txt", "r") as iris_data:
    for line in iris_data:
        currentline = line.strip().split(",")
        if not currentline:
            continue
            
        sepal_length = float(currentline[0])
        petal_width = float(currentline[3])
        iris_name = currentline[4].strip()
        
        data_rows.append([sepal_length, petal_width, iris_name])

# Convert the list to a numpy array once
feature_table = np.array(data_rows, dtype=object)

print(feature_table)
print(feature_table.shape)

This will give you the structured array you need, with the first two columns as numerical features and the third as the iris category name.

内容的提问来源于stack exchange，提问作者Emre Unsal