Feature selection using variance threshold in Pandas

Last update on December 21 2024 07:41:22 (UTC/GMT +8 hours)

Pandas: Machine Learning Integration Exercise-12 with Solution

Write a Pandas program to select feature selection using variance threshold.

This exercise demonstrates how to select features based on their variance using Scikit-learn's VarianceThreshold.

Sample Solution :

Code :

import pandas as pd
from sklearn.feature_selection import VarianceThreshold

# Load the dataset
df = pd.read_csv('data.csv')

# Select only the numeric columns for feature selection
numeric_cols = df.select_dtypes(include=[float, int])

# Initialize the VarianceThreshold with a threshold of 0.1
selector = VarianceThreshold(threshold=0.1)

# Apply feature selection based on variance
X_selected = selector.fit_transform(numeric_cols)

# Output the selected features
print(X_selected)

Output:

[[1.0e+00 2.5e+01 5.0e+04 0.0e+00]
 [2.0e+00 3.0e+01 6.0e+04 1.0e+00]
 [3.0e+00 2.2e+01 7.0e+04 0.0e+00]
 [4.0e+00 3.5e+01 8.0e+04 1.0e+00]
 [5.0e+00     nan 5.5e+04 0.0e+00]
 [6.0e+00 2.9e+01     nan 1.0e+00]]

Explanation:

• Import Libraries:
• pandas is imported for handling data in DataFrame format.
• VarianceThreshold from Scikit-learn is imported for performing feature selection based on variance.
• Load the Dataset:
• The dataset data.csv is loaded using pd.read_csv() and stored in the DataFrame df.
• Select Numeric Columns:
• select_dtypes(include=[float, int]) is used to select only the numeric columns from the dataset (e.g., Age, Salary) and exclude non-numeric columns like Name and Gender.
• Initialize VarianceThreshold:
• VarianceThreshold is initialized with a threshold of 0.1. Features with variance lower than this threshold will be removed.
• Apply VarianceThreshold:
• fit_transform() is applied to the numeric columns to perform feature selection, keeping only the features that have a variance greater than 0.1.
• Output the Selected Features:
• The resulting selected features are printed after the variance-based filtering.

Python-Pandas Code Editor:

Have another way to solve this solution? Contribute your code (and comments) through Disqus.