Mutual information for feature selection in ML Python - Model Metrics & Evaluation

Metrics & Evaluation - Mutual information for feature selection

Which metric matters for this concept and WHY

Mutual information measures how much knowing one thing helps you know another. For feature selection, it tells us how much a feature and the target share information. The higher the mutual information, the more useful the feature is for predicting the target. This helps pick features that really matter and ignore noise.

Confusion matrix or equivalent visualization (ASCII)

Mutual information is not based on a confusion matrix but on probabilities. Imagine a table showing how often each feature value pairs with each target value:

    | Feature Value | Target=0 Count | Target=1 Count |
    |---------------|----------------|----------------|
    |       A       |       30       |       10       |
    |       B       |       20       |       40       |
    

Mutual information uses these counts to calculate how much knowing the feature reduces uncertainty about the target.

Precision vs Recall (or equivalent tradeoff) with concrete examples

Mutual information helps decide which features to keep. A tradeoff is between keeping many features (high recall of useful info) and keeping only the best (high precision of relevant features).

For example, if you keep too many features with low mutual information, your model may be slow and confused by noise (low precision). If you keep too few, you might miss important signals (low recall).

Balancing this tradeoff means selecting features with mutual information above a threshold that keeps most useful info but removes noise.

What "good" vs "bad" metric values look like for this use case

Good mutual information values are higher numbers showing strong connection between feature and target. For example, a mutual information of 0.5 or above (on a scale from 0 to 1) means the feature shares a lot of info with the target.

Bad values are close to 0, meaning the feature gives almost no useful info about the target. Such features can be dropped safely.

Remember, mutual information is always >= 0. Zero means no relationship.

Metrics pitfalls (accuracy paradox, data leakage, overfitting indicators)

• Ignoring feature redundancy: Two features can both have high mutual information but carry the same info. Selecting both adds no benefit.
• Data leakage: If the feature leaks future info about the target, mutual information will be high but model will fail in real use.
• Overfitting: Selecting features based on mutual information from the test set can cause overfitting. Always compute on training data only.
• Ignoring feature interactions: Mutual information looks at one feature at a time. Some features may be weak alone but strong together.

Your model has 98% accuracy but 12% recall on fraud. Is it good?

No, it is not good for fraud detection. Even though accuracy is high, the model misses 88% of fraud cases (low recall). For fraud, catching as many frauds as possible is critical, so recall matters more than accuracy.

This shows why choosing the right metric matters. High accuracy can be misleading if the data is imbalanced or the goal is to catch rare events.

Key Result

Mutual information quantifies how much a feature tells us about the target, guiding effective feature selection by highlighting informative features.

Practice

1. What does mutual information measure in feature selection?

easy

A. The amount of shared information between a feature and the target variable

B. The correlation coefficient between two features

C. The difference between feature means

D. The number of missing values in a feature

Solution

1. Step 1: Understand mutual information concept

   Mutual information measures how much knowing one variable reduces uncertainty about another.

2. Step 2: Apply to feature selection context

   In feature selection, it measures how much information a feature shares with the target variable.

3. Final Answer:

   The amount of shared information between a feature and the target variable -> Option A

4. Quick Check:

   Mutual information = shared info [OK]

Hint: Mutual info = shared info between feature and target [OK]

Common Mistakes:

• Confusing mutual information with correlation
• Thinking it measures missing data
• Assuming it measures difference in means

2. Which Python function is used to compute mutual information for classification tasks?

easy

A. mutual_info_classif

B. mutual_info_regression

C. mutual_info_score

D. mutual_info_classifier

Solution

1. Step 1: Recall mutual information functions in sklearn

   For classification, sklearn provides mutual_info_classif.

2. Step 2: Differentiate from regression function

   mutual_info_regression is for regression, not classification.

3. Final Answer:

   mutual_info_classif -> Option A

4. Quick Check:

   Classification uses mutual_info_classif [OK]

Hint: Classification uses mutual_info_classif function [OK]

Common Mistakes:

• Using mutual_info_regression for classification
• Confusing function names
• Assuming mutual_info_score exists in sklearn

3. Given this code snippet, what is the output?

from sklearn.feature_selection import mutual_info_classif
import numpy as np
X = np.array([[1, 2], [2, 3], [3, 4], [4, 5]])
y = np.array([0, 1, 0, 1])
mi = mutual_info_classif(X, y, discrete_features=[True, True])
print(np.round(mi, 2))

medium

A. [0.0 0.0]

B. [0.69 0.0]

C. [0.0 0.69]

D. [0.69 0.69]

Solution

1. Step 1: Understand input data and parameters

   X has two discrete features, y is binary. Using mutual_info_classif with discrete_features=True for both.

2. Step 2: Calculate mutual information values

   Both features vary similarly with y, so both have similar mutual information around 0.69 (close to ln(2)).

3. Final Answer:

   [0.69 0.69] -> Option D

4. Quick Check:

   Both features share info with y ~0.69 [OK]

Hint: Discrete features with binary target give ~0.69 MI if informative [OK]

Common Mistakes:

• Assuming zero mutual information for all features
• Mixing up discrete_features parameter
• Rounding errors in output

4. Identify the error in this code snippet for mutual information feature selection:

from sklearn.feature_selection import mutual_info_classif
X = [[1, 2], [2, 3], [3, 4]]
y = [0, 1, 0]
mi = mutual_info_classif(X, y)
print(mi)

medium

A. y should be a 2D array, not 1D

B. X should be a numpy array, not a list of lists

C. mutual_info_classif requires discrete_features parameter

D. mutual_info_classif cannot handle integer data

Solution

1. Step 1: Check input data types

   mutual_info_classif expects numpy arrays or similar, not plain Python lists.

2. Step 2: Identify error cause

   Passing list of lists for X can cause unexpected behavior or errors; converting to numpy array fixes this.

3. Final Answer:

   X should be a numpy array, not a list of lists -> Option B

4. Quick Check:

   Use numpy arrays for X [OK]

Hint: Always convert input data to numpy arrays before sklearn functions [OK]

Common Mistakes:

• Thinking y must be 2D
• Assuming discrete_features is always required
• Believing mutual_info_classif rejects integer data

5. You have a dataset with 10 features. After computing mutual information scores, you find two features have the highest scores but are highly correlated with each other. What is the best approach to select features?

hard

A. Select both features because they have the highest mutual information

B. Select features randomly to avoid bias

C. Select only one of the two correlated features with the highest mutual information

D. Discard both features to avoid redundancy

Solution

1. Step 1: Understand mutual information and correlation

   High mutual information means features are informative, but high correlation means redundancy.

2. Step 2: Choose features to reduce redundancy

   To avoid redundant information, select only one of the correlated features with the highest mutual information.

3. Final Answer:

   Select only one of the two correlated features with the highest mutual information -> Option C

4. Quick Check:

   Pick one correlated feature with highest MI [OK]

Hint: Avoid redundant features by picking one with highest MI [OK]

Common Mistakes:

• Selecting both correlated features causing redundancy
• Discarding informative features unnecessarily
• Choosing features randomly without criteria