Table of Contents

Recall

Rumman Ansari June 26, 2026 26 views Subject Details

MACHINE LEARNING

Recall

Measuring how well your Machine Learning model captures all the actual positives.

What is Recall?

Recall (also called Sensitivity or True Positive Rate) is a classification evaluation metric that measures the model's ability to correctly identify all actual positive cases.

In simple words — Recall tells us how many of the actual positives the model successfully detected.

Why is Recall Important?

Measures how well the model finds all true positives.
Critical when missing a positive case is costly.
Used in medical, fraud, and safety-related domains.
Helps detect serious problems like diseases or fraud.
Works with Precision and F1-Score to evaluate model quality.

Use Recall when missing a true positive can have major consequences.

Mathematical Formula

RECALL FORMULA

$$ Recall = \frac{TP}{TP + FN} $$

Where:

TP = True Positives (correctly predicted positives)
FN = False Negatives (positives missed by the model)

Visual Intuition Using Confusion Matrix

	Predicted Positive	Predicted Negative
Actual Positive	TP	FN
Actual Negative	FP	TN

Recall focuses on the "Actual Positive" row — checking how many of them the model successfully detected.

Simple Example

Suppose a cancer detection model is tested on 100 patients with cancer:

Model correctly detects 85 patients (TP)
Model misses 15 patients (FN)

CALCULATION

$$ Recall = \frac{85}{85 + 15} = 0.85 = 85\% $$

Insight The model correctly detected 85% of actual cancer cases — a very important metric in healthcare.

When Should You Use Recall?

Use Recall when False Negatives are dangerous or costly.

Medical Diagnosis

Don't miss real diseases
Save lives

Fraud Detection

Catch every suspicious transaction
Reduce financial loss

Cybersecurity

Detect every attack attempt
Avoid security breaches

Information Retrieval

Find all relevant documents
Avoid missing useful results

Customer Feedback

Detect every negative review
Improve service quality

Anomaly Detection

Catch every unusual event
Prevent failures

Real-Life Analogy

Recall = Fishing Net

Imagine a fisherman with a wide net. If he catches 80 out of 100 fish in the lake, his recall is 80%. Recall measures how many "real positives" the model catches — even if it pulls in a few wrong ones along the way.

High vs Low Recall

High Recall

Model catches most positives
Few cases are missed
Best for life-critical apps

Low Recall

Misses many real positives
Dangerous in healthcare/fraud
Indicates poor sensitivity

Python Example — Calculating Recall

Prerequisites: Python 3.x, scikit-learn.

pip install scikit-learn

from sklearn.metrics import recall_score

y_true = [1, 0, 1, 1, 0, 1, 0, 0, 1, 1]
y_pred = [1, 0, 0, 0, 0, 1, 0, 0, 1, 1]

recall = recall_score(y_true, y_pred)
print(f"Recall: {recall * 100:.2f}%")

Output The Recall value indicates how many actual positives the model correctly detected.

Full Example — Using ML Model

from sklearn.datasets import load_breast_cancer
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import recall_score, classification_report

# Step 1: Load dataset
data = load_breast_cancer()
X, y = data.data, data.target

# Step 2: Split data
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Step 3: Train model
model = RandomForestClassifier()
model.fit(X_train, y_train)

# Step 4: Predict
y_pred = model.predict(X_test)

# Step 5: Evaluate
print("Recall:", recall_score(y_test, y_pred))
print("\nClassification Report:\n", classification_report(y_test, y_pred))

Insight Higher Recall means the model is great at finding all positive cases — essential in medical models.

Recall in Multi-Class Classification

For multi-class problems:

Macro Recall — average recall of each class (equal weight).
Weighted Recall — average recall weighted by class frequency.
Micro Recall — global recall across all classes.

recall_score(y_test, y_pred, average="macro")
recall_score(y_test, y_pred, average="weighted")
recall_score(y_test, y_pred, average="micro")

Recall vs Accuracy

Aspect	Recall	Accuracy
Focus	Actual positives	All predictions
Best For	Imbalanced data	Balanced data
Reduces	False Negatives	Total errors
Use Case	Cancer, fraud, security	General classification

Recall vs Precision

Aspect	Recall	Precision
Focus	How many real positives detected	How many predicted positives correct
Reduces	False Negatives	False Positives
Used When	FNs are dangerous	FPs are dangerous
Example	Cancer detection	Spam detection

Tip: Precision and Recall often have a trade-off. Use F1-Score to balance them.

Real-World Applications

Disease Screening

Detect all real cases
Save lives

Fraud Detection

Catch all suspicious activity
Prevent financial loss

Threat Detection

Identify all security risks
Reduce missed attacks

Search Systems

Find all relevant documents
Improve research

Anomaly Detection

Catch all unusual events
Prevent failures

Customer Support

Identify unhappy customers
Improve retention

Advantages of Recall

Ensures fewer real positives are missed.
Critical in safety-related applications.
Best metric for imbalanced datasets.
Reveals model's ability to handle rare events.
Important for ethical and high-risk decisions.

Disadvantages

Limitation 1 Doesn't consider False Positives.

Limitation 2 May increase incorrect alerts.

Limitation 3 Cannot be used alone — always combine with Precision.

Limitation 4 Higher recall may come at the cost of accuracy.

Common Mistakes to Avoid

Mistake 1 Optimizing Recall alone — leads to too many False Positives.

Mistake 2 Ignoring confusion matrix when interpreting recall.

Mistake 3 Using Recall on balanced data where Accuracy is enough.

Mistake 4 Misunderstanding Recall as overall correctness.

Best Practices

Quick Tips

Pair Recall with Precision and F1-Score.
Always check confusion matrix.
Use Recall in life-critical applications.
Tune model threshold to improve recall.
Use precision-recall curve for analysis.
Interpret in business or domain context.

Importance of Recall

Life-Saving Metric

Detects all real positives
Vital in healthcare

Reduces Critical Errors

Avoids missing fraud
Improves security

Foundation Metric

Key part of F1-Score
Used widely with Precision

Decision Confidence

Builds trust in predictions
Supports safe automation

Golden Rule

REMEMBER

Actual Positives Detected ÷ Total Actual Positives = Recall

Key Takeaway

Recall is a powerful Machine Learning metric that focuses on the model's ability to catch every actual positive. It's especially critical in healthcare, fraud detection, and cybersecurity, where missing a true positive could lead to serious consequences. Combine Recall with Precision and F1-Score for a complete view of your model's performance.