Table of Contents

    F1 Score

    Rumman Ansari June 26, 2026 16 views Subject Details
    MACHINE LEARNING

    F1 Score

    The perfect balance between Precision and Recall — one metric to rule them both.

    What is the F1 Score?

    The F1 Score is a Machine Learning evaluation metric that combines Precision and Recall into a single value using their harmonic mean.

    In simple words — F1 Score tells us how well a model balances Precision and Recall.

    Why is the F1 Score Important?

    • Balances Precision and Recall.
    • Especially useful for imbalanced datasets.
    • Helps when both False Positives and False Negatives matter.
    • Provides a single, easy-to-compare number.
    • Used in NLP, healthcare, fraud detection, and more.
    F1 Score is the best metric when you need a balance between Precision and Recall.

    Mathematical Formula

    F1 SCORE FORMULA
    $$ F1 = 2 \times \frac{Precision \times Recall}{Precision + Recall} $$

    Range:

    • 0 = Worst possible score
    • 1 = Perfect Precision and Recall
    F1 uses the harmonic mean, not arithmetic mean — to give more weight to the lower value.

    Why Harmonic Mean Instead of Average?

    If we used a simple average, a model with very high Precision and low Recall could still appear "good". But the harmonic mean ensures that:

    • Both Precision and Recall must be high to get a high F1 score.
    • A weak metric pulls the overall score down.

    Visual Intuition Using Confusion Matrix

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

    F1 Score uses both Precision (TP / (TP + FP)) and Recall (TP / (TP + FN)) to balance them.

    Simple Example

    Suppose a model has:

    • Precision = 80%
    • Recall = 60%
    CALCULATION
    $$ F1 = 2 \times \frac{0.80 \times 0.60}{0.80 + 0.60} = 2 \times \frac{0.48}{1.40} = 0.686 $$
    Insight Even though Precision is 80% and Recall is 60%, the F1 Score is 68.6% — pulled down by the lower value (Recall).

    When Should You Use F1 Score?

    Use F1 Score when:

    • You have an imbalanced dataset.
    • Both False Positives and False Negatives are important.
    • You need a single score combining Precision and Recall.
    • You're comparing multiple models.

    Where is F1 Score Used?

    Medical Diagnosis

    • Balance false alarms & missed cases
    • Improve diagnostic accuracy

    Fraud Detection

    • Detect real fraud cases
    • Avoid blocking valid users

    Spam Detection

    • Filter spam efficiently
    • Don't block valid emails

    NLP / Search Engines

    • Improve precision-recall balance
    • Show relevant results

    Anomaly Detection

    • Catch real anomalies
    • Limit false alerts

    Cybersecurity

    • Detect attacks accurately
    • Maintain user trust

    High vs Low F1 Score

    High F1 Score

    • Balanced Precision & Recall
    • Strong model performance
    • Reliable predictions

    Low F1 Score

    • Imbalanced Precision & Recall
    • Weak performance
    • Many wrong predictions

    Real-Life Analogy

    F1 Score = Teamwork

    If one player on a team is excellent but the other is weak, the team performs poorly. F1 Score works the same way — it ensures Precision and Recall work together, not separately.

    Python Example — Calculating F1 Score

    Prerequisites: Python 3.x, scikit-learn. 
    pip install scikit-learn
    from sklearn.metrics import f1_score

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

f1 = f1_score(y_true, y_pred)
print(f"F1 Score: {f1 * 100:.2f}%")
    Output The F1 Score reflects the balance between Precision and Recall.

    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 f1_score, classification_report

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

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

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

# Predict
y_pred = model.predict(X_test)

# Evaluate
print("F1 Score:", f1_score(y_test, y_pred))
print("\nClassification Report:\n", classification_report(y_test, y_pred))
    Insight A high F1 Score indicates strong, balanced predictions — ideal for healthcare and fraud detection.

    F1 Score in Multi-Class Classification

    • Macro F1 — average of all class F1 scores (equal weight).
    • Weighted F1 — average weighted by class frequency.
    • Micro F1 — calculates F1 globally.
    f1_score(y_test, y_pred, average="macro")
f1_score(y_test, y_pred, average="weighted")
f1_score(y_test, y_pred, average="micro")

    F1 Score vs Accuracy

    Aspect F1 Score Accuracy
    FocusBalance of Precision & RecallAll predictions
    Best ForImbalanced dataBalanced data
    ReducesFP + FN trade-offsTotal errors
    Use CaseFraud, healthcareGeneral classification

    F1 Score vs Precision vs Recall

    Metric Focus Best Use
    PrecisionCorrectness of positive predictionsAvoid False Positives
    RecallCapture of actual positivesAvoid False Negatives
    F1 ScoreBalance of bothBoth errors matter

    Advantages of F1 Score

    • Provides a balanced view of performance.
    • Works well for imbalanced data.
    • Easy to interpret.
    • Helps compare different models fairly.
    • Single metric to optimize.

    Disadvantages

    Limitation 1 Doesn't consider True Negatives.
    Limitation 2 Treats Precision and Recall equally — not always desired.
    Limitation 3 May hide important details about model behavior.
    Limitation 4 Not ideal for probabilistic predictions.

    F-Beta Score (Generalized F1)

    The F-Beta Score allows weighting Recall more or less than Precision.

    F-BETA FORMULA
    $$ F_\beta = (1 + \beta^2) \times \frac{Precision \times Recall}{\beta^2 \times Precision + Recall} $$
    • β = 1 → F1 Score (equal weight)
    • β > 1 → Recall more important
    • β < 1 → Precision more important
    from sklearn.metrics import fbeta_score
fbeta_score(y_true, y_pred, beta=2)

    Common Mistakes to Avoid

    Mistake 1 Comparing F1 across different datasets without context.
    Mistake 2 Optimizing F1 without checking confusion matrix.
    Mistake 3 Using only F1 — ignore the trade-off between Precision and Recall.
    Mistake 4 Not adjusting weights for imbalanced or domain-specific problems.

    Best Practices

    Quick Tips

    • Use F1 when both errors matter.
    • Combine with Precision, Recall, ROC-AUC.
    • Use Macro/Micro/Weighted F1 for multi-class.
    • Use F-Beta for custom trade-offs.
    • Always check confusion matrix.
    • Validate using cross-validation.

    Importance of F1 Score

    Balanced View

    • Combines Precision & Recall
    • Single performance number

    Critical for Imbalanced Data

    • Reveals real performance
    • Better than Accuracy

    Foundation Metric

    • Used in many ML competitions
    • Industry-standard metric

    Business Decisions

    • Improves trust in predictions
    • Supports critical operations

    Golden Rule

    REMEMBER
    Precision ⚖ Recall = F1 Score

    Key Takeaway

    The F1 Score is the perfect balance between Precision and Recall. When your dataset is imbalanced or when both False Positives and False Negatives matter, F1 Score becomes one of the most important metrics. It provides a single, balanced number to evaluate and compare Machine Learning models effectively.

    Machine Learning Course  |  Chapter: Model Evaluation & Validation  |  Topic 4