Table of Contents

Accuracy

Rumman Ansari June 26, 2026 31 views Subject Details

MACHINE LEARNING

Accuracy

The simplest and most popular metric to measure how often a Machine Learning model predicts correctly.

What is Accuracy?

Accuracy is one of the most commonly used classification evaluation metrics. It measures the percentage of correct predictions made by the model out of the total predictions.

In simple words — Accuracy tells us how often the model got the prediction right.

Why is Accuracy Important?

Provides a quick overview of model performance.
Easy to calculate and interpret.
Useful for balanced classification problems.
Helps compare different ML models.
Often used as a baseline metric.

Accuracy works best when your dataset has nearly equal classes.

Mathematical Formula

ACCURACY FORMULA

$$ Accuracy = \frac{\text{Correct Predictions}}{\text{Total Predictions}} \times 100 $$

Using the Confusion Matrix terminology:

USING CONFUSION MATRIX

$$ Accuracy = \frac{TP + TN}{TP + TN + FP + FN} $$

Where:

TP = True Positives (correctly predicted positives)
TN = True Negatives (correctly predicted negatives)
FP = False Positives (wrongly predicted as positives)
FN = False Negatives (wrongly predicted as negatives)

Simple Example

Suppose a spam detection model is tested on 100 emails:

Correctly classified emails = 90
Wrong predictions = 10

CALCULATION

$$ Accuracy = \frac{90}{100} \times 100 = 90\% $$

Insight The model correctly predicted 90 out of 100 emails — an accuracy of 90%.

Visual Intuition

	Predicted Positive	Predicted Negative
Actual Positive	TP (Correct)	FN (Wrong)
Actual Negative	FP (Wrong)	TN (Correct)

Accuracy adds the green cells (TP + TN) and divides by the total.

Python Example — Calculating Accuracy

Prerequisites: Python 3.x, scikit-learn.

pip install scikit-learn

from sklearn.metrics import accuracy_score

# Actual labels
y_true = [1, 0, 1, 1, 0, 1, 0, 0, 1, 1]

# Predicted labels
y_pred = [1, 0, 1, 0, 0, 1, 0, 1, 1, 1]

# Calculate accuracy
accuracy = accuracy_score(y_true, y_pred)
print(f"Accuracy: {accuracy * 100:.2f}%")

Output The model's accuracy is calculated directly from actual and predicted values.

Full Example — Using ML Model

from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score

# Step 1: Load dataset
data = load_iris()
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: Calculate accuracy
print(f"Accuracy: {accuracy_score(y_test, y_pred) * 100:.2f}%")

Output Random Forest typically achieves over 95% accuracy on the Iris dataset.

Where Accuracy is Used

Spam Detection

Classify spam vs ham emails
Performance check

Image Classification

Cat vs Dog
Object detection accuracy

Text Classification

Sentiment analysis
News categorization

Healthcare

Disease prediction
Diagnosis support

Banking

Loan approval
Customer classification

E-Commerce

Product recommendation
Customer segmentation

When Accuracy Can Be Misleading

Accuracy fails when the dataset is imbalanced.

Example: Suppose 95% of emails are not spam, and the model predicts everything as "not spam".

Accuracy = 95% (looks great!)
But it never detects spam at all.

Limitation Accuracy can hide poor performance on minority classes.

Better Metrics Use Precision, Recall, F1-Score, ROC-AUC for imbalanced datasets.

Accuracy vs Other Metrics

Metric	Best For	Weakness
Accuracy	Balanced data	Fails on imbalanced data
Precision	When false positives matter	Ignores false negatives
Recall	When false negatives matter	Ignores false positives
F1-Score	Balance precision & recall	Slightly complex
ROC-AUC	Overall classifier quality	Less intuitive

Real-Life Analogy

Accuracy = Exam Score

If a student answers 80 out of 100 questions correctly, their accuracy is 80%. ML models are scored the same way — but accuracy alone doesn't show how well they understood the harder topics (minority classes).

Advantages of Accuracy

Easy to understand and explain.
Simple to calculate.
Suitable for balanced datasets.
Provides a quick model comparison.
Widely supported by all ML libraries.

Disadvantages

Limitation 1 Misleading for imbalanced datasets.

Limitation 2 Doesn't show type of errors (FP or FN).

Limitation 3 Doesn't reflect business impact.

Limitation 4 Cannot evaluate probabilistic predictions.

Common Mistakes to Avoid

Mistake 1 Using only accuracy on imbalanced datasets.

Mistake 2 Comparing models without checking confusion matrix.

Mistake 3 Ignoring business cost of FP and FN.

Mistake 4 Assuming higher accuracy = better model.

Best Practices

Quick Tips

Combine accuracy with Precision, Recall, F1.
Always check confusion matrix.
Don't rely on accuracy alone for medical/financial models.
Use cross-validation for stable evaluation.
Test on unseen data.
Analyze errors before deployment.

Importance of Accuracy

Easy Insight

Quick performance check
First metric to evaluate

Model Comparison

Helps benchmark models
Quick decisions

Foundation Metric

Base of many other metrics
Easy to interpret

Business Communication

Easy to share with stakeholders
Used in reports & dashboards

Golden Rule

REMEMBER

Right Predictions ÷ Total Predictions = Accuracy

Key Takeaway

Accuracy is the simplest and most widely used metric to evaluate Machine Learning models. While it works great for balanced datasets, it can be misleading when the data is imbalanced. Always pair it with metrics like Precision, Recall, F1-Score, and ROC-AUC to get a complete picture of your model's performance.