@datascibykashi: Before training ML models, we often need to scale features so that no variable dominates others. One popular method is Normalization. ⚖️ 🔹 What is Normalization? 👉 Normalization rescales data to a fixed range, usually [0, 1] (sometimes [-1, 1]). It doesn’t change the distribution shape, just the scale of the values. 🔹 Why do we need Normalization? Consider a dataset with: • Age: 20 – 70 • Income: 15,000 – 200,000 ➡️ Without normalization, models like KNN or Gradient Descent will give more weight to Income just because numbers are bigger. ✅ Normalization ensures all features contribute equally. 🔹 How does it work? Formula (Min-Max Normalization): X’ = \frac{X - X_{min}}{X_{max} - X_{min}} Where: • X = original value • X_{min}, X_{max} = min & max of the feature Result → Value between 0 and 1. 🔹 Example: 📊 Suppose exam scores = [50, 70, 90] • Min = 50, Max = 90 ➡️ Normalized 50 = (50-50)/(90-50) = 0 ➡️ Normalized 70 = (70-50)/(40) = 0.5 ➡️ Normalized 90 = (90-50)/(40) = 1 Now all scores are in the same scale (0–1). 🔹 When to Use Normalization? ✅ Best when data does not follow a Gaussian (normal) distribution. ✅ Used in algorithms like: • Neural Networks (fast convergence) • K-Nearest Neighbors (distance-based) • Gradient Descent 🔹 Difference between Normalization & Standardization • Normalization → Rescales data to a range (0–1) • Standardization → Centers data around 0 mean & unit variance ✅ Takeaway: Normalization makes all features fit within the same scale so that ML models treat them equally and training becomes more stable. #MachineLearning #Normalization #DataScience #Preprocessing #100daysofml

Data Scientist | Kashi

Open In TikTok:

Region: PK

Wednesday 27 August 2025 04:08:01 GMT