tech@manager.com<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\nHere is the issue or problem in this table : Manager’s information is repeated for every employee in the same department and if the manager’s detail change, you must update multiple rows.<\/p>\n\n\n\n
After Normalization <\/strong><\/p>\n\n\n\nWe are splitting the data into two table. <\/p>\n\n\n\n
\nEmployee Table<\/strong><\/li>\n<\/ul>\n\n\n\nEmployee ID <\/strong><\/td>Name<\/strong> <\/td>Department ID<\/strong><\/td><\/tr>1<\/td> Emp A<\/td> 1001<\/td><\/tr> 2<\/td> Emp B<\/td> 1001<\/td><\/tr> 3<\/td> Emp C<\/td> 1002<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n\nDepartment Table<\/strong><\/li>\n<\/ul>\n\n\n\nDept. ID<\/strong><\/td>Dept.<\/strong><\/td>Manager<\/strong><\/td>Email ID<\/strong><\/td><\/tr>1001<\/td> Sales <\/td> A<\/td> sales@manager.com<\/td><\/tr> 1002<\/td> Technical<\/td> B<\/td> tech@manager.com<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\nNo Redundancy<\/strong> : <\/strong>Manager information gets stored once after normalization.<\/p>\n\n\n\nEasy Update : <\/strong>Changing the manager’s information only require adding one row.<\/p>\n\n\n\nExample 2 : Data Normalization in Machine Learning<\/strong><\/p>\n\n\n\nWhile working with ML models, features may have different scales. For instance: Original Data <\/p>\n\n\n\nAge<\/strong><\/td>Income(s)<\/strong><\/td><\/tr>30<\/td> 50,000<\/td><\/tr> 60<\/td> 100,000<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\nHere is the income range that is much larger than age. If left unnormalized, models may focus more on income than age.<\/p>\n\n\n\n
Skewing predictions: <\/p>\n\n\n\n
After Normalization (Scaling Data to Range 0 – 1)<\/strong><\/p>\n\n\n\nAge(Scaled)<\/strong><\/td>Income(Scaled)<\/strong><\/td><\/tr>0.0<\/td> 0.0<\/td><\/tr> 1.0<\/td> 1.0<\/td><\/tr><\/tbody><\/table>Normalized Table<\/strong><\/figcaption><\/figure>\n\n\n\nFor Normalized Data : step by step calculation here <\/p>\n\n\n\n
Normalization Formula : xscaled <\/sub>= (x – min (x) ) \/ ( max (x) – min (x) )<\/p>\n\n\n\n<\/span>Data Normalization Techniques<\/strong><\/span><\/h1>\n\n\n\nData normalization in machine learning involves scaling or transformation data so that all features have comparable ranges or distributions. This helps models work more effectively , especially when features carries different scales.<\/p>\n\n\n\n
Here are the most common data normalization techniques that are explained below :<\/p>\n\n\n\n
<\/span>Min Max Scaling (Feature Scaling)<\/strong><\/span><\/h2>\n\n\n\nWhat it does :<\/strong><\/p>\n\n\n\n\nScales the data to fit within a specific range, typically between 0 and 1.<\/li>\n\n\n\n It preserves the relative relationships between values by proportionally adjusting them accordingly.<\/li>\n<\/ul>\n\n\n\nNormalization Formula : xscaled <\/sub>= (x – min (x) ) \/ ( max (x) – min (x) )<\/p>\n\n\n\nIn this formula : x is the original value, min(x) is the minimum value in the dataset and max(x) is the maximum value in the dataset.<\/p>\n\n\n\n
<\/span>Standardization Scaling ( Z-Score )<\/strong><\/span><\/h2>\n\n\n\nWhat it does :<\/strong><\/p>\n\n\n\n\nTransform the data to have a mean of 0 and standard deviation of 1.<\/li>\n\n\n\n It makes the data follow a standard normal distribution.<\/li>\n<\/ul>\n\n\n\nFormula :
x is the original value.<\/p>\n\n\n\n
\u03bc is the mean of the dataset.<\/p>\n\n\n\n
\u03c3 is the standard deviation of the dataset.<\/p>\n\n\n\n
Key Characteristics:<\/strong><\/p>\n\n\n\nRange<\/strong>: The transformed values are not bounded (they can range from negative infinity to positive infinity).<\/p>\n\n\n\nEffect on Outliers<\/strong>: More robust than Min-Max for datasets with outliers, but extreme values may still have some influence.<\/p>\n\n\n\n<\/td> Min Max Scaling<\/strong><\/td>Standardization Scaling<\/strong><\/td><\/tr><\/td> There is no strict assumptions made here about data distribution <\/td> Data is normally distributed among standardization scaling<\/td><\/tr> <\/td> Neural networks , K-Nearest Neighbor<\/td> Principal Component Analysis, Support Vector Machines, and Logistic Regression<\/td><\/tr> <\/td> Min-Max scaling is sensitive to outliers<\/td> Standardization scaling handles outliers better<\/td><\/tr> <\/td> [0,1] (fixed range)<\/td> Mean=0 and standard deviation =1 <\/td><\/tr><\/tbody><\/table>Comparison table between min max scaling and standardization scaling<\/strong><\/figcaption><\/figure>\n\n\n\n
![\"data](\"https:\/\/www.iquanta.in\/blog\/wp-content\/uploads\/2024\/12\/DN-in-ML-1024x576.jpeg\")
<\/figure><\/div>\n\n\n![\"\"](\"https:\/\/www.iquanta.in\/blog\/wp-content\/uploads\/2025\/01\/dawabanner-6792322e75d5d-4-2-1024x159.webp\")
<\/a><\/figure><\/div>\n\n\n