Information Value is NOT a Widely Used Metric in Industry

While IV is a popular concept for feature selection in academic settings and some data science competitions, it's not commonly used in industry for a few key reasons:

What Industry Uses Instead:

• Feature Importance: Most industry practices rely on techniques like: Random Forest Feature Importance: This measures the decrease in model performance when a feature is removed. Gradient Boosting Feature Importance: Similar to Random Forest but based on the contribution of each feature to the gradient boosting process. Lasso/Ridge Regression Coefficients: The absolute value of coefficients in regularized linear models can indicate feature importance.
• Feature Engineering: Instead of just relying on IV, industry focuses on creating features that capture more complex relationships and interactions.
• Data Exploration & Visualization: Industry professionals heavily use data visualization and domain expertise to understand the data and identify key features.
• Model Validation & Evaluation: Emphasis is placed on evaluating model performance on unseen data using techniques like cross-validation and metrics like AUC, precision, recall, and F1-score.

The Takeaway:

While IV can be a helpful tool for initial feature screening, it's not a definitive solution for real-world data science. Industry relies on a combination of more robust feature selection techniques, feature engineering, and thorough model evaluation to make informed decisions about which features to include in their models.

Thanks Touhidul Alam Seyam

Titas De Welcome.