univariate-analysis

Univariate analysis computes IV and WoE for feature screening by applying optimal binning to your tabular data against a 0/1 binary target, generating bin-level WoE tables and keep/drop suggestions to shortlist predictive variables.

Univariate distribution analysis checks data quality by computing binning-based distribution tables, basic stats, and flags for high missing rates or low cardinality to detect problematic features before model training.

Cross distribution analysis generates a binned cross table for exactly two features, revealing their joint distribution to help understand variable interactions during exploratory data analysis.

Univariate analysis is constrained to numeric features for IV computation and requires a 0/1 binary target. Non-numeric data and multi-class targets are not supported for predictive power calculations.

Yes, the analysis leverages optbinning, pandas, and numpy to support both quantile and distance binning methods, handling special values while generating bin-level tables for exploratory data analysis.

You need univariate analysis for pre-model feature screening to rapidly inspect feature behavior, detect problematic columns, and compute IV-based predictive power to shortlist variables for finance-style supervised learning.