Statistical monitoring of dataset distributions using PSI, chi-square, KL divergence, and mean/variance shift detection — with plain-English explanations per feature.
Seven complementary metrics across numeric and categorical columns give a complete picture of distribution health — with explainability built in.
Population Stability Index with 10-bin histograms. Thresholds: <0.1 stable, 0.1–0.25 warning, >0.25 high drift.
Captures structural changes in central tendency and spread. Validated with T-test and F-test scores.
Measures asymmetric information divergence between reference and current probability distributions.
Detects distribution shift in categorical columns using chi-square test — works on string/label data.
Ranks features by drift score. Automatically surfaces the top drifting columns for investigation.
Feature-wise text explanations combining PSI severity, T-test significance, and variance instability signals.
Aggregates all metrics into a single LOW / MEDIUM / HIGH system risk indicator with colour-coded alerts.
Three steps from raw CSV to actionable drift insights.
Upload reference (baseline) and current CSVs. Numeric and categorical columns detected automatically.
FastAPI backend computes PSI, chi-square, T-test, F-test server-side and returns structured JSON.
Distribution overlays, control charts, feature contribution bars, and explainability summaries.