FX
FutuX Agri-consult ML Deduplication

Whitebox-Dedup ML Playground

Auditable piecewise MARS deduplication and Inductive Conformal Prediction (ICP) statutory circuit breaker. Upload any tabular CSV to detect entity duplicates in 100% client-side execution.

Complete Privacy Guarantee: Zero network uploads. All candidate blocking, tokenization, Jaro-Winkler distances, and MARS spline arithmetic run locally in your browser.
Use Cases & What Happens if Your CSV Has No Name Column
Show Guide

What if your dataset doesn't have a name column?
The engine does not fail. It automatically detects all columns in your CSV and selects the primary text column. You can also pick any target column from the dropdown (e.g. company_name, customer_id, title, address, phone).

Recommended Starter Columns beneficiary_id, full_name, phone_number, state, latitude, longitude.
Tier 1: Auto-Merge ($P \ge 0.898$) Statistically verified duplicates (e.g., typos, swapped names) merged automatically with auditable justification.
Tier 2: Circuit Breaker ($0.491 \le P < 0.898$) Conformal prediction set is $\{0, 1\}$ (Ambiguous). Safely pauses and flags for human audit to prevent wrongful exclusion.
Tier 3: Auto-Distinct ($P < 0.491$) Mathematically confirmed distinct individuals. Retained as valid separate records.

Upload Tabular Dataset or Run Sample Roster

The engine automatically identifies target columns, creates candidate blocks, and scores pairs.

Click to upload or drag & drop ANY CSV file
Works with any column schema (farmers, customers, patients, voter rolls, transactions)