UIDAI Aadhaar Demographic Anomaly Detection

Overview

UIDAI manages Aadhaar enrolment at a national scale, making manual identification of regional irregularities impractical.
This project implements an explainable, data-driven anomaly detection system to identify districts and states with abnormal Aadhaar enrolment patterns, enabling targeted audits and improved governance.

---

Problem Statement

To detect unusual demographic distributions and abnormal enrolment volumes in Aadhaar data that may indicate:

• Data quality issues
• Operational inefficiencies
• Potential misuse or irregular practices

The solution is designed to be transparent, interpretable, and policy-ready.

---

Dataset

Source: UIDAI Aadhaar Enrolment & Update Demographic Datasets (CSV)

Key Columns Used

• state
• district
• pincode
• demo_age_5_17
• demo_age_17_

Time-based (date) columns were excluded as the focus is on regional and demographic analysis, not time-series trends.

---

Methodology

1. Data Preparation

• Merged multiple CSV files into a single dataset
• Removed irrelevant columns
• Handled missing values
• Filtered zero-enrolment districts

---

2. District-Level Analysis

• Aggregated data so each row represents one district
• Computed:
  • Total Aadhaar enrolment
  • Age 17+ enrolment ratio

---

3. Anomaly Detection

• Demographic anomalies:
  Z-score applied on age 17+ ratio (|z| > 2)
• Volume anomalies:
  Z-score applied on total enrolment (|z| > 3)
• Final flag:
  A district is flagged if either condition is met

---

4. Severity Ranking

A severity score was calculated to prioritize districts:

5. State-Level Aggregation

• Aggregated district-level anomalies at the state level
• Identified states with a higher proportion of anomalous districts
• Revealed systemic regional risk patterns

---

6. Pincode-Level Drill-Down

• Performed pincode-level analysis for the most anomalous districts
• Localized the exact pincodes contributing to the anomalies
• Enabled precise and actionable, ground-level insights

---

Key Results

• Approximately 1000 districts analyzed
• 29 districts flagged with significant demographic anomalies
• Extreme outliers observed with Z-scores greater than 10
• State-level analysis revealed systemic regional risks
• Pincode-level drill-down enabled targeted and actionable insights

---

Visual Analysis

• Demographic ratio distribution plots
• Demographic vs enrolment anomaly scatter plots
• Visualizations of top anomalous districts and states
• Pincode-level anomaly charts (case study)

---

Tech Stack

• Python
• Pandas, NumPy
• SciPy
• Matplotlib, Seaborn
• Jupyter Notebook

---

Relevance to UIDAI

This solution helps UIDAI to:

• Prioritize audits efficiently
• Detect demographic irregularities at an early stage
• Focus investigations at district and pincode levels
• Improve overall data quality and governance