A forensic data analytics project investigating 21 years of NEET/AIPMT examination integrity, paper leak patterns, state-level involvement, and student impact — built with Python, SQL, SQLite, and Power BI.
India's NEET examination is the gateway for 22+ lakh medical aspirants every year. This project forensically analyzes 20+ years of exam integrity data, breach events, and their cascading human impact — from cancelled exams to student suicides — using a full data analytics pipeline.
Key Questions Answered:
- How has exam integrity changed across the CBSE era (2006–2018) vs NTA era (2019–2026)?
- Which states are most frequently implicated in breach events?
- Is there a statistically significant difference in candidate impact between disrupted and clean exam years?
- What does the correlation between breach severity and infrastructure gaps reveal?
| Metric | Value |
|---|---|
| Years Analyzed | 21 (2006–2026) |
| Total Candidates Affected | 6M+ across breach events |
| Confirmed Clean Exams | 10 / 21 years |
| Arrests Documented | 470+ across all breach events |
| Statistical Tests Run | 7 (with p-values) |
| SQL Forensic Queries | 12 |
- Breach severity spikes have become more frequent post-2020 (NTA era)
- Chi-square test confirms integrity statuses are not equally distributed (p < 0.05)
- Conducting body comparison reveals CBSE era had significantly fewer major disruptions
- 2024 and 2026 both resulted in full or partial exam cancellations — first back-to-back occurrence in 21 years
| Tool | Purpose |
|---|---|
| Python (pandas, numpy, scipy) | Data loading, cleaning, statistical analysis |
| Matplotlib / Seaborn | Publication-quality visualizations |
| SQLite3 | Local forensic database |
| SQL | 12 analytical queries with rolling averages, JOINs, window functions |
| Power BI | Interactive dashboard |
| HTML / Chart.js | Standalone browser dashboard (no install needed) |
neet-exam-integrity-analysis/
│
├── data/ # Source CSVs (verified public data)
│ ├── exam_timeline.csv # 21 years of NEET integrity records
│ ├── breach_events.csv # Detailed breach event log
│ ├── student_impact.csv # Suicide and mental health data
│ ├── state_infrastructure.csv # State-level investigation involvement
│ └── exam_benchmarking.csv # CBSE vs NTA comparison
│
├── python/
│ ├── analysis.py # Full 5-module analytics pipeline
│ ├── setup_and_analyze.py # One-command automation script
│ ├── requirements.txt # Python dependencies
│ └── output/ # Generated charts and JSON exports
│
├── sql/
│ ├── create_database.sql # Schema and views
│ └── forensic_queries.sql # 12 validated analytical queries
│
├── tests/
│ └── test_cases.md # 25 validation test cases
│
├── NEET_Dashboard.pbix # Power BI dashboard file
├── index.html # Standalone interactive dashboard
└── README.md
# Just open in any browser:
open index.html# 1. Clone the repo
git clone https://github.com/wildtigress/neet-exam-integrity-analysis.git
cd neet-exam-integrity-analysis
# 2. Install dependencies
pip install -r python/requirements.txt
# 3. Run the full pipeline (creates DB + charts + stats)
python python/setup_and_analyze.pyThis will:
- Create
neet_forensic.dbfrom CSVs - Run all 12 SQL forensic queries and export results to CSV
- Run 7 statistical tests and save p-values
- Generate 6 charts in
python/output/ - Export
dashboard_data.json
Open NEET_Dashboard.pbix in Power BI Desktop.
| Chart | Description |
|---|---|
01_integrity_timeline.png |
21-year integrity score with 3-year rolling average |
02_appeared_vs_affected.png |
Candidate participation vs affected population |
03_suicide_trend.png |
Student mental health impact over time |
04_state_investigations.png |
State-wise investigation involvement heatmap |
05_correlation_heatmap.png |
Cross-variable correlation matrix |
06_cbse_vs_nta.png |
Disruption severity: CBSE vs NTA era boxplot |
7 tests were run with explicit null hypotheses and p-values:
- Chi-Square Goodness-of-Fit — Are integrity statuses equally distributed?
- Mann-Whitney U — Candidate counts in disrupted vs clean years
- Spearman Rank Correlation — Investigation involvement vs infrastructure gaps
- Fisher's Exact Test — Breach type independence
- Kruskal-Wallis — Severity across different states
- Wilcoxon Signed-Rank — Before/after NTA transition
- Point-Biserial Correlation — Breach binary vs affected count
Results saved to: python/output/statistical_test_results.csv
12 forensic queries including:
- Rolling 3-year integrity score average (window functions)
- Affected population as % of total appeared
- State investigation involvement ranking
- Financial impact estimates (with
⚠️ estimates clearly flagged) - CBSE vs NTA era comparison JOIN
All data compiled from verified public sources:
- NTA official notifications and press releases
- CBI FIRs (2019, 2024, 2026)
- Supreme Court orders (AIPMT 2015, NEET 2024)
- NCRB Accidental Deaths & Suicides in India reports
- Parliamentary Standing Committee Report, Dec 2025
- State police FIRs (Tamil Nadu CB-CID, Rajasthan SOG, Jaipur Police)
- News sources: The Hindu, Indian Express, NDTV, Careers360
Note: Financial figures (₹ estimates) are clearly marked as estimates. No causal claims are made from correlational analysis.
Source dataset (Google Sheets):
https://docs.google.com/spreadsheets/d/1KIo8H_Q26Y-pj8iiiGOeMrOLhxeeKK1iLBEGf-dRSjw/
25 test cases documented in tests/test_cases.md covering:
- Data integrity (year completeness, no duplicates, positive counts)
- Statistical validity (p-value ranges, sample size adequacy)
- Dashboard accuracy (KPI match, chart data match)
- SQL query validation (JOIN integrity, NULL handling)
- Ethical methodology (no causal claims, estimates labeled, neutral language)
Python · Pandas · NumPy · SciPy · SQLite · SQL (Window Functions) · Matplotlib · Seaborn · Power BI · Data Cleaning · Statistical Testing · Outlier Detection · Data Storytelling · HTML/CSS · Chart.js
Samiksha Barnwal
BCA Graduate | Aspiring Data Analyst & AI Engineer
099samiksha@gmail.com
LinkedIn · GitHub
This project is for educational and portfolio purposes. Data is compiled from public sources. See individual source citations in data/breach_events.csv and sql/forensic_queries.sql.