Border Intelligence — Common Operating Picture

Business Analysis Dashboard · CBP Border Crossing Data · Built for GDIT IHS Division Context

---

Overview

A full-stack business intelligence project analyzing 273,391 records of U.S. border crossing data spanning April 1996 – January 2026 across 114 active ports of entry. Built to demonstrate end-to-end business analysis capability — from raw government data ingestion to executive-ready deliverables.

The dashboard is modeled after GDIT's Common Operating Picture framework used in Intelligence and Homeland Security programs, combining a Power BI-style executive layer with a Tableau-style analyst drill-down and an ML-driven port risk register.

Live dashboard → nithink-pixel.github.io/border-intelligence-dashboard

---

Dashboard Panels

Panel	Layer	Description
Executive View	Power BI style	KPI cards, monthly trend, port rankings, YoY comparison
Analyst Drill-Down	Tableau style	Port breakdown, mode share, COVID impact, anomaly table
Port Risk Clustering	ML layer	KMeans clustering, scatter plot, full risk register
BA Findings Memo	Deliverable	3 findings, 2 risks, 1 tiered recommendation

---

Key Findings

Finding 1 — San Ysidro is the sole Critical Hub 274.4M crossings (2020–2025) — nearly 2× the next highest port (El Paso, 147.2M). Isolated as the only "Critical Hub" port among 114 analyzed via KMeans clustering. Volatility index of 0.995 indicates consistent, high-density throughput.

Finding 2 — US-Canada border dropped 19.2% in 2025 US-Canada crossings fell from 74.6M (2024) to 60.3M (2025) while US-Mexico held flat at ~266M. Statistically significant divergence requiring root-cause investigation before staffing decisions are made at northern ports.

Finding 3 — 10 ports flagged with anomalous spikes Z-score analysis flagged 10 ports exceeding z > 2.0, with 8 of 10 clustering in July–August 2024 on the US-Canada border — a temporal concentration that exceeds normal summer seasonality.

---

Technical Stack

Data Processing (Python)

# Core pipeline
pandas          # ETL, aggregation, feature engineering
scikit-learn    # KMeans clustering (k=4), StandardScaler normalization
numpy           # Z-score anomaly detection per port

Clustering Methodology

• Algorithm: KMeans (k=4, n_init=10, random_state=42)
• Features: total_volume, avg_monthly_volume, volatility_index, measure_type_diversity
• Normalization: StandardScaler (zero mean, unit variance)
• Anomaly Detection: Per-port z-score on monthly volume (threshold: z > 2.0)

Frontend

• Vanilla HTML/CSS/JS — zero framework dependencies
• Chart.js 4.4.0 for all visualizations
• IBM Plex font family
• Fully responsive — works on any device

---

Dataset

Attribute	Value
Source	Bureau of Transportation Statistics (BTS)
Portal	data.gov — CBP Border Crossing Entry Data
Records	273,391 rows
Date range	April 1996 – January 2026
Borders	US-Mexico · US-Canada
Ports	114 active ports of entry
Measures	Personal Vehicles, Pedestrians, Trucks, Buses, Trains + passengers
License	Public domain (U.S. Government work)

---

Methodology

Raw CBP Data (273,391 rows)
        │
        ▼
Python ETL Pipeline
├── Date parsing & normalization
├── Port-level feature engineering
│   ├── total_volume
│   ├── avg_monthly_volume
│   ├── std_monthly (volatility)
│   └── measure_type_diversity
        │
        ▼
KMeans Clustering (k=4)
├── StandardScaler normalization
├── Cluster assignment → risk tier labeling
└── Output: Critical Hub / High Volume / Moderate / Low Activity
        │
        ▼
Z-Score Anomaly Detection
├── Per-port monthly baseline (mean, std)
├── Z-score computation per port-month
└── Flag: z > 2.0 → anomaly event
        │
        ▼
Dashboard (HTML/Chart.js)
├── Executive View — Power BI layer
├── Analyst Drill-Down — Tableau layer
├── Risk Clustering — ML layer
└── BA Findings Memo — deliverable layer

---

Business Recommendations

Based on the analysis, a 4-tier resource allocation model is recommended:

Tier	Classification	Ports	Resource Strategy
1	Critical Hub	1	AI monitoring · Dedicated analyst · Real-time alerts
2	High Volume	11	Automated anomaly alerts · Monthly review cycles
3	Moderate	45	Quarterly review · Surge capacity on-call
4	Low Activity	57	Annual review · Automated reporting only

---

Project Structure

border-intelligence-dashboard/
│
├── index.html              # Full dashboard (self-contained)
├── README.md               # This file
│
└── analysis/               # Python scripts (local, not deployed)
    ├── etl_pipeline.py     # Data cleaning & feature engineering
    ├── clustering.py       # KMeans port risk classification
    └── anomaly_detect.py   # Z-score flagging per port

---

About

Built by Nithin Krishna as part of a business analysis portfolio project.

• MS Business Analytics — UMass Isenberg School of Management (May 2027)
• Skills demonstrated: SQL · Python · Power BI · Tableau · Business Analysis · KMeans Clustering · Z-score Anomaly Detection · Executive Reporting
• LinkedIn: linkedin.com/in/nithin-krishna145
• GitHub: github.com/nithink-pixel

---

Data source: Bureau of Transportation Statistics · U.S. Customs and Border Protection · data.gov · Public domain