Food Establishments Inspection Analysis

Analyzed 1.9M+ food safety inspection records across Chicago and Dallas to identify violation patterns, risk factors, and operational insights for public health departments.

Problem Statement: Public health departments struggle to prioritize inspections and allocate resources effectively due to:

• High volume of establishments requiring routine inspections
• Inconsistent violation tracking across jurisdictions
• Lack of visibility into recurring violation patterns
• Manual processes for risk assessment and resource allocation

Business Impact: Inefficient inspection scheduling leads to delayed identification of high-risk establishments, putting public health at risk and wasting limited health department resources.

Proposed Solution: A dimensional data warehouse integrating food inspection data from multiple cities, enabling:

• Standardized violation tracking and analysis
• Risk-based inspection prioritization
• Comparative analysis across jurisdictions
• Automated reporting for health department leadership

Key Findings:

Chicago (267,866 inspections analyzed):

• 27.4% of records missing violation data - data quality issue requiring investigation
• 196,048 high-risk establishments (Risk Category 1 & 2) requiring priority inspections
• 73,445 violations documented across 166K+ unique violation types
• Top violation categories: Food temperature control, sanitation, pest management

Dallas (78,400 inspections analyzed):

• 72.2% missing data across violation fields - systematic data collection issue
• 42 duplicate inspection records identified and removed
• 25 violation categories consolidated into standardized format
• Average inspection score: 85/100 (passing threshold: 70)

Cross-City Insights:

• Violation patterns differ significantly between cities (coding standards inconsistent)
• Chicago has more granular violation tracking (better data quality)
• Dallas uses numeric scoring system (more quantitative assessment)
• Need for standardized national food safety violation taxonomy

Dataset Overview:

Chicago Food Inspections:

• Source: City of Chicago Open Data Portal
• Records: 267,866 inspections (2010-2024)
• Establishments: 32,165 unique food businesses
• Variables: 17 columns (Inspection ID, License, DBA Name, Address, Risk, Results, Violations, etc.)

Dallas Restaurant Inspections:

• Source: Dallas Open Data Portal
• Records: 78,400 inspections (2019-2024)
• Establishments: 9,101 unique restaurants
• Variables: 114 columns (25 violation slots, scores, addresses, etc.)

Data Quality Challenges:

• Missing values (1.8% Chicago, 72.2% Dallas)
• Inconsistent violation formats (free text vs structured codes)
• Duplicate records (Dallas: 42 duplicates)
• Non-standardized risk classifications

Tech Stack:

• ETL: Talend Open Studio (integration), Python (pandas)
• Database: Microsoft SQL Server
• Data Modeling: ER/Studio (dimensional design)
• Visualization: Tableau
• Data Profiling: ydata-profiling, Alteryx

Project Structure

Food-Establishments-Inspection/
├── README.md
├── Datasets/
│   ├── Raw/
│   │   ├── Chicago.tsv                 # Chicago food inspections (268K records)
│   │   └── Dallas.tsv                  # Dallas restaurant inspections (78K records)
│   └── Processed/
│       ├── Chicago - Cleaned.csv       # Normalized violation data
│       └── Dallas - Cleaned.csv        # Standardized format
├── Python/
│   └── Staged.py                       # Python cleaning script
├── ETL/
│   ├── Chicago/                        # Chicago ETL workflows
│   ├── Dallas/                         # Dallas ETL workflows
│   ├── Dimension/                      # Dimension table workflows
│   ├── Facts/                          # Fact table workflows
│   └── ETL_Workflow.docx               # Talend workflow documentation
├── Profiling/
│   ├── Chicago_Profiling.yxmd          # Alteryx data profiling workflow
│   ├── Dallas_Profiling.yxmd           # Alteryx data profiling workflow
│   └── Profiling_Workflows.docx        # Profiling process documentation
├── Model/
│   ├── Dimensional_Model.png           # Star schema diagram
│   └── Model_Script.docx               # DDL documentation
├── Viz/
│   ├── Food Establishments Inspections # Tableau Document
|   ├── Master Script.py                # Python script for Combined dataset
│   └── Master Load .csv                # Dataset(Combined)
└── 

Database Schema: Star Schema Design

Fact Tables (2):

• Fact_Inspection - Core inspection events with scores and results
• Fact_Violation - Individual violations linked to inspections

Dimension Tables (7):

• Dim_Business_License - Establishment details (name, address, license)
• Dim_Date - Date hierarchy (year, quarter, month, week)
• Dim_Facility_Type - Restaurant, grocery, school cafeteria, etc.
• Dim_Inspection_Type - Routine, complaint-driven, license, reinspection
• Dim_Risk_Category - Risk levels (High, Medium, Low)
• Dim_Food_Inspections_Results - Pass, Fail, Pass with Conditions
• Dim_Violation - Violation codes, descriptions, comments

Key Relationships:

Dim_Business_License ──┐
Dim_Date ──────────────┤
Dim_Facility_Type ─────┤
Dim_Inspection_Type ───┼──> Fact_Inspection ──> Fact_Violation
Dim_Risk_Category ─────┤                              │
Dim_Results ───────────┘                              │
                                                      │
Dim_Violation ────────────────────────────────────────┘

Total Records in Warehouse:

• 346,266 inspections
• 1.2M+ violation records (after normalization)
• 40K+ unique establishments

Technical Implementation:

ETL Pipeline:

Phase 1: Data Extraction:

• Downloaded TSV files from city open data portals
• Initial profiling with ydata-profiling and Alteryx

Phase 2: Data Quality Assessment: Identified issues:

• Missing violation data (27% Chicago, 72% Dallas)
• Inconsistent violation formats (text blocks vs structured fields)
• Duplicate records
• Invalid date formats
• Truncated text fields

Phase 3: Data Transformation:

Chicago Processing:

• Normalized violations column (split pipe-delimited text into separate rows)
• Extracted violation code, description, comment using regex
• Added DI metadata (Process ID, Current Date, Workflow name)
• Calculated risk scores and violation points

Dallas Processing:

• Consolidated 25 violation columns into single normalized structure
• Mapped violation points to standardized scale
• Standardized facility types
• Created default risk categories

Phase 4: Data Loading:

• Loaded staged data into SQL Server
• Populated dimensional model using Talend
• Applied surrogate keys and SCD Type 1 logic
• Validated referential integrity

Data Profiling Results:

Chicago Data Quality:

• 17 variables, 267,866 observations
• 84,184 missing cells (1.8%)
• 0 duplicates
• Key issues: Violations column missing in 27% of records

Dallas Data Quality:

• 114 variables, 78,400 observations
• 6.4M missing cells (72.2% - sparse violation matrix)
• 42 duplicate records removed
• Key issues: 99%+ missing in violation detail fields 11-25

SQL Capabilities Demonstrated:

• Dimensional Modeling: Star schema with 7 dimensions, 2 facts
• DDL Design: Foreign keys, constraints, indexes
• Data Validation: Integrity checks, duplicate detection
• ETL Logic: Staging → integration → dimensional model
• Data Quality: Profiling, missing value handling, normalization

Visualizations:

Tableau Dashboards:

1. Executive Overview - Total inspections, pass/fail rates, trend analysis
2. Violation Analysis - Top violations by frequency and severity
3. Geographic Distribution - Heat map of high-risk establishments by neighborhood
4. Comparative Analysis - Chicago vs Dallas metrics side-by-side

For dashboard previews, see Tableau folder

Installation & Reproducibility:

Prerequisites:

• Python 3.8+ (pandas, re, datetime)
• Microsoft SQL Server
• Tableau Desktop (for .twb file) or Tableau Public

Setup Instructions:

1. Clone repository:

git clone https://github.com/lakshmi14k/Food-Establishments-Inspection.git

2. Run Python cleaning script:

cd data/raw
python ../../python/clean_data.py

Output: Chicago - Cleaned.csv, Dallas - Cleaned.csv in data/processed/

3. Create database and load schema:

-- Run in SQL Server Management Studio
CREATE DATABASE FoodInspections;
GO

-- Execute DDL script
USE FoodInspections;
GO
-- Run sql/DDL_Script.sql

4. Load data using ETL tool (Talend):

• Original implementation: Talend Open Studio (workflows in etl/ folder)
• Alternative: Use Python pandas to load CSVs directly to SQL Server

5. Run validation checks:

-- Execute sql/Validation_Script.sql

6. Open Tableau dashboards:

• Connect to SQL Server database OR
• Use processed CSV files from data/processed/

Results:

Data Pipeline:

• 346K+ inspection records integrated
• 1.2M+ violation records normalized
• 40K+ unique establishments profiled
• 9 dimensional/fact tables populated

Insights Delivered:

• Identified systematic data quality issues (72% missing Dallas data)
• Documented top 10 violation categories per city
• Mapped high-risk establishment concentrations
• Enabled trend analysis (inspection frequency, pass rates over time)

Business Value:

• Prioritized inspection scheduling based on risk scores
• Identified establishments requiring follow-up (failed inspections)
• Enabled resource allocation optimization
• Provided comparative benchmarking across cities

Key Technical Achievements

• ETL Pipeline: Multi-source integration (2 cities, different schemas)
• Data Quality: Handled 72% missing data without losing analytical value
• Normalization: Transformed wide format (114 columns) to normalized structure
• Dimensional Modeling: Star schema with proper grain and relationships
• Data Profiling: Comprehensive quality assessment before transformation
• Regex Parsing: Extracted structured data from unstructured text fields

Built with SQL Server, Python, Alteryx, Talend, and Tableau