🏆 Fabric Data Days — Notebook Submission
✨ Overview
This project demonstrates an end-to-end pipeline that evaluates and predicts the City of Tomorrow using a four-pillar framework: Sustainability, Infrastructure, Safety, and Livability. I have considered only urban_planning_dataset.csv as my only Data source. The notebook includes data loading, cleaning, exploratory data analysis (EDA), feature engineering, custom metric derivation, model training, evaluation, visualization, and an interactive ipywidgets application that uses a saved model from Models/*.pkl.
🔎 What I built
- Interactive Application: An ipywidgets-based UI that ingests user inputs and returns pillar predictions using the trained model (saved as a .pkl file).
- EDA & Statistical Analysis: Thorough EDA with distribution plots, correlation analysis, box/violin plots, density maps, and statistical summaries to understand feature behavior by land use and pillar.
- Custom Metrics: Derived domain-specific metrics (e.g., resilience-adjusted sustainability, transport equity index) to capture nuanced urban attributes beyond raw features.
- Interpretation & Visuals: All key findings are supported by clear visuals and explanatory insight blocks for each pillar.
- Conclusion & Recommendations: A concise, evidence-backed summary with policy- and design-oriented recommendations for future-ready cities.
🧪 Models & Evaluation
Below are the model names and evaluations:
Overal Average Accuracy: 97.63% (R2)
1. Sustainability Model (RandomForest): 94.43% (R2)
2. Infrastructure Model (RandomForest): 98.66% (R2)
3. Safety Model (RandomForest): 99.91% (R2)
4. Livability Model (RandomForest): 97.52% (R2)
📂 Artifacts & Notebook
The full notebook and artifacts are included with this submission. Couldn't upload the .pkl file to github but can be found in the zip attachment with this submission.
Open the notebook here: Nov2025_Jai_Rathinavel_CitiesOfTomorrow.ipynb
Application Demo GIF: https://github.com/jai-rathinavel/Fabric-Notebooks/blob/main/Fabric-Data-Days/Demo-App-Urban.GIF
The model file(s) used by the ipywidgets app are stored in the Models/ folder inside the notebook workspace (e.g., Models/final_model.pkl).
📌 Key Findings & Conclusion
- Renewable energy adoption and green space strongly correlate with higher sustainability scores indicating ecological planning drives the sustainability pillar.
- Industrial zones are the primary contributors to carbon footprint; targeted interventions and cleaner processes are recommended.
- Transport infrastructure shows generally balanced coverage; where gaps exist, targeted public-transit investments reduce inequity and congestion.
- Income vs crime correlations are weak policy and social interventions likely have outsized impact on safety compared to income alone.
- Disaster risk is a major negative driver of livability and overall sustainability resilience planning is essential for vulnerable residential zones.
- Custom metrics such as the Transport Equity Index and Resilience-Adjusted Sustainability helped surface non-obvious trade-offs and prioritize interventions.
- The interactive app demonstrates how stakeholders can explore “what-if” scenarios by adjusting feature sliders and immediately seeing pillar outcomes.
- Overall, cities that balance all four pillars perform best sustainable, resilient, inclusive, and livable urban design is the path to the City of Tomorrow.
Proud to be a Super User! |  |
