Demand forecasting and venue ops automation for an indoor slide park — built during a real internship, solving real operational problems.
Built for Slick City Action Park (Slide Park Georgia LLC) as part of a Data Science & AI Automation internship under CEO Mehtab Wasi.
Indoor entertainment venues face highly variable demand — weekends vs weekdays, weather, school calendars, local events. Without a model, they overstaff slow days and understaff busy ones. Every wrong call costs money.
An ensemble forecasting model trained on synthetic-but-realistic venue data:
- Feature engineering: day of week, month, school calendar flags, weather proxy, local events, holiday indicators
- Models: Random Forest + Gradient Boosting ensemble with cross-validation
- Results: R² = 0.91, MAPE ≈ 12% on validation set
- Rolling validation to simulate real-world deployment
Linear programming model that takes a demand forecast as input and outputs the minimum-cost staffing plan:
- Built with Python + PuLP
- Constraints: minimum coverage per hour, max shift length, staff availability
- Outputs optimal shift assignments for each role (cashier, attendant, safety)
- Reduces theoretical overstaffing by ~18% vs heuristic scheduling
Delivered to the operations team for live planning:
| Tab | Content |
|---|---|
| Forecast | Predicted demand by day with confidence bands |
| Schedule | Optimal staff schedule output |
| KPIs | Revenue, utilization, labor cost metrics |
| Actuals | Historical demand data |
| Sensitivity | What-if analysis on demand assumptions |
| Capacity | Slide capacity and throughput model |
| Dashboard | Summary for leadership review |
| Metric | Value |
|---|---|
| Forecast R² | 0.91 |
| MAPE | ~12% |
| Theoretical overstaffing reduction | ~18% |
| Delivery format | Python scripts + Excel workbook |
| Layer | Tech |
|---|---|
| ML | scikit-learn (RandomForest, GradientBoosting) |
| Optimization | PuLP (Linear Programming) |
| Analysis | Python, Pandas, NumPy |
| Reporting | Excel (openpyxl), Matplotlib, Seaborn |
| Platform | ROLLER venue management API |
git clone https://github.com/MuhammadFarid1990/SlidePark-AI
cd SlidePark-AI
pip install -r requirements.txt
# Run the forecasting pipeline
python forecasting/train.py
# Run the scheduling optimizer
python scheduling/optimize.py --forecast forecasting/output/forecast.csvSlidePark-AI/
├── forecasting/
│ ├── train.py # Model training + validation
│ ├── features.py # Feature engineering pipeline
│ ├── predict.py # Inference on new dates
│ └── output/ # Forecast CSVs
├── scheduling/
│ ├── optimize.py # LP scheduling optimizer
│ ├── constraints.py # Staffing constraints
│ └── output/ # Schedule outputs
├── data/
│ └── generate_synthetic.py # Synthetic data generator
├── requirements.txt
└── README.md
Muhammad Farid — MS Business Analytics & AI @ UT Dallas. Data Science & AI Automation Analyst intern @ Slick City Action Park.