Park Wait Forecasting Repository

Business Context

Amusement parks face two operational challenges every day:

• Highly variable attraction queue times by hour.
• Large swings in daily attendance driven by seasonality, weather, and calendar effects.

Without reliable forecasts, operations teams react too late, staffing is less efficient, guest experience degrades, and revenue opportunities are missed.

Business Goal

Build a forecasting system that predicts:

• Next 7 days of park attendance (daily).
• Next 7 days of attraction waiting times (hourly) per ride.

These forecasts are used to support operational planning and power dashboard/UI modules for decision-making.

Business Value

• Better staffing and resource planning for peak/off-peak periods.
• Lower queue pressure by anticipating high-load attractions.
• Better guest satisfaction through more reliable wait-time information.
• Faster planning cycles with automated forecast generation.

Solution Overview

The repository includes:

• Data preparation pipeline from historical attendance, waiting times, and weather.
• Attendance forecasting model (LightGBM-based).
• Attraction-level waiting-time forecasting models (XGBoost-based).
• Inference pipeline that combines weather + attendance + attraction models.
• Frontend apps for visualization and communication of forecast outputs.

Key Dependencies

• NumPy & pandas: Core tabular processing and feature engineering for forecasting workflows.
• scikit-learn, XGBoost, and LightGBM: Main machine learning stack for attendance and wait-time modeling.
• statsmodels & pmdarima: Time-series analysis and ARIMA/SARIMAX experimentation.
• FastAPI & Uvicorn: Backend/API-serving stack for exposing model outputs.
• Lovable (lovable-tagger): UI scaffolding workflow used to quickly generate and iterate demo interfaces.
• React + TypeScript + Vite: Runtime frontend stack used across dashboard/UI modules.
• Tailwind CSS + Radix UI (shadcn/ui): Component system and styling foundation used by the generated frontend code.
• Recharts & Leaflet: Runtime visualization libraries for charts and park map rendering.

Repository Structure

• modeling/: data, notebooks, forecasting code, and trained artifacts.
• modeling/wait_time_forecasting/: production Python pipeline modules and CLIs.
• modeling/data/raw/: input datasets (attendance.csv, waiting_times.csv, weather_data.csv, etc.).
• modeling/data/processed/: generated intermediate outputs (including weather_forecasted_data.csv and wait-times.csv).
• modeling/data/forecasts/: per-attraction forecast CSVs.
• modeling/artifacts/models/: saved trained model bundles.
• park-wait-radar/: wait-time UI.
• Average-Wait-Time-UI-Module/: average wait-time UI.
• Attendance_and_Revenue_UI_Module/: attendance/revenue UI.

Prerequisites

• Python 3.12
• Node.js 20+
• npm

Forecasting Workflow

1) Generate weather forecast input

python modeling/wait_time_forecasting/weather_forecast.py

This creates:

• modeling/data/processed/weather_forecasted_data.csv

2) Train attendance model (if needed)

python -m modeling.wait_time_forecasting.attendance_cli --mode train

Saved model:

• modeling/artifacts/models/attendace_gradient_boosting.joblib

3) Train waiting-time models (if needed)

Example for selected attractions:

python -m modeling.wait_time_forecasting.waiting_time_cli --mode train --attractions "Bumper Cars,Dizzy Dropper,Free Fall"

4) Run waiting-time forecast

python -m modeling.wait_time_forecasting.waiting_time_cli --mode forecast --attractions "Bumper Cars,Dizzy Dropper,Free Fall" --horizon-days 7

This produces:

• Per-attraction files in modeling/data/forecasts/ named like Attraction_Name_waiting_dd_mm_yyyy.csv.
• Combined UI feed at modeling/data/processed/wait-times.csv.

Optional: Train and Forecast in One Command

python -m modeling.wait_time_forecasting.waiting_time_cli --mode train_forecast --attractions "Bumper Cars,Dizzy Dropper,Free Fall" --horizon-days 7

Frontend Apps

Install and run each UI separately.

Park Wait Radar

cd park-wait-radar
npm install
npm run dev -- --host 0.0.0.0 --port 5173

Open: http://localhost:5173

Average Wait Time UI

cd Average-Wait-Time-UI-Module
npm install
npm run dev -- --host 0.0.0.0 --port 5174

Open: http://localhost:5174

Attendance and Revenue UI

cd Attendance_and_Revenue_UI_Module
npm install
npm run dev -- --host 0.0.0.0 --port 5175

Open: http://localhost:5175

Feeding Wait-Time Data to park-wait-radar

If needed, copy the generated combined forecast file into the app data path:

Copy-Item modeling/data/processed/wait-times.csv park-wait-radar/src/data/wait-times.csv -Force

Notebooks

Notebook workflows are under:

• modeling/notebooks/01_research/
• modeling/notebooks/02_validation/
• modeling/notebooks/03_train/
• modeling/notebooks/04_inference/

Notes

• Raw data location is expected at modeling/data/raw/.
• Waiting-time model files are attraction-specific and saved under modeling/artifacts/models/.
• wait-times.csv in modeling/data/processed/ is the main combined output for UI consumption.