matanzig/Disney-Theme-Park-Queue-Dynamics

🎢 Disney World Queue Dynamics

A Comprehensive EDA & Strategic Analysis

Author: Matan Zigelman • University: Reichman University • Date: March 2026

---

📋 Project Introduction: Disney Theme Park Queue Dynamics

This project analyzes a numeric-heavy operational dataset from a major Disney theme park, sourced from kaggle and containing approximately 3,757,301 records. The dataset is primarily driven by time-based and operational metrics ( e.g., TOTAL_DAILY_ATTENDANCE , CAPACITY , GUEST_CARRIED , and DOWNTIME ), making it ideal for statistical modeling while allowing for the easy removal of non-essential categorical columns.

Research Question: How can queue dynamics and operational data be leveraged to optimize time management for park visitors, and how can management use these same patterns to maximize operational efficiency and prevent bottlenecks?

Target Variable: The primary target variable to predict is WAIT_TIME_MAX (the maximum wait time in minutes per hour). This continuous variable serves as the ultimate benchmark for measuring both customer friction (visitor time lost) and park operational efficiency.

---

🧹 Data Preprocessing & Cleaning

Before conducting the exploratory analysis, the raw data (spanning over 3.5 million records) underwent a rigorous preprocessing pipeline to ensure statistical accuracy and integrity:

• Data Integration: The data was originally scattered across multiple sources. We dynamically loaded and merged three primary datasets (waiting_times, attendance, and weather_data) into a single unified master dataframe using Left Joins based on specific WORK_DATE and DEB_TIME_HOUR keys.
• Cleaning & Imputation: We eliminated completely identical duplicate rows and dropped records missing the primary target variable (WAIT_TIME_MAX). For missing numerical features (such as weather metrics), we imputed the gaps using the column median to preserve the natural distribution without skewing the data. Standardized text formatting and Datetime objects were also applied.
• Outlier Handling: We utilized the Interquartile Range (IQR) method to identify and remove extreme upper-bound outliers and negative values (likely system errors). Crucially, zero-minute wait times were explicitly preserved, as they accurately reflect "walk-on" scenarios during early park opening or late closing.
• Initial Correlation Findings: Early descriptive statistics revealed a strong correlation between wait times and operational features (like CAPACITY and GUEST_CARRIED), naturally identifying the park's "Mega-Attractions." Interestingly, total daily attendance showed a weaker correlation to wait times, indicating that headliner rides maintain long queues even on historically "slow" days due to localized crowd dispersion.
• Additionally, maximum wait times (WAIT_TIME_MAX) strongly correlate with operating hours (OPEN_TIME and UP_TIME). This reflects a clear operational strategy: the park's most demanded "blockbuster" rides are deliberately kept open longer to absorb massive crowds. Consequently, these extended hours increase their exposure to peak-time rushes, naturally driving up their maximum daily wait times.

👨‍👩‍👧‍👦 Part 2: Visitor Strategies (Queue Dynamics)

Uncovering behavioral patterns to help guests optimize their itineraries and avoid the "herd mentality."

🧠Question 1: How does the average wait time evolve throughout a typical day?

I used a Bar Chart to reveal the classic "bell curve" of park attendance. It highlights when the park truly wakes up, the peak midday congestion window, and the gradual decline towards closing.

💡Conclusion: The data reveals a massive midday congestion curve. The park wakes up slowly at 09:00, but queues aggressively peak exactly at 12:00 (reaching nearly 19 minutes). Wait times remain intensely high until 16:00, proving that arriving late and staying until closing is much smarter than arriving at noon.

🧠Question 2: Which days of the week are statistically the best to visit?

By plotting the days using a Bar Chart, this insight exposes the "weekend premium" (surge in wait times on Saturdays and Sundays), allowing visitors to choose the optimal day for their trip.

💡Conclusion: The "weekend premium" is heavily skewed towards Saturday. While weekdays (Monday-Friday) remain remarkably stable at around 11-12 minutes, Saturday spikes dramatically to almost 15 minutes. Surprisingly, Sunday is much quieter than Saturday, making it the superior choice for a weekend trip.

🧠Question 3: Does extreme temperature drive crowds away?

This Line Plot demonstrates the correlation between weather (Celsius) and queue lengths. It reveals that extreme temperatures can create a strategic window with shorter queues for visitors willing to brave the weather.

💡Conclusion: There is a clear breaking point. As temperatures rise from freezing to a warm 30-35°C, park queues steadily grow to their maximum. However, once the heat becomes extreme (pushing past 38°C), crowds surrender to the weather and wait times drop sharply, creating a unique opportunity for heat-tolerant visitors

🧠Question 4: Is there a "Queue Momentum" where current crowds predict future crowds?

Using a Scatter Plot with a Regression Line, I proved that a high wait time in the previous hour strongly predicts a high wait time in the current hour. The crowd doesn't disappear instantly; congestion rolls over.

This plot proves Queue Momentum. The red regression line shows a strong positive correlation: high wait times in one hour strongly predict high wait times in the next.

💡Conclusion:

Management: Congestion won't clear itself. Queue spikes require immediate intervention.

Visitors: Don't wait around for a crowded line to drop; move to a different attraction.

(Note: Data was sampled to 50,000 records for rendering performance).

🧠Question 5: When exactly are the park's "Dead Zones"?

I created a Heatmap crossing the day of the week with the hour of the day. This acts as the ultimate tactical tool, pinpointing the exact "Dead Zones" (light yellow) versus peak crunch times (dark blue).

💡Conclusion: The heatmap exposes Saturday midday (11:00 - 16:00) as the ultimate trap (darkest blue). To find true "Dead Zones", visitors should exclusively target the first hour of opening (09:00) or late evenings after 19:00, which are consistently empty across all days of the week.

🧠Question 6: Is it better to fight the Morning Rush or wait for the Evening Lull?

This Bar Chart compares two distinct visitor strategies: fighting the morning crowd (herd mentality) versus waiting for the evening when families leave, showing which timeframe offers statistically shorter queues.

💡Conclusion: The data clearly favors the evening. Morning wait times average over 10 minutes, while evening wait times drop below 7 minutes. Arriving late and staying until closing is a statistically superior strategy to rushing the gates in the morning.

🧠Question 7: Does the "Lunchtime Drop" actually exist, or is it a myth?

I used a Line Plot with a shaded area to track wait times during lunch hours. The plot clearly debunks this myth, showing that wait times actually remain at their absolute peak during the 12:00-14:00 window.

💡Conclusion: The data confirms that the "Lunchtime Drop" is real. While wait times peak at 12:00, the graph shows a steady decline in queues between 12:00 and 14:00 as visitors take a break to eat. Once lunch ends at 15:00, wait times immediately spike back up, making 14:00 a strategic sweet spot for rides.

🧠Question 8: Can visitors save time by using a "Contra-Flow" strategy?

By contrasting a "morning peak" ride with an "afternoon peak" ride on a Multi-Line Plot, we prove visitors can save hours by navigating the park in reverse to the general herd.

💡Conclusion: Yes. The Roller Coaster has massive lines from 10:00 to 16:00, but crashes to near-zero by 19:00. By doing the park in reverse saving headliner coasters for the evening visitors can avoid 20-minute waits entirely.

🧠Question 9: What is the exact optimal hour to ride each specific attraction?I created a Heatmap crossing the top attractions with the hours of the day. This creates a tactical matrix where green zones indicate the exact hour a specific queue collapses.

💡Conclusion: Rides do not peak equally. High-demand rides like the Haunted House and Skyway turn deep red midday and must be visited before 10:00 or after 19:00. Conversely, filler rides like the Circus Train or Crazy Dance remain green all day and can be ridden anytime.

---

⚙️ Part 3: Operational Optimization for Management

Identifying systemic bottlenecks to assist park management in resource allocation and crisis mitigation.

🚨Question 1: At what hour are rides most likely to suffer a mechanical breakdown?

This Bar Chart reveals the operational pressure points. Breakdown probability surges as rides run continuously under peak load, dictating exactly when to deploy maximum technical staff.

🎯Conclusion: Surprisingly, breakdowns peak early in the day at 10:00 AM (over 4% probability). This likely happens when rides experience their first full-capacity stress of the day. Maintenance teams must be on high alert immediately after the park fills up.

🚨Question 2: When is the highest risk for a "Black Swan" event (>40 min queues)?

This Bar Chart maps the park's absolute breaking points, highlighting the specific hours where wait times have the highest probability of spiraling completely out of control.

🎯Conclusion: Extreme queue events are highly concentrated in a specific midday window (11:00 - 16:00), peaking at over 16% probability. Management must proactively deploy surge capacity (extra staff/cars) specifically for this 5-hour window, as evenings are statistically safe.

🚨Question 3: Are evening queues creating an "Overtime Bleed" for park management?

I utilized a Horizontal Bar Chart with a threshold line to examine wait times exactly at park closing. The visual proves there is no overtime bleed, as queues are practically zero at closing time.

🎯Conclusion: Overtime bleed is a myth. By 22:00, the longest queue in the park (Tilt-A-Whirl) is under 8 minutes, well below the 60-minute threshold. The park's soft-closing procedures are highly efficient and do not generate overtime costs.

🚨Question 4: How long does it actually take a queue to recover from a short mechanical breakdown?

By comparing perfect days to breakdown days using a Multi-Line Plot, I uncovered the "Recovery Lag". The visualization demonstrates that a short downtime creates a backlog that takes several hours to clear.

🎯Conclusion: A breakdown doesn't just cause a temporary spike; it elevates wait times for the entire day. The green line shows that once a ride breaks, its wait time remains permanently higher than a perfect day until the park closes. Preventative morning maintenance is critical.

🚨Question 5: Does opening the gates cause an immediate, unmanageable stampede?

This Slope Graph debunks the operational fear that opening the park causes an immediate crush of visitors. It shows wait times at 09:00 are actually near zero, rising smoothly towards 11:00 as crowds slowly filter in.

🎯Conclusion: There is no 09:00 AM stampede. Wait times at opening are practically zero across all rides. The real crowd crush happens around 11:00 AM (where the Aeroplane Ride spikes to 40 minutes) as visitors slowly filter in. Heavy security is not needed at 09:00.

---

🛠️ Technical Stack & Repository Contents

Technology	Usage
Python	Core Language
Pandas & NumPy	Data Manipulation & High-Speed Logic
Seaborn & Matplotlib	High-Fidelity Data Visualization
Colab & Hugging Face	Development & Deployment Environment

• 📄 Disney_Theme_Park_Queue_Dynamics.ipynb: The complete source code, featuring step-by-step technical explanations and data wrangling logic.
• 💾 disney_cleaned_dataset.csv: The fully processed 3.7M+ row dataset used for this study.

Thank you for reviewing this analysis.

🚀 Quick Start: Load Data with Pandas

You don't need to download the massive CSV file locally to explore it! You can load this dataset directly into your Python environment (such as Google Colab or Jupyter Notebook) using the official Hugging Face URL integration.

1. Install Prerequisites

First, ensure you have the required file-system libraries installed in your environment:

pip install fsspec huggingface_hub -q

2. Load the Dataset

Use the hf:// protocol to stream the cleaned data directly into a Pandas DataFrame:

import pandas as pd

# The official Hugging Face path to the dataset
dataset_url = "hf://datasets/matanzig/Disney-Theme-Park-Queue-Dynamics/disney_cleaned_dataset.csv"

# Load the data directly into your environment
df = pd.read_csv(dataset_url)

# Display the first 5 rows to verify successful loading
display(df.head())