Problem Description

• Goal: Understand the relationships between park amenities, visitation rates, and visitor experiences
• Relevance: Insights can help improve visitor experience, optimize resources, and guide conservation efforts
• Scope: Patterns related to high visitation periods, amenity influence, and visitor satisfaction indicators

Project Overview

Objective

• Analyze visitor patterns and experiences across U.S. National Parks.
• Investigate factors influencing park popularity, visitor satisfaction, and the relationship between amenities and visitation rates.

Significance

• Insights can help improve park management, resource allocation, and visitor experience.

Analytics Plan

1. Data Collection: Using httr and jsonlite packages in R to request and parse API data
2. Data Wrangling: Clean and transform datasets for analysis
3. Exploratory Analysis: Identify data distributions and key patterns
4. Predictive Modeling: Use H2O to build models analyzing visitation influences

Data Preparation

Steps Taken

• Preprocessed the dataset to clean and normalize variables.
• Engineered features:
  • activity_count_log: Log-transformed count of park activities.
  • topic_count: Number of unique topics associated with parks.
  • average_fee: Average entrance fees.
• Selected a total of 20 predictors relevant to visitor engagement and park amenities.

Final Dataset

• Target Variable: activity_count_log
• Predictors: topic_count, average_fee, and geographic/amenity-based features.

Machine Learning Models

Evaluation Plan

• Metrics:
  • Model accuracy and F1-score
  • Cross-validation for robustness
• Data Visualizations: Highlight relationships and comparisons across predictors (e.g., amenities, visitor demographics)
• Outcome: Insightful data to support strategies for park management and sustainable tourism
• Update: Unable to implement aspects of our evaluation plan due to technical difficulties when running our model.

Gradient Boosting Machine (GBM)

• Goal: Predict activity_count_log using other predictors.
• Key Parameters: 100 trees, max depth of 5, learning rate of 0.1.
• Evaluation:
  • Trained on 80% of the data and validated on 20%.
  • Metrics used: RMSE, R².

Deep Learning Model

• Architecture:
  • Two hidden layers with 128 neurons each.
  • Rectifier activation function with dropout for regularization.
• Performance Evaluation:
  • Focused on reducing overfitting using L1/L2 regularization.
  • Validated on the test set.

Conclusion

• Expected Impact:
  • Aid National Park Service in understanding visitor trends
  • Enhance park management and visitor satisfaction strategies
• Real Impact -Unable to get proper modeling due to errors and technical difficulties. -Also had issues with the API
  • Upon further reflection, there are many ways we could have improved upon the project by cleaning the data more thoroughly
  • We could alse see if we could resolve the errors in our models.

Next Steps

• Extend the analysis to include time-series trends.
• Implement models for real-time resource optimization.