BA 322 | Business Analytics | Spring, 2022
Project #3: Forecasting Visitors to National Parks
Selecting forecasting methods for datasets with different characteristics
Background
When we forecast time series data using moving averages, exponential smoothing or linear regression, we assume that historic patterns in data can be used to predict the future. This assumption can be reasonable during normal circumstances, but is tenuous in the time of a global pandemic. In this uncertain environment, we need more than historic demand to make a good forecast, we must also try to understand the ways in which the coronavirus has impacted this demand.
Before the pandemic, there was a lot of variation in attendance to national parks. Some parks had few visitors, while others were required to ration permits to keep visitation at sustainable levels. Many parks experienced strong seasonal variation in attendance. The coronavirus disrupted park attendance both by spurring official closures and by changing the face of summer travel. In the face of the coronavirus, people traveled regionally and engaged in more outdoor related activities. These changes are likely to impact all parks, but are not likely to impact all parks in the same way.
Assignment Summary:
Identify and explain trends and seasonal patterns in attendance for a National Park of your choosing (not Yosemite or Great Smoky National Park).
Forecast park attendance for the remainder of 2022 – using forecasting methods appropriate to your dataset.
Present your findings in a presentation recorded on ZOOM and posted to Canvas (3 minutes max)
Review the presentations of 3 other students and draw from their work to explain the ways in which the coronavirus has influenced park attendance.
Instructions:
Download data
Visit the National Parks statistics website and download data (as an Excel file) on recreation visits by month for any park of your choosing (do not choose Yosemite or Great Smoky Mountains). When choosing among the parks, think about the characteristics of the park and take a guess about the ways in which covid might have decreased or increased park attendance.
Some questions to ask are: Is this park close to an urban center, or is it remote? How do visitors reach it and how long are they likely to stay? Think about the ways in which the coronavirus may have changed attendance for the different National Parks and choose one that interests you.
Look at historical demand & make a forecast:
Begin by identifying existing patterns and trends in attendance data for your park (10 POINTS):
Sort all of the data from oldest to newest. The newest data should be at the bottom
Create at least three plots to evaluate trends and seasonality in the data. Specifically, plot:
Annual visitation by year (line chart)
Average visitation by month (column chart)
Monthly visitation for the last 5 years (clustered-column chart)
Describe what you see. Are there trends or seasonality in the data? Discuss them in two text boxes (one for the trend, one for seasonality).
Calculate a monthly index and deseasonalize demand. Plot deseasonalized demand.
Identify an appropriate forecasting method that suits your data (15 POINTS):
Choose two forecasting methods that might work on your data (i.e. Moving Averages and Linear Regression)
Create historical forecasts using both methods
Calculate the ME, MAE, MSE and MAPE for each method. Discuss the performance of the two forecast types.
Using the method that produced the BEST HISTORICAL FORECAST (lowest MSPE) (5 POINTS):
Forecast data for the remainder of 2022
Compare your forecast for 2021 to actual attendance for 2020 (10 POINTS):
Discuss how the coronavirus pandemic impacted park attendance.
Talk about whether or not you think that 2022 attendance will be above or below the numbers you forecasted. Explain.
Create a Zoom presentation to present your findings:
Include the following slides:
Title slide with your name, the name of your park and an introduction to what you’re doing (5 POINTS)
A slide introducing your park. Include its location and key details about what it offers. Discuss who goes, when they go and why. Include graphs that show the presence (or absence) of seasonality and trends. (10 POINTS)
One slide showing your forecast. Mention the forecasting method used and explain why it’s appropriate for the data. Highlight what visitation would have been without the pandemic and how the pandemic reduced/increased traffic. Show your projections for the rest of the year. Explain how you arrived at your estimates for the remainder of 2020. (25 points)
A final slide that discusses the impact of the coronavirus on park attendance and wraps up your analysis. (10POINTS)
Guidelines for creating good presentations:
Use bullet points and graphs on slides – do not type paragraphs.
Create a script to minimize filler words and errors.
Keep narration for each slide short. Average 30 seconds or less per slide with a little more time spent on slides 3 and 4. Your whole presentation should be about 3 minutes long.
Review other’s Zoom Presentations and write a synopsis
Formulate an answer to the question: “In what ways has the coronavirus impacted attendance at national parks, and how/why does that vary across parks?”
Watch the Zoom presentations of at least 3 other students
Write a perfectly-constructed one paragraph synopsis of the impact of covid on park attendance. There are multiple ways in which covid has impacted the ways we travel and recreate. These changes are reflected in attendance to our national parks. Each park will have been impacted differently, and your summary should reflect this. This should be of professional quality: grammatically perfect, nicely formatted and impeccably organized.
Include a list of the presentations that you viewed including the name of the student and the park that they reviewed.
Deliverables
Zoom presentation – 40 points
Excel – 40 points
Synopsis – 15 points
Self-Assessment – 5 points
Extra resources
Links to videos:
Time series forecasting #1| Plotting Time Series Data (Zoom #1)
Time series forecasting #2 | Moving Average Forecasting
Time series forecasting #3 | How Good is Your Forecast? (Calculating errors, me, mad, mse maps)
Time series forecasting #4 | Simple Exponential Smoothing Forecast
Time series forecasting #5 | National Park Forecasting (Comparable to first Zoom, plus linear regression)