Research Question

The site, www.hikingproject.com is crowd-sourced and curated, similar to alltrails.com.

”At Hiking Project, our job and our goal are the same — to curate guidebook-quality information that inspires hikers like you to plan an adventure with confidence using the most accurate data available.”

Can we trust the trail ratings? What factors impact the rating?

Dataset

From each trail page, we scraped the Trail Name, Difficulty, Rating
Number of Ratings, Area, Trail Length, Trail Type, High Elevation, Low Elevation, Elevation Up, Elevation Down, Average Grade, Max Grade, Dogs, & Features.

Data Scraping

When you look for trails, the site shows only 10.

This code keeps pressing the “Show More Trails” button until it disappears.

Once all the trails are displayed, the link to each trail’s description page is placed in a list. For each entry in the list, that page is opened, and the designated attributes are appended to a pandas dataframe. That dataframe is then saved as a pickle file.

Raw Data

From this, I can see there is a lot of work that needs to be done to make it easily analyzable.

Data Cleaning

Just a couple of specific code snippets that show some the cleaning done.

The trail feature column contains a list. For logistic regression, these should be converted to True (1) or False (0) values.

Trail difficulties are mapped to different numbers.

Visualizations

Logistic Regression Conclusions

Statistically significant (p < .05) factors for lm model with all factors:

• Positive impact: Intermediate/Difficulty, Max Grade, Number of Reviews, River (TRUE), Views (TRUE), Lake (TRUE), Geologic Significance (TRUE)
• Negative impact: Easy Difficulty, Average Grade, All Dog Factors

Best subset with all factors (9 of 28):

• Positive impact: Trail Type (LOOP), Number of Reviews, River (TRUE), Intermediate/Difficulty, Max Grade, Views (TRUE)
• Negative impact: Easy Difficulty, Average Grade, Dog (Unknown)

Best subset without features (5 of 14):

• Positive: Elevation Up, Trail Type (LOOP), Intermediate/Difficulty, Max Grade,
• Negative: Easy Difficulty

Conclusions

Answer to our question: can we trust the ratings?

Possibly. There is a variety in ratings, so it seems like people are giving fair ratings. Trails with views and rivers are rated higher. Harder difficulty trails are rated higher than easy trails. Not enough ratings for the central tendency theorem to occur.

Further research possibilities

The entire website could be scraped, or a random sample of trails from the entire database could be analyzed to confirm our conclusions or improve the model.

Implications for natural resource managers

When planning trails, resource managers should prioritize loops with views and/or water features over other types of trails and features.

Implications for hikingproject.com

Because this is a crowd-sourced website, hikingproject.com should add an automated process to locate nearby geographic features for a submitted trail and add them to the trail report.