Airbnb prices: What’s the deal?
While working on the Udacity Data Scientist Nanodegree, I had to choose a dataset to investigate and share some interesting findings! Turns out that there’s a lot of data available from Airbnb, and I dove into the listings data for Boston and Seattle to see if I could answer these three key questions:
- What is the distribution of Airbnb prices in Boston and Seattle?
- What are the top contributing features, and how does it vary from city to city?
- Could we use this analysis to help people make their Airbnb rentals more attractive?
The Boston and Seattle datasets were of similar size, with over 3500 listings contained in each. There were over 90 features available in each dataset, it wasn’t clear right off the bat which ones would be the most important. Luckily, for the first question we can just check out the price feature!
What is the distribution of Airbnb prices?
Taking a look at the distribution of nightly prices, it’s clear right away that the distributions are different! The average price in Boston is around $175 and in Seattle it is around $125, although have a fair number of apartments in the 50–200 dollar range. Boston seems to just have a higher proportion of apartments in the 200–400 dollar range! I wonder which features people in Boston and Seattle seem to value?
What are the top contributing features, and how does it vary from city to city?
To find the feature importances I cleaned up the data and tried to predict the price using decision tree regression, which is an effective way to take a lot of features and come out with a price estimation. You can even extract the features to find out which is the most important!
Looking at the extracted features in the plot, a few interesting patterns emerge that aren’t surprising: the size of the listing and the location of the listing have a massive impact on its price. There are a couple other common features that are interesting though, in both Seattle and Boston having an elevator and a smoke detector are likely to raise your price. A TV helps too!
When looking at the difference between the two cities, it seems only a minor question of guest priorities. In Seattle people look for bigger apartments, and in Boston it’s much more important for them to rent the whole apartment. Although in both places size AND renting the whole apartment are important.
Could we use this analysis to help people make their Airbnb rentals more attractive?
Now we get to the real meat of the analysis: can we make use of what we’ve done to provide some advice to hosts or guests? As it turns out, there isn’t too much we can recommend for hosts. The price you should list most depends on the listings size and location! As a host do be sure to have smoke detectors, a TV, and respond promptly!
As a guest, this analysis only provides some really basic insight. When you are looking for a listing, be sure to only get the Airbnb that fits the specific number of people that are going. Also importantly: check different locations to see if one is much less expensive!
That wraps up the analysis — I’d say we learned a few things but nothing earth shattering. If you enjoyed reading and want to check out the data, you can find it on Kaggle here and here. And if you want to see the nitty gritty details of the analysis, check that out in the github repo. Thanks for reading!
