Anaconda 4.10.3 was used. Specifically, the versions of the various parts used were Python 3.8.12, Jupyter Lab 3.2.4, numpy 1.21.4, pandas 1.3.4, seaborn 0.11.2, and matplotlib 3.5.0. I also used scikit-learn 1.0.1 and Andreas Mueller's WordCloud package (version 1.8.1).
There's a ton of Airbnb data available online here. I enjoy visiting Chicago and staying in Airbnbs there, and if I ever lived there I might consider listing my place on Airbnb from time to time. So, why not take a good look at the data on Chicago's listings?
To be concrete, I came up with some problems to work on:
- What are the factors that contribute most heavily to the price of a listing? Here I'm interested in modeling the price and isolating the most important factors in setting prices, so if I hypothetically got a place in Chicago, I could price it in line with similar listings. I'm also interested in knowing what the most important factors are, so I could decide where to buy and what to do to reasonably set a higher price.
- Which neighborhoods have the most variability in the prices of listings there over the course of the year? If I'm only interested in listing my place on Airbnb from time to time (say, when I'm a few hours out of town), I might want to do that at times when I can get the highest price for my listing. And if that's my strategy for making money using Airbnb, I want to buy a property in a neighborhood where there's a lot of price variability over time.
- What can we learn about the character of some of the neighborhoods using the descriptions in the listings? This one is more of a sotfball. I haven't been to Chicago many times, and don't know a lot about its different neighborhoods. The idea is to look at common words in descriptions given by Airbnb hosts and see if I can get a sense of what the neighborhood is like (or, at least, how it sees itself).
The main notebook is "Chicago Airbnb Analysis.ipynb", where I attempt to make progress on these problems. My code and extensive comments and reasoning are in this notebook.
I've also provided "my_listing.csv", which is a hypothetical listing I created in the proper format for my price predictor at the end of Part 1 in my notebook. If you want to price your own Chicago property, try changing the fields appropriately. My tests indicate that my price predictor does pretty well (20% error or less for half of listings given to it), but of course I make no guarantees about its accuracy for your particular case.
A summary of my findings is here on Medium.
This code is freely available under a GPL-3.0 License. Licensing information for the open Airbnb data can be found here; it is under a Creative Commons Attribution 4.0 International License. I want to thank Inside Airbnb for making this data freely available. I also want to thank all the developers who created and maintain the open-source software that makes this work and so much like it possible.