A Quick Picture of Boston and Seattle Airbnb Data
Introduction
Are you wondering how Airbnb host set up rental price for their apartment/house? What can we discover from Boston and Seattle Airbnb dataset?
There are several questions I found interesting before diving into the data, hope we can figure them out by the end of the blog:
Is there any noticeable difference between Seattle and Boston Airbnb?
What are the most important features to estimate Airbnb rental price?
What are the top amenities people needs most?
We might be able to answer these questions based on our own understanding, but let’s also take a look at the answer from data.
Quick Glance at Seattle and Boston
From the data provided by mid August 2018, Boston has 6036 listings with an average of $184/night while Seattle has 8494 listings with an average of $152/night.
To give a more detailed sense of price range, 75% of Boston listings lies below $219/night while 75% of Seattle listings lies below $189/night which suggests the rental price for Airbnb in Seattle is slightly cheaper than in Boston.
The most expensive list in Boston is $3999/night while in Seattle the most expensive one is $5400/night.
Airbnb hosts can list entire homes/apartments (red), private (green) or shared rooms (blue). In Boston, hosts listed their room type as 62.2%, 36.6% and 1.2% respectively for the above three room types. While in Seattle the room type percentage is 66.6%, 30.4% and 3.1% respectively.
Let’s take a deeper look at data and try to use them to estimate the rental price of Airbnb in Boston and Seattle.
Predicting Rental Price in Boston and Seattle
As 99% of the listing price are below 500, so I drop the rows above 500 to get a more stable prediction. Missing values were filled by median value or most frequent value based on some other related features.
GradientBoostingRegressor was used as clasiifier for both datasets and a five-fold GridSearchCV was applied to find the best hyperparameter for the classifier.
On the other hand, 1/5 of the preprocessed dataset was used as test data and the remaining 4/5 are used to train machine learning model.
What are the Most Important Features to Predict Rental Price?
I listed top 20 features for both cities which plays the most important role in predicting rental price. Let’s take a look how they look like.
In Seattle, features are more equally important. There are some new features such as host_response_rate, extra_people and guest_included. But most of the top features are shared with Boston machine learning model. We can see cleaning_fee is the most important feature here, as more cleaning needs result from larger room, probably more bathrooms and more services.
What are the Top Amenities to Pay Attention?
After training machine learning model, we can also check how amenities influence the price prediction/how important are they.
In Boston, amenities such as TV, Internet lockbox and hangers are important as expected. Familiy/kid friendly is also a feature which most traveller will consider. We can see if you have a pool in your property, you might have more confident to raise your rental price.
One interesting phenomenon is Airbnb has more amenities listed in Seattle (168) than Boston (120).
Seattle hosts have some extra amenities than Boston’s such as Amazon Echo, Heated Floors, Formal Dinning Area and Memory Foam Mattress. Noting that they are very host dependent and not a lot of hosts provide these amenities so they might not influence the prediction a lot. We can see the importance of amenities are not as high as other features in the last section. As there are more than 100 amenities provided by the hosts and most of them are always provided as they are needed by most guest.
In Seattle, the most important amenities seems to be First aid kit, while TV and Internet are also in the top 10 as they should be.
Conclusion
In this blog, we dive into the most recent Airbnb Boston and Seattle dataset and found many interesting phenomenom. Here I listed what we have done:
We gathered the Boston and Seattle Airbnb data, and compare the two dataset.We established a machine learning model to predict the rental price for both cities.We took a look at the feature importance of the trained model and check if they make sense.We list all the important amenities to get a better feeling how host can make more money by providing better services to meet customers’ need.
There’s a lot can be done in the future such as studying the reviews provided by the customer. This is my first data science blog, hope you enjoy it, thank you for your attention.
