Introduction

Airbnb is a popular way for homeowners to make money by renting out their properties or even spare rooms in their own home. More people are considering joining Airbnb to profit by investing in new properties to transform into Airbnbs. However, how will they know what to consider to make their property an attractive proposition for customers? How will they identify which variables can increase their listing price and profit?

There is a problem though, you see. Hosts remove their listings for various reasons such as a lack of bookings or if the property is currently occupied. This means we must find a way to predict that data and recommend them a reasonable price so they can attract more guests. Before we can answer those questions, we need to find relevant variables to use.

Since 2008, guests and hosts have used Airbnb to travel in a more unique, personalized way. As part of the Airbnb Inside initiative, this dataset describes the listing activity of homestays in Boston, MA and Seattle, WA. We will use the CRISP-DM process to work through this project, CRISP-DM stands for Cross-Industry Process for Data Mining. The first stage of the CRISP-DM process is to understand what we want to accomplish from a business perspective.

Project resources

https://www.kaggle.com/airbnb/boston
https://www.kaggle.com/airbnb/seattle

The following Airbnb activity is included in these datasets:

  • listings: including full descriptions and average review score
  • reviews: including unique id for each reviewer and detailed comments
  • calendar: including listing id and the price and availability for that day

Business Objectives

How can we find the best location to start our new Airbnb?

  1. Can we find out if a review will be positive or negative based on the text?
  2. What factors can affect listing prices?
  3. How do those factors affect listing prices?

Project Plans

Our plan to achieve the data mining and business goals is to use a Jupiter Notebook as the method of handling our project. We will use:

  • Python 3 and packages such as Pandas, NumPy to handle dataframes
  • Matplotlib and Seaborn to plot visualisations of our findings
  • Plotly to display interactive dashboards and to show map data with overlays
  • Scikit-learn packages to perform machine learning operations

Airbnb Prices around Boston

We can see 75% of prices are below 220 USD but from the plot, we can see some prices rise to around 500-600 on average. The max value of 4000 USD is clearly an outlier if we look at the plot. Generally, prices can range from roughly 20 to 500 USD which is acceptable for an Airbnb listing price.
After removing the outliers, we see the prices are more closely

Observation: As no surprise, we can notice the prices increase as we go towards the city center and waterfont. The suburbs have low prices since there is not as much demand for those areas.

  • Neighbourhood: The highest-priced properties seem to be located in Leather District, Downtown, Chinatown and South Boston Waterfront. If we look at the map we can see they are located near the business district, with shops and restaurants dotted around the waterfront.
  • Property type: Boats tend to list for higher prices than houses but Villas seem to have the most variance in price. We also observed Houses, Dorms and Camper Vans have the lowest prices, probably because they are located in the suburbs. Houses also have the highest variance in terms of outliers.
  • Room type: As a surprise to no one we can see that an entire house has a higher listing price to a private or shared room.
  • Bed type: Real beds by far correlate to higher listing prices so we can comfortably say comfort is king!
  • Cancellation policy: The lowest-priced properties have the most flexibility in terms of cancellation policies.
  • Host is superhost: Superhosts are experienced hosts who provide a shining example for other hosts, and extraordinary experiences for their guests. Superhosts charge slightly more than others but the difference is almost negligible.
  • Instant bookable: It seems instantly bookable properties have a slightly higher price than non-instant bookable properties.
  • Is location exact: Properties with exact locations in their listings fetch higher prices
  • Require guest verification Properties that require guest phone verification tend to go for higher prices
  • Require guest profile picture Properties that require guest profile pictures tend to go for higher prices

Data Understanding Results

Through analysis of the categorical variables above, we see all the factors that affect the listing price. As an aspiring Airbnb host, a few factors can be taken into consideration to increase the listing price. From our observations above we can conclude:

  • Location, location, location: Location has a significant factor in price. The further out we go from the city centre and its amenities we have to accept the price will fall due to lower demand. Also, remember that providing accurate descriptions can help your property stand out from the rest. Try to explain what makes it unique and why?
  • People don’t like sharing much: We can see that most people opt for an entire home over a private room only. It seems people will pay more regardless to get more privacy as the cost savings are significant with private rooms.
  • Comfort is king: Comfort is so important to customers and should be the primary focus once the above factors have been satisfied. The type and size of the bed can affect the quality of a customers sleep and overall experience, we can look into this more when we use NLP to look at review sentiment. The listing should include details such as the type of bed and the comfort that can be expected to increase customer interest.
  • Experience is key: Experienced Airbnb owners called ‘Superhosts’ can charge more than ones that aren’t. They may also provide better services such as providing information on Boston’s local culture or history or, personalising their customer’s experiences which could attract more business.

If you want to invest in an Airbnb home you should first pay attention to the Neighbourhood based on your budget. Next, you must consider the four factors listed above, also within budget:

  • Location and accurate description
  • Privacy
  • Comfort
  • Experience

The other factors we need to look at may or may not increase your listing price as the potential profit increase is marginal:

  • Cancellation policy
  • Instant bookable
  • Guest profile picture

The cancellation policy may have a relationship with luxury properties which explains the increase in price. It also makes sense since the more expensive a property is to rent the more money it loses sitting vacant. This is probably not something that will increase our profit.

Instantly bookable homes tend to have a slightly higher price but the difference is negligible. I could indicate that less cleaning and preparation is needed for those homes if they are smaller apartments. Most of these factors can be useful for our price prediction.

Conclusion

Can we find out if a review will be positive or negative based on the text?

It is possible to use just raw text as input for making predictions but the most important factor is being able to extract relevant features from the data. I would say it should be used as a complementary source of data in order to extract more machine learning features to increase the predictive power of a current model.

What factors can affect listing prices?

We’ve found a few factors which can affect listing price:

  • Location
  • Privacy
  • Comfort
  • Superhosts

How do those factors affect listing prices?

  1. The location has a significant factor in price. The further out we go from the city centre and its amenities we have to accept the price will fall due to lower demand.
  2. People don’t like sharing much. We can see that most people opt for an entire home over a private room only. It seems people will pay more regardless to get more privacy as the cost savings are significant with private rooms.
  3. Comfort is so important to customers and should be the primary focus once the above factors have been satisfied. The type and size of the bed can affect the quality of a customers sleep and overall experience.
  4. Experienced Airbnb owners called ‘Superhosts’ can charge more than ones that aren’t. They may also provide better services such as providing information on Boston’s local culture or history or, personalising their customer’s experiences which could attract more business.

Our recommendation for a first time Airbnb homeowner is to find a property in a location close to the city centre. If budget is an issue then the next best thing would be to buy a property with good transport routes to the centre. The next thing to think about is the type of property. It would be best to go for the house as a first choice or an apartment as a second choice, if the budget permits.

The next most important thing to consider is whether to rent out the rooms or the home as a whole. From research, I think the best option is to rent out the home as a whole. There may be some risk in having periods of the home vacant instead of spreading the risk with multiple rooms. If the property has more than 3 rooms this may be a consideration.

Comfort is a very important part of the customers’ experience in an Airbnb from our research so good beds should be considered. Being a Superhost seems to greatly benefit exposure and potential profits since customers want great experiences. Being consistent and working on personalising customers’ experiences should be the main focus when starting as being a Superhost should be the objective.

Author

Write A Comment