Renting in Barcelona

Renting in Barcelona

Predict yearly occupation of night that an AirBNB could be rented.

How many days a year will a property be rented out depending on its price, conditions and position?

This case of use offers a data-analysis and machine-learning approach regarding a possible air-bnb real state investment . It aims to solve the question.

01. Airbnb Data Analysis

We start by dividing the map by districts showing the number of properties available for renting, showing an analysis of the most frequent price given by the neighborhood geolocated in the map. Even though the most frequent price overall per night is 30 euros in the city, there is a extreme variance regarding prices and most common ones depending on the neighborhood and the Airbnb characteristics. Besides and as an overview, there is an overall frequency of prices that ranges from 15 to 250 in the city.

Number of AirBNBs and median price in Barcelona

02. Prices & Neighbourhoods

In a more granular setting, during this case of use, we explore 10 neighborhoods from different 4 Barcelona districts, given the quality information that the dataset provided. Regarding this, we show the median and the minimum price of neighborhoods per night.

Regarding statistical analysis, the median stands for the middle point in the dataset -half of the data points are smaller than the median and half of the data points are larger-. We plot a comparison with the minimum price per night, giving us an overview of the decisions that could be made regarding the prices per neighborhood.

Note that in some neighborhoods there is an incremental going from the 5th to the 6th night, giving us the hypothesis that rents might have 2 days, 5 days and 7 days length, and from the 5th to the 6th rentability might be broken, therefore the price is incremental.

With this overview and a granular setting in the map, we could have an idea of the type of Airbnb and overall price of a location.

03. Airbnb Types

Conditions and Airbnb room types play a significant role when it comes to renting. Therefore, the radial chart shows the types of Airbnbs (Entire Apartment, private Room and Other that sums up).

04. Conclusions

Regarding the machine learning challenge proposal, we have built a occupation prediction model ( in days ) in the shape of a form, selecting the most relevant features regarding occupation, aiming to solve the question ‘How much days might my AirBNB property be occupied based on the price I might charge per night as a host?‘ You can select the price per night, type of neighborhood your AirBnB property is located in, giving you the opportunity to design the better pricing proposal for your property based also on the locations and property conditions.

For the machine learning prediction models, we have studied the linear dependency and the distribution of the data, as well as possible correlations, and choose coherently and run automl-jar having RSME – Root Mean Squared Error –  as key metric for evaluation. Having this baseline and criteria, we have selected  XGBoost as our algorithmic solution for predicting occupation. Find in the form above, selecting the price, other features and location of the AirBnB,and find the occupation that the models give as a result.

Note that price is the most relevant feature of the model as the different plot trees result have shown and occupation might vary depending on this price and the maximum number of nights.

As a conclusion we might say that AirBnB occupation relies on prices, the type of neighbourhood the property is located, the availability in terms of maximum nights and the quality of the AirBnB measured in the scores given by the reviews and the number of reviews per month.

Do you want a data-driven solution?