Using Microsoft AI to find the best Airbnb

Recently we have been looking at an experimental labs feature of Microsoft Cognitive Services called Project Wollongong. This API enables you to rank physical locations based on an attractiveness score. This attractiveness score is calculated by the proximity to surrounding attractions that are of interest to you.


An example

We used this API to rank different locations for going on holiday; whether you are going on a package holiday, city break, caravan site or staying at an Airbnb, you want to know what is around you. If you look at services like Airbnb or Hotels.com you can filter by different amenities but within an already geographic area of a city or town. With project Wollongong, we can apply the same filtering at a country level enabling people to decide based on their interests and requirements rather than defining the location first.

We take 10 Airbnb locations spread across the USA to visit and specify the amenities that are important to the visitor.

  • If you have kids, you want to be near a hospital or 24-hour grocery store
  • Interested in the arts, then museums and the arts within 30-mile drive
  • Public transport is a 10-minute walk, and there is an spa nearby

image

As you can see from the map, this shows the ranking scores for our locations and make it very visual and easy to start comparing locations across the country.

For this lab project, amenities are grouped into twenty-five categories with more coming; the current categories cover everything from Bars to Banking, Churches, Grocery, Museums, Medical Centres and Supermarkets.

How does it work

There is a great technical video by Scott Peterson on the Microsoft virtual academy that walks through the API in good depth if your interested in taking it for a spin.


The Microsoft labs project only covered the USA currently, but it offers some interesting possibilities even in its infancy and use-cases to pre-vet locations for both consumer and business solutions.