Using Machine Learning to Find Paradise

We used machine learning to find paradise lost and the place might surprise you!

Ever dream of living somewhere that has free public transportation? Or the highest concentration of self-made millionaires?

Maybe an area where every resident has the equivalent of five tennis courts of green space, and there are 84,274.6 km of walkways?

You can have it all in West Perth!

Using Machine Learning to Find Paradise Lost

At least from an analytics perspective, this small district in the Western Australian boomtown of Perth is the best place in the world to live.

No surveys or existing list of criteria determined this result, but rather objective data alone. SAS examined no less than 148,233 locations in 193 countries for the Paradise Found project, without predetermined aspects to investigate or even a hypothesis.

Instead, we let the data speak for themselves. Over 5 million pieces of data from 1,124 data sources spoke up, including both structured and unstructured data (for instance, in the form of texts from statistics agencies). Overall, 1,060 international data services, three online geodata services, four social media services, and 57 urban studies contributed to the project.

Related: Machine Learning the Same as AI? 

Data wrangling and powerful data management software from SAS helped cleanse, structure and prepare the data. More information on how we dealt with the challenge of the diversity and volume of the data will soon be appearing in a blog entry by my colleague Andreas Gödde .

Using machine learning, the missing values for the individual locations were determined. A forecasting model was then developed that predicts locations that would be assessed as good places to live.