Tools and Techniques
To understand why recommenders don’t always work as well as expected, the team at Fast Forward Labs built some basic recommendation systems and dissected them. Here's what they found.
Booking.com is massive and as it continues to refine its products, new changes produce smaller and smaller effects. Even so, with over 1.5 million transactions each day, even a fraction-of-a-percent change in conversion can make a big difference to profits. Here's a practical look at how Booking.com's data scientists use a technique called CUPED to optimize experiments and squeeze new profits out of its products.
Shirin Glander's latest post describes an easy way to create a web API for a trained machine learning model. Once you have the API, other programs can access your model and use it to make predictions. This post is easy to follow and includes information about handling security on a production server.