Online travel marketplaces have a particularly complex challenge when it comes to fighting fraud: bad behavior can come from multiple directions. On one hand, scammers may sign up on the site to list fake properties and con unsuspecting travelers. And on the other hand, fraudsters may also book rooms using stolen credit card info.Read More
Let’s start with a simple truth, universally acknowledged: no one at a business – no matter what department they work in – wants to turn good customers away.
No fraud analyst shows up at work saying, “Today, let’s keep our sales down.” But sometimes the fraud and risk department ends up gaining a reputation for being the people who say no. That’s because, on the surface, they may seem to have completely different goals:Read More
You’re a worldly, 21st century business. You keep up with your customers by offering real-time purchasing, allowing customers to exchange virtual cash, and encourage easy checkout by app or website. You’re doing everything you can to be competitive and buzz-worthy in this digital minefield of e- and m-commerce.
But because you’re doing everything right, you might be inadvertently opening yourself up to malicious users. From something as “benign” as coupon or referral abuse to straight-up credit card fraud, websites and apps are seeing it all. Even the bad behavior that seems relatively harmless – like spammy content or fake accounts – ends up damaging your business’ reputation and your customers’ experience.Read More
At Sift Science, we use a variety of popular machine learning models to detect fraud for our customers. However, until recently we relied exclusively on a combination of linear models and sophisticated feature engineering. As we were reaching the limits of this setup, we began experimenting with our first non-linear model: random decision forests. Several months and over 100 experiments later, we were thrilled to announce the addition of random decision forests to our ensemble of models used to fight fraud. Along the way we learned quite a few things about designing a random decision forest classifier for the fraud detection use case. Here we detail several of these learnings, including how we handled sparse and missing features, useful model visualization techniques, heuristics we used to improve class separation, specialized feature engineering, and how we combined our random decision forest with our existing models. All told, these learnings resulted in an 18% reduction in error for our customers.Read More
“I love filling out forms!” …. said nobody, ever. Especially when shopping for something online. And SUPER-especially when shopping for something on mobile. All that pinching, zooming, and fat-fingering. It’s painful.Read More