2020 is almost over, and we’re already fully looking forward to 2021. Even so, we can’t help but take one last look at the year that was. At the Explorium blog, we covered quite a bit of ground, exploring some interesting areas of data science, diving into machine learning (ML) for business, and even throwing in a few of our favorite TV shows. Before we turn the page on 2020, let’s look back at our top ten data science blogs of the year that was.
What happens when the rules of the game change overnight and leave you completely exposed? 2020 was a year of surprises, and it taught us the importance of having risk models and the right data to face even an unexpected storm. Read more here.
It’s tempting to think our ML models will last forever once we deploy them, but that’s just wishful thinking. In reality, factors like data and concept drift mean that the longer you leave your models, the less effective they’ll be. Read more now to see how you can avoid model drift.
Rules-based risk models are okay in some situations, but they really miss out when it comes to spotting nuances and slight shifts in established patterns. This post is all about dynamic risk models and why they should be a frontline of defense for organizations that deal with risk. Learn more here.
We love to root for the bad guy in TV shows, and Ozark’s Marty Byrde is as sly as criminals come. His money laundering schemes work for the Mexican cartels and avoid the law through three seasons (and counting), but could Marty evade ML-powered risk models? Read more to find out.
How smart are Marty Byrde and Walter White, really? This article looks at both Ozark and Breaking Bad’s money-laundering operations and pits them against the best risk models to see if Walter White and Marty Byrde are as smart as they claim they are. See how they stack up here.
Data acquisition is a must in today’s landscape, but data can be costly, and organizations can sometimes drag their feet about writing the check. So, how can you change that? By reframing the problem. You shouldn’t think of the costs that come with data, but the ROI and impact it can provide. Read more now.
If you want good results from your ML models, you need to prepare data the right way. Clustering is an easy and fast way to do some pre-training prep, but it’s not always the right choice. This post is a great primer if you’re curious about clustering and when to use it. Learn more now.
What can you do if you have urgent questions but the data you have doesn’t hold answers? It’s time to find new data, but it’s not as easy as it sounds. Data acquisition is a priority for data leaders, but it’s one that can come with landmines and roadblocks. This post breaks down how you can avoid them.
Machines are supposed to be completely objective, but unfortunately for ML, that’s not always the case. Indeed, though they might not be emotionally attached to a problem, ML models may develop biases that impact your ability to make predictions. It’s key to understand how and why so you can mitigate its effects. Read more about it now.
Choosing the right features to train your ML models is a tough step, and getting it wrong can severely impact your results. But it’s not just about picking the best-sounding features. You need to make sure the features you choose are optimal for your question and will be stable throughout the ML process. Learn more about it here.
It’s been a great year at the Explorium blog; with so many new posts and top data science blogs, we could have easily picked our top 20. However, it’s time to turn the page and see what 2021 has in store for us in terms of our upcoming top data science blogs. Tune in next year (in a few days, that is) to see what we have prepared for you. We promise you won’t be bored, and you’ll learn some great new things.
A good data scientist is a bit like a good journalist: they know how to ask questions so precise and to the point that there can
Who could possibly have known, back in those carefree days of early January 2020, that the best-selling items of the coming months — the products to
