Strong Blog

Building and deploying machine learning solutions can be a messy business. In this post, we review hallmarks of rigorous, reproducible approaches to machine learning development that encourage model iteration, simplify model validation, and ease deployment. We also introduce `strong-bootcamp`, a simple Python package that implements these principles.

Black-box and derivative-free optimization is a discipline in mathematical optimization that does not use derivative information in the classical sense to find optimal solutions. Unlike most optimization algorithms, a derivate-free algorithm may query the value of objective function at a given point, but it does not obtain gradient information, and in particular it cannot make any assumptions regarding the analytic form of the objective function. Here we overview one class of derivative-free algorithms, evolutionary algorithms (EA), and present a suite of robust Python implementations of each.

Visualizing graphs as pairwise relations in a network can be a challenging task. In this post, we discuss how simulations of Coulomb's Law can be used to draw graphs in an aesthetically-pleasing way without requiring sophisticated knowledge about graph theory.

One of R’s strengths is its powerful formula-based interface to modelling. It saves you the headache of one-hot-encoding categorical variables, manually constructing higher-level interactions, and rewriting models just to add or drop terms. Instead, you just specify the model you want in high-level notation, and R will handle transforming it into the raw numerical model-matrix that’s needed for model-fitting. We wrote keanu to make tackling situations where R's built-in interfaces fall short. It allows you to focus on the core aspects of your models, providing the tools that handle the busywork of translating abstract model-specifications into the nuts-and-bolts of model-fitting.

Our co-founder, Jacob, wrote on Medium recently about why data science consulting is his dream job.

Strong Analytics is seeking a data scientist to join our team in developing machine learning pipelines, building and validating statistical models, and helping our clients discover value in their data.

It's easy to see how deep learning is a powerful tool for companies like Google, Twitter, Facebook, and Amazon. In this post, I review several ways that deep learning can have immense value for smaller SaaS companies. The same technology that powers self-driving cars can even help you win new customers, make recommendations, and decrease churn.

Understanding the value that a predictive model can add to your business is often challenging. We wanted to make things simpler by building a free tool to calculate the ROI ($) of your predictive model based on its performance, number of possible events, baseline rate of events, and profits/costs of your intervention.

Data science is a new and growing field. As such, most organizations are still figuring out how to use these new tools and strategies to unlock new opportunities. We want to help! Take our free data science audit to find weaknesses and opportunities for you to pursue and, in turn, leverage the power of data science in your organization.

Optimail offers its users the opportunity to continuously and automatically optimize their email campaigns using artificial intelligence. Marketers can simply draft the emails they want to send (or many versions thereof), define the goals they want these emails to achieve, then sit back and watch as Optimail learns the strategy that drives customers most effectively toward those goals.

In this post we talk about customer lifetime value, a critical metric for any business. We describe problems with the traditional way of calculating this metric, and go into detail about a method we prefer, based in survival analysis.

We used machine learning to predict whether customers on a free trial would convert to a paid subscription, significantly increasing the conversion rate at a small SaaS company.

We worked with a software-as-a-service startup to analyze their customers' landing pages. With a little help from machine learning, we determined what kinds of pages had the highest conversion rates, and made our customer (and their customers) happy in the process.

Some think that making data-driven decisions requires 'big data,' leaving smaller organizations out of the recent data science movement. I argue to the contrary that thinking about data science early on can not only prepare your organization for future growth, but offer insights even from the small data you have now.

We lay out a comprehensive data strategy for your business to make data-driven decisions about strategy, marketing, and product development.

What's the difference between posters of good movies and bad movies? We use a Convolutional Neural Network to classify good vs. bad posters and explore the visual features that drive this distinction.

I took a look at movie ratings in IMDB and Metacritic from the last fifteen years to examine agreement between critics' and fans' movie ratings, how they've changed over time, and which movies they disagree on most.