The word innovation often draws to mind images of self-driving cars, new phones, and shiny tech. Yet, innovation often happens behind the scenes, especially in advanced analytics.
Around the world, industries like healthcare, government, banking, manufacturing, and more rely on the latest advancements in analytics.
At SAS Explore, an event for technologists, Udo Sglavo, Vice President of Advanced Analytics Research and Development, shared four key areas of innovation happening at SAS.
Throughout the general session on day two at SAS Explore, Sglavo interviewed various experts about how SAS is paving the way in advanced analytics and machine learning. Together, they covered the speed and repeatability of advanced analytics, proactively preventing biased decisions in AI, analytics on the go, and the possibilities of synthetic data.
Making advanced analytics faster and more productive
In the past, advanced analytics was limited to large-scale, high-dollar projects. With advancements made in the last decade and digitalization's ongoing impact in response to the pandemic, adoption has skyrocketed. Businesses now regularly use advanced analytics for decision making, demand planning, and more. Thankfully, analytics in the cloud helps to meet demand.
The speed and agility of SAS® Viya® 4 in the cloud allow data scientists to test multiple solutions faster and more productively.
DIVE DEEPER: Watch this full demo with Josh Griffin, who heads the Advanced Analytics Foundation Department team, to learn more.
Responsible innovation: AI and bias
It’s clear advanced analytics and AI are already changing the world. But AI poses risks and can cause unintentional harm to marginalized groups if not handled responsibly. For example, this use of a hiring AI unintentionally discriminated against female candidates.
In response to these concerns, SAS created the responsible innovation program. We must help customers innovate with AI in a responsible, trustworthy and fair manner.
Some responsible AI features have been standard in SAS software for years, including automatic detection of private and sensitive information in data, model interpretability, and natural language-generated explanations of results.
In addition to these important features, SAS now offers bias detection and mitigation in models. Data scientists can run the model to assess bias and accuracy for various groups. SAS is giving technologists the tools to identify this bias to prevent building biased models into the large-scale decision-making process.
DIVE DEEPER: Watch this demo from Jonathan and Justin about how we’re supporting our customers in using AI responsibly.
Analytics on the go
Pelin Cay, manager in the SAS Advanced Analytics Center of Excellence, demonstrated how she used SAS Analytics Pro to quickly solve one of the oldest optimization problems in the book: the traveling salesman problem. This problem famously asks, “Given a list of cities and the distances between each pair of cities, what is the shortest possible route that visits each city exactly once and returns to the origin city?" In her example, Cay asked how to find the best route between all the baseball stadiums on a three-week trip. She also complicated the problem by adding specific time slots for each stadium.
Even with the extra variable, Pelin could solve this problem quickly on her laptop. Creating easily accessible data analytics is an innovation with a wide-reaching impact. SAS customers use similar techniques to optimize the delivery of holiday packages and to optimize hospital resources during disease surges.
DIVE DEEPER: Watch the full demo (from install to solution!) with Pelin Cay.
The power of synthetic data for innovation
We need data for machine learning to help us solve problems like medical diagnostics or fraud prevention. (And lots of it!) While machine learning models are already sophisticated, they are also increasingly data-hungry. Data scientists don't lack machine learning algorithms, but data. (Especially high-quality data.)
Gathering real data poses many challenges. First, it’s costly and time-consuming. Additionally, privacy concerns and issues of representation limit which data is used for accurate modeling.
During SAS Explore, Mary Osborne, Senior Product Manager of Advanced Analytics and AI, shared Synthetic data can be part of the solution.
Synthetic data artificially manufactures data sets with special-purpose machine learning models that capture the data distributions and patterns while also helping to maintain privacy. Generative Adversarial Networks (GANs) learn the patterns and relationships in existing data to generate new observations that are indistinguishable from real data. GAN models work great for synthetic data generation, particularly for image data. We can use this same technology for tabular data, which trains predictive models with machine learning algorithms.
While synthetic data has limitations and will require regulations, generating accurate synthetic data can increase the rate of innovation and the use of machine learning.
DIVE DEEPER: Watch this demo with Jonathan and Mary about SAS’s innovative approach to composite AI to learn more.
Continued innovation
The innovations covered here just scratch the surface of what’s to come for analytics in the next few years. As the founder and future of analytics, SAS continues to invest in R&D to provide customers access to the latest innovations.
Ready to see more innovative ways data is changing the world?
