Page 28 - FCW, March/April 2020
P. 28

Artificial Intelligence
Democratizing
data science
With automated machine learning, any employee can reap the benefits of AI
Chad Cisco
General Manager, Federal, DataRobot
address potentially hundreds of problems a year.
In any area that collects a lot of data with a known outcome, automated machine learning can help predict a future outcome. In addition to preventing fraud, some major initiatives we’re working on include human resource planning in the form of selecting the best candidates quickly and predicting retirements so that agencies know where they might need to staff up.
Automated machine learning also helps with predictive maintenance for vehicle fleets, logistics planning, cybersecurity and complex workflows in places like hospitals.
ACROSS THE PUBLIC
sector, the amount of data that
agencies collect continues to grow exponentially. The question is: How do we channel the information that comes from all this data toward the most effective outcomes?
Artificial intelligence allows the government to harness the power that’s already available in various datasets to make more informed and more accurate decisions on behalf of citizens. AI does that by giving agency subject-matter experts the ability to quickly solve complex problems without relying on manpower in the form of dedicated data science teams.
Ideally, agency employees should be able to take advantage of AI without having to get a Ph.D. in data science. That’s where automated machine learning can help.
Using automation to find the best data models
An automated machine learning platform like DataRobot’s can create a new class of citizen data scientists and put the power
to create advanced machine learning models into the hands of agency domain experts without having to learn to code or understand when and how to apply certain algorithms. We help citizen data scientists leverage their agency’s data, automate
the process of testing the effectiveness of different algorithms, and determine the best models for predicting and describing the data.
For instance, if we want to reduce fraud in loan applications and we know that 10,000 of the 1 million applications submitted in the previous year were
fraudulent, automated machine learning can quickly identify the attributes that signal fraud in those applications. DataRobot’s platform can take in the data, prep it and test it against over 400 types of models to figure out which are the best for zeroing in on fraudulent applications.
Typically, a data scientist would go through that process manually. Automated machine learning helps make data scientists more productive because repetitive steps in the model-building process are automated. This frees them to use their unique expertise for selecting and fine-tuning models, helping them accelerate how they
davooda/Shutterstock/FCW Staff
S-28 SPONSORED CONTENT


































































































   26   27   28   29   30