DataRobot Announces Automated Time Series Solution

Print Friendly, PDF & Email

DataRobot, the pioneering architects of automated machine learning, announced the general availability of DataRobot Time Series. Following an extensive collaboration with more than 75 customers and world-class data scientists, this latest breakthrough completely automates sophisticated time series modeling. Finally, front line business people can quickly build highly accurate forecasts for a diverse array of time-dependent prediction opportunities.

Until now, handling time-dependency modeling required highly specialized expertise to even set up the problem correctly. Conventional modeling uses randomly selected records from a dataset to build and evaluate predictive models. For example, a representative sample of loan records and whether a customer defaulted. Each record is relatively similar.

Not so with time series. Temporal relationships matter. And there are so many potential ones to consider. Do you want to predict demand for a product three days from now? What day of the week is that? How does demand on that day of the week vary with demand on the other days? For the same day in previous weeks? Where does the week fall within the month? Does that matter? How about within the quarter? How is this month and this quarter comparing so far to last month and last quarter? The same month and quarter a year ago? Two years through 10 years ago? Is that a holiday in any location of interest and how does holiday performance differ? What promotions have you run and when did you run them? How have such promotions performed under similar temporal conditions?

Building on the 2017 acquisition of Nutonian, Inc. and its proven Eureqa modeling engine, DataRobot Time Series understands all these questions and how to set up the problem based on the answers. Then the DataRobot automation platform constructs and evaluates hundreds to thousands of different time series models and scores their performance – taking into account all the different temporal conditions to determine real world accuracy.

DataRobot Time Series beta customers, including Fortune 2000 retailers, banks, and hospital networks, have quickly built accurate models for staffing, inventory management, demand forecasting, financial applications, and more – all without the need for manual forecasting, specialized data science expertise, and custom coding.

Forecasting underpins most critical business functions. If you can predict the future, you can usually win the game. But it is one of the hardest problems in data science. Since the Nutonian acquisition last May, we’ve been on a massive undertaking to combine Nutonian and DataRobot innovations into the best time series product in the world. This fourth version, which has been extensively tested by customers in production, automates a wide array of advanced best practices in areas like feature engineering and thereby achieves a whole new level of accuracy,” said Michael Schmidt, Chief Scientist, DataRobot.

This new version, which is available now, includes advanced machine learning models for forecasting, as well as essential time series methods like ARIMA and Facebook Prophet. Full API support helps AI engineers integrate modeling and prediction directly into business processes and applications.

Time series machine learning has historically resisted automation,” says Srikant Datar, Professor of Business Administration and Faculty Chair of the Harvard Innovation Lab at Harvard Business School. “Having worked with DataRobot’s Time Series product for the past several months, including delivering real financial applications, I’m amazed at what is possible and how easily models can be built.”

Steward Health Care, the largest for-profit private hospital operator in the United States, is using DataRobot to significantly improve operational efficiency and reduce costs among their network of 38 hospitals across the nation. Sixty percent of hospital operational expenses come from staffing alone. With DataRobot’s improved forecasts for patient volume, Steward’s potential labor savings amount to $2 million by reducing hospital overstaffing by 1% for eight of the 38 hospitals in Steward’s network.

We have data – a lot of data – and we want to use it to our advantage,” said Erin Sullivan, executive director of information systems and software development at Steward Health Care. “DataRobot has the tools to help us take historical data, manipulate it, and learn from it. We’ve already experienced tremendous cost and time savings with DataRobot, and these latest advancements will further transform how we forecast nurse staffing and patients’ length of stay—both of which will yield significant benefits for our hospital network.”


Sign up for the free insideBIGDATA newsletter.

Speak Your Mind