SAS automates data modeling for fast analysis
- 18 June, 2015 05:22
SAS Factory Miner now includes machine learning capabilities
SAS wants to supercharge your business analysis, through new software that automatically builds multiple models of data and picks those that best predict future events.
"If you take too long coming up with a model, you lose a lot of value," said Sascha Schubert, SAS technology marketing director, speaking of today's competitive environment. "You want to be efficient in your analytics."
With a customer base of over 75,000 organizations, SAS has long been known for its advanced statistical analysis software. With this new offering, called SAS Factory Miner, the company is trying to go one step further to help its customers, by generating models for their data.
Traditionally, business analysts construct data models by hand, choosing from a data table, or multiple tables, the variables to examine closely, and trying to understand how they work together to produce a desired outcome.
With so much data today, however, it can be difficult to pinpoint the specific factors that are key indicators. Were sales last weekend good because of some discount pricing, or because people had more time to shop during the weekend, or some other, hidden factor?
As its name indicates, SAS Factory Miner automates this process of building and testing models, which could lead to better models that are generated more quickly than what could have been done by hand. It could also help alleviate the need for an organization to hire more data scientists, who are much in demand.
SAS Factory Miner can use any source of data, as long as the data itself can be formatted into a table. The software, run from a server and accessed with a browser, offers a graphical point-and-click interface. It comes with a set of customizable templates for creating baseline models. Analysts can fine tune or revise any of the computer-generated models.
To help pick the best models, the software uses a number of machine learning algorithms that, through repeated testing of the models, can recognize patterns to anticipate future performance. One unnamed customer used an early version of the software to build 35,000 different models in order to find the best approach for a marketing campaign.
This approach also helps enable what Schubert called stratified modeling, in which large sets of data, such as sales, can be divided into smaller segments. Stratified modeling can offer more accurate results, though its use has been limited by the time it takes to build the models, Schubert said.
A financial institution could build different models for different sets of potential users, based on spending ability, buying behavior, or other factors. These models could then be used to generate more appealing credit card offers.
The technology can be used in multiple ways by business, especially in the field of marketing, Schubert said. It could be key to reducing customer churn, predicting future demand, personalizing offers, and managing risk.
For instance, a manufacturer could use the software to build more accurate predictions for when equipment will fail. A firm in a fiercely competitive field could build models for each of its competitors, and then run them in unison to get a full view of the industry.
SAS Factory Miner will be generally available around the middle of July. The company did not provide pricing, noting the price varies by the size of the installation, and the work it will do.