Vendors are responding with holistic platforms that help integrate the process of big data analysis with analytics efforts across all areas of the organization. Kirsch points to IBM's SPSS Analytic Server, which helps companies get fast results for predictive analytics of big data, as an example.
9. Analytics Services Are Increasingly Hosted in the Cloud
Advanced analytics vendors are increasingly turning to the cloud to deliver analytics capabilities in a more affordable way, making them practical beyond the large enterprises able to afford the significant expense of complex, on-premise solutions.
"Some of these offerings are for specific use cases," Kaufman and Kirsch say. "For example, Angoss, Pega and SAP all offer salesforce.com applications through the AppExchange to perform analytics on CRM data. Angoss, IBM and SAS also offer more flexible software as Software as a Service (SaaS) that allows customers to do general-purpose analytics with cloud-based software."
10. In-Database Analytics Sidestep ETL Challenges
Performance, data governance and security can present serious challenges to performing advanced analytics on massive data sets. In-database analytics can ease many of these challenges by giving users the ability to deploy their models in the database itself rather than moving data to an analytics environment. By performing analytics on the data in place, the users can recognize performance and efficiency gains while simplifying security and data governance because the data never leaves the secure database.
"Many vendors are offering in-database capabilities for a number of data platforms, including Hadoop," Kaufman and Kirsch say. "IBM, SAS, RapidMiner, Revolution Analytics, Predixion, StatSoft, SAS and Angoss all support in-database mining. When evaluating a vendor based on in-database capabilities, it is important to investigate their support for the data platform your organization is using. Some vendors only support Hadoop while others support nearly every common data platform."
11. Companies Are Turning to Predictive Model Markup Language (PMML)
As companies make the switch from batch analytics to using real-time feedback to continuously improve the accuracy of their models, they are increasingly leveraging Predictive Model Markup Language (PMML). PMML is a standard for statistical and data mining models developed by the Data Mining Group (DMG), an independent consortium led by vendors. IBM and SAS are full members of the DMG, while SAP, StatSoft, RapidMiner and Angoss contributed to the development of PMML. Kirsch notes that the standard makes it easy to develop a model on one system with a particular application and then deploy it on a different system using a different application.
"These companies find that deploying models in applications with PMML helps to overcome delays and speed up the process of moving models more quickly into production," Kaufman and Kirsch say. "One of the major benefits of using PMML is that it eliminates the need for costly and time-consuming custom coding and proprietary processes."
Sign up for Computerworld eNewsletters.