DataSonar’s off-the-shelf Machine-learning building blocks let you deal with huge amounts of data and tackle any kind of business problem.
Machine-learning engine and workflow
The data management and machine-learning engine algorithms are implemented in the Complex Event Processing framework, and constitute DataSonar architecture’s middle layer. Implemented at this level, the machine-learning workflow is interfaced and controlled, through dashboards, by the user.
Dashboards are there to give the user, including the non-IT-expert user, full control of the data mining process.
Simply defined, workflow consists of a series of stages:
- data acquisition.
Making data available, from a variety of sources, to the machine-learning workflow
- data enrichment.
Extracting and transforming data so that work is cut out for the machine-learning engine
- feature selection.
Defining a specific feature subset that best characterizes data for a specific purpose
This is machine-learning in the strict sense: applying algorithms and storing their outcomes.
A big DataSonar advantage over the competition is that it addresses expert and lay users alike by offering them a number of options:
- using the wide range of built-in patterns and tools to let them get down to work right away
- integrating external math and/or machine-learning libraries for special tasks or team consistency
- using any of the available APIs and writing their own special-purpose code.
Another big advantage is that it fills a market gap. Many machine-learning and data mining tools are available today. Some of them have an impressive range of statistical methods; others focus on big data challenges; some provide generic tools to implement data analytics, such OLAP cubes. However, few tools provide off-the-shelf machine-learning building blocks capable of dealing with huge amounts of data and generic enough to tackle any kind of business problem. DataSonar does.
Users can further break into, and compose, the workflow; and then redefine it by mapping each step to components in the framework (e.g. built-in machine-learning algorithms, operators, computing resources, precomputed data, external libraries, etc.). This unparalleled flexibility will save users’ time and their organizations’ money.
See DataSonar Data Discovery module.