|
PRODUCTS
|
|
SOLUTIONS
|
|
OPEN USE TOOLS
|
|
INDUSTRIES
|
| SERVICES
|
| NEWS
|
| EWA SYSTEMS
|
Data Mining
EWA Systems offers a comprehensive set of integrated, modular data mining algorithms utilizing a common task and data management frameworks. EWA Systems' algorithms are meant to work together, allowing similar algorithms solving similar problems to be swapped out. The task management framework controls and limits the number of threads to a manageable number, and schedule repetitive or delayed tasks. The data management framework simplifies the process of gathering data from various data sources, filter and manipulate the data into tables, and even allow multiple algorithms to access the data set without replication either locally or remotely. This modular design is illustrated below. EWA Systems' Data Mining Package Modular Architecture Performance: All EWA Systems' algorithms include all of the latest features and are lightning fast running on a desktop, or running on the latest server, supercomputer, or cluster. EWA Systems' Java algorithms have comparable performance to that of C-based algorithms, which only improves as additional local or remote processors are made available, unlike most C-based algorithms. EWA Systems' data mining algorithms in other languages (C#, C, C++, Fortran) show similar levels of performance.
Problem Size: EWA Systems' algorithms are optimized for both in-core, where the whole problem is stored in memory, and out-of-core solutions, where the problem size exceeds the available RAM. Small problems can be run in memory achieving the utmost performance. With the Data Management Framework, large problems can be cached or run with all of the data left on local or remote disks with just a small speed penalty. EWA Systems' algorithms are commonly run on terabyte- and even petabyte-sized data sets. All EWA Systems' Data Mining algorithms use the same Data Management Framework (DMF). The DMF also provides a single comprehensive point of interface back to flat files, databases, and data streams, and a host of filtering options and many other SQL-like operations. The DMF handles the multi-threading synchronization, caching, security and clustering of these data sources. The DMF Architecture document details these supported operations. EWA Systems' Data Management Framework Architecture Robust Accuracy: Each of the six data mining problem type, classification, regression, clustering, association, outliers, and time series, is represented by multiple algorithms so the learning bias of one algorithm can be mitigated by the other algorithms. These algorithms are offered as individual data mining components, as individual data mining GUI applications, as part of a data mining package, or integrated as part of EWA Systems' Data Mining Suite, Visual Data eXplorer (all below). The following two tables detail the various data mining algorithms and the additional modules available:
EWA Systems' text mining package and Bayesian Decision Support package extends these data mining algorithms. Text can be analyzed based upon wording, an unstructured text analysis, or taking into account how the words are used in sentences or in reply to certain questions. Bayesian Decision Analysis supports the modeling of decision making situations. Using this package, one can determine what is the best decision based on current information, and the value of finding out more information regarding any remaining uncertainties.
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
