MACHINE ANALYTICS, INC. — Department of Defense SBIR Phase II: A12-030

Description

Under Phase I, we have designed and developed a limited-scope Mobile Agent-based Distributed Search and Analytics Tool (MADSAT) to support the Armys intelligence and operational information and analysis requirements in the era of net-centric warfare and cloud computing. MADSAT is built on: 1) agent-based, distributed execution of optimized query plans; 2) data modeling for integrating heterogeneous, distributed data sources; 3) distributed fusion for complex analytics based on Bayesian network (BN) models; and 4) web-based and customized user interfaces integrated within Ozone Widget Framework. We have demonstrated that MADSAT can query distributed SALUTE data sources and perform complex analytics via distributed belief propagation on a hierarchical BN model to determine nuclear proliferation threat. We have initiated an effort to develop cloud-based MADSAT within the Apache Hadoop framework. For Phase II, we propose full-scope agent- and cloud-based development of MADSAT and demonstration and evaluation using actual represented data sources. We will enhance the data modeling capability in coherence with the Armys cloud data strategies, accessing DataSpace through DSMS API. MADSAT will support natural-language-based search and analytics queries. Query planning and execution in the cloud will be built on the MapReduce framework. We plan to transition MADSAT into the DCGS-A Standard Cloud (DSC).