The team of Optillel Solutions, Inc and Michigan Aerospace Corporation (MAC) propose to develop a generic cloud-based analytic tool, the Nimbus Analytic Toolset (NAT), a fundamental set of composable functions spanning the range of analytic activity. This toolset will employ virtualization techniques to facilitate transparent operation across the growing array of cloud computing infrastructures, including Hadoop, Amazon AWS and others. We will demonstrate the power of the Cloud for deploying Predictive Analytics algorithms including supervised learning (classification|regression), unsupervised learning (clustering|one-class learning) and discovery operations (anomaly/outlier detection|novelty discovery). NAT will support a graphical programming paradigm to compose processes in which petabytes of distributed data will fuel the generation of predictive models to support crucial tasks such as situational awareness, surveillance and data mining. MAC researchers will work closely with Optillel''s distributed computing experts to develop a viable strategy for deploying an advanced approach to the construction and utilization of Ensembles of Decision Trees into the Cloud model. MAC has evolved this pattern recognition framework over the past several years on programs with various government agencies. The proposed toolset will provide a flexible framework for the analyst that will allow easy adaptation to the rapidly shifting wartime environment.
Keywords: Cloud Computing, Task Scheduling, Intelligence Community, Predictive Analytics, Parallel Processing, Graphical Programming, Composable Workflow, Ensembles Of Decision Trees