Data Analytics &
Machine Learning

Our Data Analytics and Machine Learning teams understand the power and challenges of collecting, analyzing and adapting data, no matter how large and complex, to delver the insights you need.

Data Analytics

Expression Networks Data Analytics provide value at the heart of logistics, operations, and management by enabling your organization to leverage its data instead of big data controlling your organization. Our Data Analytics solutions help customers derive actionable knowledge from the vast array of available information resulting in effective real-time decision making and enhanced productivity. Today, many factors and related demands interact, generating an enormous amount of interconnected data relevant to a customer’s mission. Properly harnessed, this data empowers decision makers and end users to achieve results quickly and effectively – enhancing productivity and mission success. Yet commonly used software tools are often obsolete and unable to capture, analyze, store, distribute, secure, and manage the vast store of increasingly complex big data (both structured and unstructured) in a timely manner.

Our data analytics experts work closely with customers to understand the intricacies of their data and the relevance of accessible external data. We provide them with the tools to organize and analyze all applicable data, no matter how vast and complex, to achieve recommended solutions to specific organizational or mission objectives. Our analytical solutions allow decision makers to quickly and efficiently cut through increasing data volume and gain keen insights to improve operational efficiency. Our platforms are scalable and flexible, capable of collecting, organizing and analyzing ever-expanding data sources, allowing customers to continuously adapt to new challenges and opportunities.

Data Analytics Capabilities

Our Data Analytics capabilities include:

  • Structured and Unstructured Data Store Design/Implementation
  • Search and Query Algorithm Design/Optimization
  • Data Aggregation
  • Data Visualization
  • Schema Optimization
  • ETL Design and Operations

Certifications

While innovation is a major focus at Expression, our expertise is based on recognized professional standards. Our certifications include, but are not limited to:

  • Cloudera Certified Professional – Data Scientist
  • MongoDB Certified Developer/DBA Associate
  • Certified Analytics Professional – INFORMS
  • Elasticsearch OEM and Technology Partner

Data Analytics Customers

  • Defense Information Systems Agency (DISA)
  • U.S. Department of Homeland Security
  • U.S. Department of Treasury
  • U.S. Department of Defense
  • U.S. Department of Army
  • U.S. Department of Navy

 

Machine Learning / Artificial Intelligence Capabilities and Tools

 Tensorflow

  • Machine learning library created by Google
  • Supports computation using tensor flow graph
  • Complex neural network model development
  • Deep learning supported across N distributed nodes
  • Developed in C++, provides C bindings
  • Principal development API is in Python
  • Easily deployable in both Python and Go

 Scikit-learn/Numpy/matplotlib

  • Python library for supervised and unsupervised machine learning model development
  • Aids dimensionality reduction and feature engineering: principal component analysis and matrix decomposition
  • Algorithms used: Support Vector Machines (SVM); K-means; and linear methods

DeepDive

  • Framework for applying structure to “dark” data
  • Accepts user-defined functions to implement adherent schemas
  • Accepts rulset logic (heuristic or statistically-generated) for input parameters
  • Generates probabilistic graphical model on training data for application to live data

Use Cases

Document classification:

  • Application accepts stream of unlabeled, unstructured documents
  • Runs NPL techniques to classify documents of particular type or source

Compliance check:

  • Given document type, use SVM against document input dimensions to classify as in/out of compliance

Outlier detection:

  • Application accepts stream of structured/unstructured log events
  • Determines whether individual event or set of events falls in the “acceptable” usage pattern
  • Acceptable usage pattern determined by operational history and prior classification
  • Tuning allows optimization for use-based determination of true positives/true negatives

Structuring unstructured data:

  • Application accepts stream of unstructured plaintext documents
  • Applies schema and probabilistic graphical model from deep dive to generate structure
  • Outputs structured, parsed data in relational format