Measure of Effect

What are the MOE problems?

  • Effectiveness of an effort?
  • Suggested way ahead
  • Trend efforts
  • Culturally segment out efforts
  • Mission Grade (overall)

The Process

Input/Collection - use of internet harvesters, server access logs, blogs, articles, video/audio, etc…

Pre-processing - data collection and sentiment analysis, which includes formatting data for processing

Processing - linking data to efforts/campaigns - calculating effectiveness of effort, trending the efforts

Presentation - Palantir, custom application(s), etc…

  1. NLP (Natural Language Processing) - keyword identification, adjectives describe opinion = value of subject(sentiment). provide confidence in the value of the sentiment
  2. Link Analysis - Data Mining - connection of the opinion to the effort
  3. Trending - Data Mining - before and after the effort that will highlight the impact of our effort
  4. Summation - Belief Network - this represents the summary of the effort with confidence


What are the underlying algorithms that we can utilize to solve the MOE problems?

Data collected will consist of accurate/completely believable data, such server access logs. I.E. We know there were X number of visitors because we ahve the access lgo files. Other data collected will be more subjective. This data will be derived from Natural Language Process and will interpolate the sentiment of collected internet data. I.E. The sentiment of article X taken from website Y is: “the subject of the article is generally positive about yellow trucks”. The ability to differentiate between the confidence levels of these two examples of data is paramount, the server lgos we know to be 100% accurate and the sentiment analysis we believe is X% accurate and there are algorithms that do and some that don't take “confidence” into account.

In examining the accuracy of our resutls we also need to determine, to the best of our ability, the level of confidence of our accuracy. Not all equations take “confidence” into accoutn and thus need to be supplemented for accuracy to be considered.

  • Bayesian
  • Shannon's information theory
  • Dempster-Shafer


Reuse of existing Northrop Grumman Technology in the MOE IRAD:

  • Cyber Hawk
  • OME
  • COE
  • Other???
  • companies on the net
  • BrightPlanet
  • Data Miner
  • Hunter
  • Scrapinghub
  • Clearbit
  • Oxylabs
  • Parsehub
  • Visual Web Ripper

Way Ahead

Given the combination of current funding, staffing and other challenges, combining efforts with another IRAD(s) may provide a greatly added benefit to the MOE IRAD as well as the others.

  • investigate existing possible IRAD solutions
  • do the current IRADS provide the needed foundation for MOE?
wiki/projects/research/moe.txt · Last modified: 2023/01/12 20:28 by jeff
Back to top
CC Attribution-Share Alike 4.0 International
Driven by DokuWiki Recent changes RSS feed Valid CSS Valid XHTML 1.0