Supervised Machine Learning

Supervised Machine Learning 

 

SIGINT’s Machine Learning technology can review up to 1,000,000 documents an hour with unbelievably high levels of accuracy

 

How it works?

SIGINT’s Machine Learning process uses example documents and machine learning to categorize large sets of documents.  A subject matter expert will go through the documents and identify which documents are relevant.  The adaptive machine learning then examines these now categorized documents and extracts predictive “rules” or models to classify other documents accordingly.  Every machine learning instance requires sample documents as its training set. The training sets normally consist of random samples and the process is tracked and monitored in real time. Once the subject matter expert has trained the system, it then goes through and automatically categorizes the rest of the documents with extremely high levels of accuracy.

In order to ensure that the machines levels of precision and recall are sufficient, SIGINT provides the ability to monitor and track: 

  • Precision moving average
  • Recall moving average 
  • Total documents reviewed
  • Responsive documents
  • Global precision
  • Global recall