Skip Navigation
Center Innovation Fund: ARC CIF

Ask-the-Expert: Minimizing Human Review for Big Data Analytics through Active Learning

Completed Technology Project

Project Introduction

Ask-the-Expert: Minimizing Human Review for Big Data Analytics through Active Learning
In order to learn the operational significance of anomalies using active learning, we will first get a ranked list of statistically significant anomalies by running a data driven anomaly detection method such as NASA’s Multiple Kernel Anomaly Detection (MKAD) or Inductive Monitoring System (IMS) on our dataset. A very small percentage of these anomalies (~5) will then be given to an SME to assess their operational significance. We will build a classifier taking only these few labeled examples. We plan to incorporate SME rationale by engineering new features as conjunctions and disjunctions of original features into the iterative learning process. Data points, about which the classifier is most uncertain, will then be presented interactively to the SME, and the classifier updated after each input from the SME. The process will continue until a desired accuracy is reached or the expert has analyzed ‘enough’ examples More »

Anticipated Benefits

Project Closeout – Executive Summary

Primary U.S. Work Locations and Key Partners

Project Library

Share this Project

Organizational Responsibility

Project Management

Project Duration

Technology Maturity (TRL)

Technology Areas

Target Destination

Supported Mission Type

A final report document may be available for this project. If you would like to request it, please contact us.

^