Machine Learning to Augment the Fusion Process
for Data Classification

  • Dr. James Crowder, CAES AT&E
  • Dr. Michael Hirsch, ISEA TEK LLC

The fusion of feature data has the ability to greatly aid the task of data classification. However, in most situations, some features are better able to aid in the classification then others. In this research, we utilize a model-free reinforcement learning approach, coupled with Dempster-Shafer fusion calculus, to learn the subset of features to use for classification of data from multiple classes. Our approach is compared with using all features for data classification on an automobile feature data set, and the results show the benefits of our approach.

Full access to this whitepaper requires completing and submitting the following form: