Information fusion predictor

• US 20070136224A1
• Filed: 12/08/2005
• Published: 06/14/2007
• Est. Priority Date: 12/08/2005
• Status: Active Grant

• Current Assignee: Northrop Grumman Systems Corporation (Northrop Grumman Corporation)
• Original Assignee: Northrop Grumman Corporation
• Inventors: Aboutalib, Omar
• Granted Patent: US 7,558,772 B2
• Time in Patent Office: Days
• US Class Current: 706/52
• CPC Class Codes: G06F 18/25 Fusion techniques
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

A method of data fusion comprises the steps of: receiving a plurality of decisions and associated confidences; determining probabilities of a set of the decisions; if the probabilities are not zero, then choosing the decision having a highest likelihood from the plurality of decisions; and if the probabilities are zero, then selecting an algorithm to be used for choosing one of the decisions.

Claims

1. A method of data fusion comprising the steps of:

receiving a plurality of decisions and associated confidences;

determining probabilities of a set of the decisions;

if the probabilities are not zero, then choosing the decision having a highest likelihood from the plurality of decisions;

if the probabilities are zero, then selecting an algorithm to be used for choosing one of the decisions; and

outputting the chosen decision.

2. The method of claim 1, wherein the algorithm is selected from a table of best algorithms.

3. The method of claim 1, further comprising the steps of:

determining a plurality of performance metrics for each of a plurality of algorithms and each of a plurality of targets; and

construct a table of best algorithms.

4. The method of claim 3, wherein the algorithms are characterized using one or more of the performance metrics.

5. The method of claim 4, wherein the performance metrics include one or more of:

an assurance metric;

a deception metric;

a probability of correct classification metric;

a credibility metric;

a value-added metric; and

a confusion metric.

6. The method of claim 4, wherein the performance metrics are determined using ground truth data.

7. The method of claim 3, further comprising the steps of:

constructing a confusion matrix for each algorithm; and

using the confusion matrix to produce a fused output.

8. The method of claim 7, wherein the step of constructing a confusion matrix comprises the steps of:

training the algorithms.

9. The method of claim 7, further comprising the step of:

scaling the confusion matrix.

10. The method of claim 1, further comprising the steps of:

comparing the confidences to a first predetermined threshold; and

declaring a not-in-library decision if none of the confidences exceeds the first predetermined threshold.

11. The method of claim 1, further comprising the steps of:

comparing a difference between two of the associated confidences to a second predetermined threshold; and

declaring no decision if the difference is less than the predetermined threshold.