Computation of receiver operating characteristic curves

Patent metadata

Field	Value
Publication number	US-10192166-B2
Application number	US-201414208302-A
Country	US
Kind code	B2
Filing date	Mar 13, 2014
Priority date	Apr 27, 2013
Publication date	Jan 29, 2019
Grant date	Jan 29, 2019

Abstract

A method of determining a false and/or a true positive rate is provided. A true count value and a false count value are initialized for probability bins. For a plurality of records, a truth of event occurrence and a probability of occurrence are read; a probability bin that includes the probability of occurrence is determined; the true count value of the determined probability bin is incremented when the truth of event occurrence indicates true; and the false count value of the determined probability bin is incremented when the truth of event occurrence indicates false. A true positive rate and a false positive rate are computed for each probability bin based on the true count value, the false count value, a determined total number of true event occurrences, and a determined total number of false event occurrences.

First claim

What is claimed is:

1. A method for finding a distributed computed solution for true positive rates and false positive rates comprising:
receiving, by each computing device of a plurality of computing devices, a value defining a number of probability bins, wherein a plurality of unsorted records is distributed across the plurality of computing devices;
initializing, by each computing device of the plurality of computing devices, a true count value for each of the defined number of probability bins;
initializing, by each computing device of the plurality of computing devices, a false count value for each of the defined number of probability bins;
(a) reading, by a current computing device, a truth of event occurrence from a current record of a subset of the plurality of unsorted records stored on the current computing device;
(b) reading, by the current computing device, a probability of occurrence from the current record;
(c) determining, by the current computing device, a probability bin of the probability bins that includes the probability of occurrence;
(d) incrementing, by the current computing device, the true count value of the determined probability bin when the truth of event occurrence indicates true;
(e) incrementing, by the current computing device, the false count value of the determined probability bin when the truth of event occurrence indicates false;
(f) repeating (a) to (e) with each record of the subset of the plurality of unsorted records stored on the current computing device as the current record;
(g) after processing the subset of the plurality of unsorted records stored on the current computing device, sending, by the current computing device, the true count value and the false count value for each probability bin of the probability bins to a master computing device,
wherein (a) to (g) is performed by each computing device of the plurality of computing devices as the current computing device;
receiving, by the master computing device, the true count value and the false count value for each probability bin of the probability bins from each computing device of the plurality of computing devices;
determining, by the master computing device, a total number of true event occurrences;
determining, by the master computing device, a total number of false event occurrences;
computing, by the master computing device, a true positive rate for each probability bin of the probability bins based on the true count value of the probability bin and the determined total number of true event occurrences;
computing, by the master computing device, a false positive rate for each probability bin of the probability bins based on the false count value of the probability bin and the determined total number of false event occurrences; and
outputting, by the master computing device, the true positive rate and the false positive rate for each probability bin of the probability bins for selection of an operating point to determine occurrence of an event.

2. The method of claim 1 , further comprising incrementing, by each computing device of the plurality of computing devices, a total true count value when the truth of event occurrence indicates true, and a total false count value when the truth of event occurrence indicates false.

3. The method of claim 1 , wherein computing the true positive rate for each probability bin of the probability bins comprises:
accumulating, by the master computing device, the true count value for successive probability bins of the probability bins;
subtracting, by the master computing device, the accumulated true count value for each probability bin from the determined total number of true event occurrences to define a number of true positives for each probability bin; and
dividing, by the master computing device, the number of true positives for each probability bin by the determined total number of true event occurrences to define the true positive rate for each probability bin.

4. The method of claim 3 , wherein computing the false positive rate for each probability bin of the probability bins comprises:
accumulating, by the master computing device, the false count value for successive probability bins of the probability bins;
subtracting, by the master computing device, the accumulated false count value for each probability bin from the determined total number of false event occurrences to define a number of false positives for each probability bin; and
dividing, by the master computing device, the number of false positives for each probability bin by the determined total number of false event occurrences to define the false positive rate for each probability bin.

5. The method of claim 1 , further comprising sending, by the master computing device, the value of the number of probability bins to define to each computing device of the plurality of computing devices.

6. The method of claim 1 , further comprising:
accumulating, by each computing device of the plurality of computing devices, the true count value for successive probability bins of the probability bins before sending the true count value and the false count value, wherein the true count value sent to the master computing device is the accumulated true count value; and
accumulating, by each computing device of the plurality of computing devices, the false count value for successive probability bins of the probability bins before sending the true count value and the false count value, wherein the false count value sent to the master computing device is the accumulated false count value.

7. The method of claim 1 , further comprising:
accumulating, by each computing device of the plurality of computing devices, the true count value for successive probability bins of the probability bins before sending the true count value and the false count value;
determining, by each computing device of the plurality of computing devices, a total number of true event occurrences; and
subtracting, by each computing device of the plurality of computing devices, the accumulated true count value for each probability bin from the determined total number of true event occurrences to define a number of true positives for each probability bin, wherein the true count value sent to the master computing device is the defined number of true positives.

8. The method of claim 7 , further comprising:
accumulating, by each computing device of the plurality of computing devices, the false count value for successive probability bins of the probability bins before sending the true count value and the false count value;
determining, by each computing device of the plurality of computing devices, a total number of false event occurrences; and
subtracting, by each computing device of the plurality of computing devices, the accumulated false count value for each probability bin from the determined total number of false event occurrences to define a number of false positives for each probability bin, wherein the false count value sent to the master computing device is the defined number of false positives.

9. The method of claim 2 , wherein the total number of true event occurrences is determined by receiving the total true count value from each computing device of the plurality of computing devices; and the total number of false event occurrences is determined by receiving the total false count value from each computing device of the plurality of computing devices.

10. The method of claim 1 , wherein the probability of occurrence for each unsorted record is determined by executing a model with other data associated with the current record prior to reading the probability of occurrence.

11. The method of claim 2 , f

Assignees

• Sas Inst Inc

Inventors

• Lewis Lawrence E

Classifications

• G06N7/01 Primary

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N7/005 Primary

  Physics · mapped topic