Model based data processing

Patent metadata

Field	Value
Publication number	US-10832162-B2
Application number	US-201615259866-A
Country	US
Kind code	B2
Filing date	Sep 8, 2016
Priority date	Sep 8, 2016
Publication date	Nov 10, 2020
Grant date	Nov 10, 2020

Abstract

Techniques for model based data processing are proposed. In one example, a computer-implemented method is as follows. A first model is determined from a first category of data processing models and a second model is determined from a second category of data processing models. The second category can be different from the first category in some instances. Performance of a first combined model of the first and second models can be compared with performance of the first model. A target model for processing data can be selected from the first combined model and the first model based on the comparing.

First claim

What is claimed is:

1. A computer-implemented method comprising:
determining, by an electronic device operatively coupled to a processing unit, a first model from a first category of data processing models;
determining, by the electronic device, a second model from a second category of data processing models, the second category being different from the first category;
comparing, by the electronic device, performance of a first combined model of first and second models with performance of the first model, wherein the first combined model is computed based on weighting the first and second models by a value associated with the weight parameter, wherein the weight parameter is between 0 and 1, and wherein the first model is weighted by 1 minus the weight parameter and the second model is weighted by the weight parameter; and
selecting, by the electronic device, a target model for processing data from the first combined model and the first model based on the comparing.

2. The computer-implemented method of claim 1 , wherein the selecting a target model comprises:
in response to determining that the performance of the first combined model is higher than the performance of the first model, selecting the first combined model as the target model; and
in response to determining that the performance of the first combined model is equal to or lower than the performance of the first model, selecting the first model as the target model, wherein the selecting the first combined model or the selecting the first model results in reduced processing unit processing time.

3. The computer-implemented method of claim 1 , further comprising:
determining, by the electronic device, a third model from a third category of data processing models, the third category being different from the first category and the second category;
comparing, by the electronic device, performance of a second combined model of the first combined model and the third model with the performance of the first combined model; and
in response to determining that performance of the second combined model is higher than the performance of the first combined model, selecting, by the electronic device, the second combined model as the target model.

4. The computer-implemented method of claim 1 , further comprising:
comparing, by the electronic device, performance of the target model with a defined threshold; and
in response to determining that the performance of the target model is lower than the defined threshold, increasing, by the electronic device, a number of categories of the data processing models.

5. The computer-implemented method of claim 4 , further comprising:
in response to determining that the performance of the target model is higher than or equal to the defined threshold, ceasing, by the electronic device, updating the target model.

6. The computer-implemented method of claim 1 , further comprising:
obtaining, by the electronic device, first and second categories of data processing models by aggregating data processing models selected from a group consisting of a type, a predicting result, a performance and a defined number of categories of the data processing models.

7. The computer-implemented method of claim 1 , wherein the performance represents similarity between a predicting result of a respective model and an actual result.

8. An electronic device comprising:
a processing unit; and
a memory operatively coupled to the processing unit and that stores computer executable instructions that, based on execution by the processing unit, facilitate performance of acts, comprising:
determining a first model from a first category of data processing models;
determining a second model from a second category of data processing models, the second category being different from the first category;
comparing performance of a first combined model of the first and second models with performance of the first model, wherein the first combined model is computed based on weighting the first and second models by a value associated with the weight parameter, wherein the weight parameter is between 0 and 1, and wherein the first model is weighted by 1 minus the weight parameter and the second model is weighted by the weight parameter; and
selecting a target model for processing data from the first combined model and the first model based on the comparing.

9. The electronic device of claim 8 , wherein the acts further include:
in response to determining that the performance of the first combined model is higher than the performance of the first model, selecting the first combined model as the target model; and
in response to determining that the performance of the first combined model is equal to or lower than the performance of the first model, selecting the first model as the target model.

10. The electronic device of claim 8 , wherein the acts further include:
determining a third model from a third category of data processing models, the third category being different from the first category and the second category;
comparing performance of a second combined model of the first combined model and the third model with the performance of the first combined model; and
in response to determining that performance of the second combined model is higher than the performance of the first combined model, selecting the second combined model as the target model.

11. The electronic device of claim 8 , wherein the acts further include:
comparing performance of the target model with a defined threshold; and
in response to determining that the performance of the target model is lower than the defined threshold, increasing a number of categories of the data processing models.

12. The electronic device of claim 11 , wherein the acts further include:
in response to determining that the performance of the target model is higher than or equal to the defined threshold, ceasing updating the target model.

13. The electronic device of claim 8 , wherein the acts further include:
obtaining first and second categories of data processing models by aggregating data processing models based on a type, a predicting result, performance and a defined number of categories of the data processing models.

14. The electronic device of claim 8 , wherein the performance represents a similarity between a prediction result of a respective model and an actual result, and wherein the processing unit efficiently determines the prediction result based on the selecting the target model.

15. A computer program product for model based data processing, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by an electronic device to cause the electronic device to:
determine a first model from a first category of data processing models;
determine a second model from a second category of data processing models, the second category being different from the first category;
compare performance of a first combined model of first and second models with performance of the first model, wherein the first combined model is computed based on weighting of the first model and the second model by a weight parameter, wherein the weight parameter is between 0 and 1, and wherein the first model is weighted by 1 minus the weight parameter and the second model is weighted by the weight parameter; and
select a target model for processing data from the first combined model and the first model based on the comparing.

16. The computer program product of claim 15 , wherein the program

Assignees

• IBM

Inventors

• Chen Boliang
• Chu Stephen Mingyu
• Huang Jing Chang
• Li Dong Sheng
• Yan Junchi
• Zhang Weipeng

Classifications

• G06N20/00 Primary

  Machine learning · CPC title

• G06N5/02 Primary

  Knowledge representation; Symbolic representation · CPC title

• G06N20/20

  Ensemble learning · CPC title

• G06N20/10

  using kernel methods, e.g. support vector machines [SVM] · CPC title