Methods and systems for intelligent evolutionary optimization of workflows using big data infrastructure

1. A computer-implemented method for intelligent evolutionary optimization of a workflow, comprising:

receiving, by a Big Data infrastructure from a client computer, a run request comprising workflow input parameters, an objective function, a pool of initial configuration parameters, and completion criteria;

running, by the Big Data infrastructure, multiple parallel instances of a workflow based on the pool of initial configuration parameters resulting in corresponding parallel output results;

storing, by the Big Data infrastructure, the workflow input parameters and the corresponding output results;

modeling, by the Big Data infrastructure utilizing at least one of a statistical technique and a parametric technique, a relationship between changes in the workflow input parameters and the corresponding output results to gain insights into which input changes have the greatest impact and which input changes have the least impact on quality of the output results;

determining, by the Big Data infrastructure based on at least one of the objective function and the completion criteria as compared to results of the modeling, that optimal output results have been achieved;

tracking by the Big Data infrastructure the relationship between the change in the workflow input parameters and the corresponding output results and providing decisions regarding which workflow input parameters to use during additional workflow run requests which optimize the output results to reduce time, effort and computing expenses; and

transmitting, by the Big Data infrastructure, the optimal output results and input-output variable relationships to the client computer.

2. The computer-implemented method of claim 1 , further comprising storing, by the Big Data infrastructure, the optimal output results and the workflow input parameters.

3. The computer-implemented method of claim 1 , wherein the workflow input parameters and associated output results are stored in one of a relational database, a flat file, or multiple flat files distributed across a cluster.

4. The computer-implemented method of claim 1 , wherein modeling the relationship between changes in the workflow input parameters and the corresponding output results comprises utilizing at least one of linear regression and singular value decomposition.

5. The computer-implemented method of claim 1 , further comprising, subsequent to modeling the relationship between changes in the workflow input parameters and the corresponding output results:

scoring each input parameter based on an accuracy of an associated output; and

selecting at least one input parameter to mutate based on the score.

6. The computer-implemented method of claim 1 , wherein the completion criteria comprises at least one of a predetermined output accuracy score based on the objective function, a determination that no further improvement occurs regarding the output results, and a determination that a predetermined computer run time limit expired.

7. The computer-implemented method of claim 1 , wherein the optimal output results comprise at least one of a set of scientific data, a set of biomedical images classified as cancerous or normal, a list of power generation equipment and a predicted time to next failure, a list of customers placed into customer segments, a list of finance customers and their associated probability of default scores, and customer insight data.

8. The computer-implemented method of claim 1 , further comprising, subsequent to modeling the relationship between changes in the workflow input parameters and the corresponding output results:

determining, by the Big Data infrastructure, based on at least one of the objective function and the completion criteria, that the optimal output results have not been achieved;

developing, by the Big Data infrastructure, a pool of updated workflow input parameters;

running multiple instances of the workflow based on the pool of updated workflow input parameters;

storing, by the Big Data infrastructure, the updated workflow input parameters and corresponding updated output results;

modeling, by the Big Data infrastructure utilizing at least one of a statistical technique and a parametric technique, a relationship between changes in the updated workflow input parameters and the corresponding updated output results to gain insights into which input changes have the greatest impact and which input changes have the least impact on quality of the updated output results;

determining, by the Big Data infrastructure, based on at least one of the objective function and the completion criteria as compared to results of the modeling, that the optimal output results have been achieved; and

transmitting, by the Big Data infrastructure, the optimal output results and input-output variable relationships to the client computer.

9. The computer-implemented method of claim 8 , further comprising, subsequent to modeling the relationship between changes in the updated workflow input parameters and the corresponding updated output results:

determining, by the Big Data infrastructure, based on at least one of the objective function and the completion criteria, that the optimal output results have not been achieved;

developing, by the Big Data infrastructure, a second pool of updated workflow input parameters;

running, by the Big Data infrastructure, multiple instances of the workflow based on the second pool of updated workflow input parameters;

storing, by the Big Data infrastructure, the second pool of updated workflow input parameters and corresponding second output results;

modeling, by the Big Data infrastructure utilizing at least one of a statistical technique and a parametric technique, a relationship between changes in the second pool of updated workflow input parameters and the corresponding second output results to gain insights into which input changes have the greatest impact and which input changes have the least impact on quality of the second results;

determining, by the Big Data infrastructure based on at least one of the objective function and the completion criteria as compared to results of the modeling, that optimal output results have been achieved; and

transmitting, by the Big Data infrastructure, the optimal output results and input-output variable relationships to the client computer.

10. A non-transitory computer readable medium storing instructions for intelligent evolutionary optimization of a workflow, in which the instructions are configured to cause a processor to:

receive, from a client computer, a run request comprising workflow input parameters, an objective function, a pool of initial configuration parameters, and completion criteria;

run multiple parallel instances of a workflow based on the pool of initial configuration parameters resulting in corresponding parallel output results;

store the workflow input parameters and the corresponding output results;

model, by utilizing at least one of a statistical technique and a parametric technique, a relationship between changes in the workflow input parameters and the corresponding output results to gain insights into which input changes have the greatest impact and which input changes have the least impact on quality of the output results;

determine, based on at least one of the objective function and the completion criteria as compared to results of the modeling, that optimal output results have been achieved;

track the relationship between the change in the workflow input parameters and the corresponding output results and providing decisions regarding which workflow input parameters to use during additional workflow run requests which optimize the output results to reduce time, effort and computing expenses; and

transmit the optimal output results and input-output variable relationships to the client computer.

11. The non-transitory computer readable storage medium of claim 10 , further comprising instructions configured to cause the processor to store the optimal output results.

12. The non-transitory computer readable storage medium of claim 10 , wherein the instructions for storing the workflow input parameters and associated output results further comprise instructions configured to cause the processor to store the workflow input parameters and associated output results in one of a relational database, flat file, or multiple flat files distributed across a cluster.

13. The non-transitory computer readable storage medium of claim 10 , wherein the instructions for modeling the relationship between changes in the workflow input parameters and the corresponding output results further comprise instructions configured to cause the processor to utilize at least one of linear regression and a singular value decomposition.

14. The non-transitory computer readable storage medium of claim 10 , further comprising, subsequent to the instructions for modeling the relationship between changes in the workflow input parameters and the corresponding output results, instructions configured to cause the processor to:

score each input parameter based on an accuracy of an associated output; and

select at least one input parameter to mutate based on the score.

15. The non-transitory computer readable storage medium of claim 10 , wherein the instructions for determining that optimal output results have been achieved further comprise instructions configured to cause the processor to determine at least one of that a predetermined output accuracy score based on the objective function has been met, that no further improvement has occurred regarding the output results, and that a predetermined computer run time limit expired.

16. The non-transitory computer readable storage medium of claim 10 , further comprising, subsequent to the instructions for modeling the relationship between changes in the workflow input parameters and the corresponding output results, instructions configured to cause the processor to:

determine, based on at least one of the objective function and the completion criteria, that the optimal output results have not been achieved;

develop a pool of updated workflow input parameters;

run multiple instances of the workflow based on the updated workflow input parameters;

store the updated workflow input parameters and corresponding updated output results;

model, by utilizing at least one of a statistical technique and a parametric technique, a relationship between changes in the updated workflow input parameters and the corresponding updated output results to gain insights into which input changes have the greatest impact and which input changes have the least impact on quality of the updated output results;

determine, based on at least one of the objective function and the completion criteria as compared to results of the modeling, that the optimal output results have been achieved; and

transmit the optimal output results and input-output variable relationships to the client computer.

17. The non-transitory computer readable storage medium of claim 16 , further comprising, subsequent to the instructions for modeling the relationship between changes in the updated workflow input parameters and the corresponding updated output results, instructions configured to cause the processor to:

determine, based on at least one of the objective function and the completion criteria, that the optimal output results have not been achieved;

develop a second pool of updated workflow input parameters

run multiple instances of the workflow based on the second pool of updated workflow input parameters;

store the second pool of updated workflow input parameters and corresponding second output results;

model, by utilizing at least one of a statistical technique and a parametric technique, a relationship between changes in the second pool of updated workflow input parameters and the corresponding second output results to gain insights into which input changes have the greatest impact and which input changes have the least impact on quality of the second output results;

determine, based on at least one of the objective function and the completion criteria as compared to results of the modeling, that optimal second output results have been achieved; and

transmit the optimal second output results and input-output variable relationships to the client computer.

18. A system for the intelligent evolutionary optimization of a workflow, comprising:

a client server computer; and

a Hadoop cluster comprising a plurality of server computers operably connected to the client server computer;

wherein the Hadoop cluster operates to:

receive, from the client computer, a run request comprising workflow input parameters, an objective function, a pool of initial configuration parameters, and completion criteria;

run multiple instances of a workflow many times in parallel based on the pool of initial configuration parameters resulting in corresponding parallel output results;

store the workflow input parameters and the corresponding output results;

model, by utilizing at least one of a statistical technique and a parametric technique, a relationship between changes in the workflow input parameters and the corresponding output results to gain insights into which input changes have the greatest impact and which input changes have the least impact on quality of the output results;

determine, based on at least one of the objective function and the completion criteria as compared to results of the modeling, that optimal output results have been achieved;

track the relationship between the change in the workflow input parameters and the corresponding output results and providing decisions regarding which workflow input parameters to use during additional workflow run requests which optimize the output results to reduce time, effort and computing expenses; and

transmit the optimal output results and input-output variable relationships to the client computer.

19. The system of claim 18 , wherein the Hadoop cluster, subsequent to modeling the relationship between changes in the workflow input parameters and the corresponding output results, further operates to:

determine, based on at least one of the objective function and the completion criteria, that optimal output results have not been achieved;

develop a pool of updated workflow input parameters;

run multiple instances of the workflow based on the updated workflow input parameters;

store the updated workflow input parameters and corresponding updated output results;

model, by utilizing at least one of a statistical technique and a parametric technique, a relationship between changes in the updated workflow input parameters and the corresponding updated output results to gain insights into which input changes have the greatest impact and which input changes have the least impact on quality of the updated output results;

determine, based on at least one of the objective function and the completion criteria as compared to results of the modeling, that the optimal output results have been achieved; and

transmit the optimal output results and input-output variable relationships to the client computer.

20. The system of claim 19 , wherein the Hadoop cluster, subsequent to modeling the relationship between changes in the updated workflow input parameters and the corresponding updated output results, further operates to:

determine, based on at least one of the objective function and the completion criteria, that the optimal output results have not been achieved;

develop a second pool of updated workflow input parameters

run multiple instances of the workflow based on the second pool of updated workflow input parameters;

store the second pool of updated workflow input parameters and corresponding second output results;

model, by utilizing at least one of a statistical technique and a parametric technique, a relationship between changes in the second pool of updated workflow input parameters and the corresponding second output results to gain insights into which input changes have the greatest impact and which input changes have the least impact on quality of the second output results;

determine, based on at least one of the objective function and the completion criteria as compared to results of the modeling, that optimal second output results have been achieved; and

transmit the optimal second output results and input-output variable relationships to the client computer.