Management of data in a hybrid cloud for use in machine learning activities

1. A computer implemented method for managing hybrid cloud resources, the method comprising:

clustering data elements of a data set according to attribute sensitivity into clusters of data elements;

computing a resource allocation impact of transferring each of the clusters of data elements to public cloud resources;

computing an information gain associated with the transferring of each of the clusters of data elements to a machine learning data set in the public cloud resources;

allocating cloud resources according to the resource allocation impact and the resource information gain.

2. The computer implemented method according to claim 1 , further comprising determining the attribute sensitivity.

3. The computer implemented method according to claim 1 , wherein the resource allocation impact is computed according to the attribute sensitivity, data cluster size, data transfer cost, and trust of a public cloud of the public cloud resources.

4. The computer implemented method according to claim 1 , wherein the information gain is computed as a mean of mean information gains of each of the clusters of data elements of the data set.

5. The computer implemented method according to claim 1 , further comprising optimizing attribute transfer according to a reinforcement learning problem.

6. The computer implemented method according to claim 1 , further comprising transferring at least one of the clusters of data elements to the public cloud resources.

7. The computer implemented method according to claim 1 , further comprising:

determining the attribute sensitivity including a sensitivity value of each of the data elements relating to confidentiality of the data elements and a desire to maintain security of the data elements to maintain a business advantage;

optimizing attribute transfer according to a reinforcement learning problem evaluating outcomes associated with differing data attribute transfers for optimization of the allocating of the cloud resources; and

transferring the data elements to the public cloud resources.

8. A computer program product for managing hybrid cloud resources, the computer program product comprising one or more computer readable storage devices and stored program instructions on the one or more computer readable storage devices, the stored program instructions comprising:

program instructions to group at least a portion of data elements of a data set according to attribute sensitivity into a cluster of data elements;

program instructions to compute a resource allocation impact of transferring of the cluster of data elements from a private cloud entity to a public cloud entity;

program instructions to compute an information gain associated with the transferring of the cluster of data elements, the information gain is from adding the cluster to a machine learning data set in the public cloud; and

program instructions to allocate cloud resources according to the resource allocation impact and the information gain.

9. The computer program product according to claim 8 , further comprising program instructions to determine the attribute sensitivity by parsing documents in the data set to identify matches to policies and regulatory documents,

program instructions to compute the information gain of the attribute sensitivity of the cluster and a mean information gain for the cluster,

program instructions to determine that transferring the cluster is worthwhile based on the attribute sensitivity, the resource allocation impact, and the information gain, and

program instructions to transfer the cluster of data elements to the public cloud entity for use by a machine learning element.

10. The computer program product according to claim 8 , wherein the resource allocation impact is computed according to the attribute sensitivity, data cluster size, data transfer cost, and trust of the public cloud entity.

11. The computer program product according to claim 8 , wherein the information gain is computed as a Kullback-Leibler divergence representing an amount of change in the computed information gain between an initial probability distribution and a subsequent probability distribution.

12. The computer program product according to claim 8 , further comprising program instructions to optimize attribute transfer according to a reinforcement learning problem.

13. The computer program product according to claim 8 , further comprising program instructions to transfer the cluster of data elements to the public cloud entity.

14. The computer program product according to claim 8 , further comprising program instructions to:

determine the attribute sensitivity;

optimize attribute transfer according to a reinforcement learning problem; and

transfer the cluster of data elements to the public cloud entity.

15. A computer system for managing hybrid cloud resources, the computer system comprising:

one or more computer processors;

one or more computer readable storage devices; and

stored program instructions on the one or more computer readable storage devices for execution by the one or more computer processors, the stored program instructions comprising:

program instructions to group at least a portion of data elements of a data set according to attribute sensitivity into a cluster of data elements;

program instructions to compute a resource allocation impact of transferring the cluster of data elements to public cloud resources;

program instructions to compute an information gain associated with the set transferring of the cluster of data elements, the information gain being computed from a difference between an initial probability distribution for the data elements in the data set that are in a public cloud prior to the transferring of the cluster of data elements to the public cloud resources and a subsequent probability distribution for an initial data set expanded by the cluster of data elements; and

program instructions to allocate cloud resources according to the resource allocation impact and the information gain.

16. The computer system according to claim 15 , further comprising program instructions to determine the attribute sensitivity utilizing an attribute sensitivity scale to place a value on the attribute sensitivity of each of the data elements.

17. The computer system according to claim 15 , wherein the resource allocation impact is computed according to the attribute sensitivity, data cluster size, data transfer cost, and trust of a public cloud of the public cloud resources.

18. The computer system according to claim 15 , wherein the information gain is computed as a mean information gain of the cluster of data elements.

19. The computer system according to claim 15 , further comprising program instructions to optimize attribute transfer according to a reinforcement learning problem yielding a maximum value for ratio of the information gain to the resource allocation impact.

20. The computer system according to claim 15 , further comprising program instructions to transfer data elements to the public cloud resources.