Automatic and blind segmentation of diverse data

1. An electronic device, comprising:

a processor;

memory, coupled to the processor, configured to store program instructions, wherein, when executed by the processor, the program instructions cause the electronic device to perform operations comprising:

accessing one or more records associated with an individual based at least in part on an identifier of the individual;

assigning the individual to a predetermined segment based at least in part on values of features associated with the predetermined segment in a population that includes the individual and values of the features in the one or more records, wherein the features are identified based at least in part on count statistics of the features in pairwise combinations of all the features for the population, the count statistics of the features comprise a number of occurrences of the features in the combinations, and the predetermined segment is based at least in part on distances computed using values of the features; and

providing a recommendation for the individual, wherein the recommendation is based at least in part on one or more predefined attributes associated with one or more other individuals in the predetermined segment;

wherein the predetermined segments comprise first individuals associated with a first type of event and second individuals associated with a second type of event and

wherein the recommendation is based at least in part on counterfactual analysis that determines factors that are predicted to change the individual from the first type of event to the second type of event.

2. The electronic device of claim 1 , wherein the distances comprise Euclidean distances.

3. The electronic device of claim 1 , wherein the predetermined segment is based at least in part on an unsupervised-learning technique.

4. The electronic device of claim 3 , wherein the unsupervised-learning technique comprises a dendrogram analysis or another clustering technique.

5. The electronic device of claim 1 , wherein the one or more predefined attributes comprise one or more interventions of the one or more other individuals.

6. The electronic device of claim 1 , wherein, after being identified based at least in part on the count statistics, the features are further identified based at least in part on inclusion of the features in a predictive model by a supervised-learning technique.

7. The electronic device of claim 6 , wherein the supervised-learning technique comprises L 1 -normalized logistic regression.

8. The electronic device of claim 6 , wherein the predictive model is trained using the segment of the population.

9. The electronic device of claim 1 , wherein the combinations of features for the population are above a noise floor or threshold.

10. A non-transitory computer-readable storage medium for use in conjunction with an electronic device, the computer-readable storage medium storing program instructions that, when executed by the electronic device, causes the electronic device to perform operations comprising:

accessing one or more records associated with an individual based at least in part on an identifier of the individual;

assigning the individual to a predetermined segment based at least in part on values of features associated with the predetermined segment in a population that includes the individual and values of the features in the one or more records, wherein the features are identified based at least in part on count statistics of the features in pairwise combinations of all the features for the population, the count statistics of the features comprise a number of occurrences of the features in the combinations, and the predetermined segment is based at least in part on distances computed using values of the features; and

providing a recommendation for the individual, wherein the recommendation is based at least in part on one or more predefined attributes associated with one or more other individuals in the predetermined segment;

wherein the predetermined segments comprise first individuals associated with a first type of event and second individuals associated with a second type of event; and

wherein the recommendation is based at least in part on counterfactual analysis that determines factors that are predicted to change the individual from the first type of event to the second type of event.

11. The non-transitory computer-readable storage medium of claim 10 , wherein the predetermined segment is based at least in part on an unsupervised-learning technique.

12. The non-transitory computer-readable storage medium of claim 10 , wherein the one or more predefined attributes comprise one or more interventions of the one or more other individuals.

13. The non-transitory computer-readable storage medium of claim 10 , wherein, after being identified based at least in part on the count statistics, the features are further identified based at least in part on inclusion of the features in a predictive model by a supervised-learning technique; and

wherein the predictive model is trained using the segment of the population.

14. The non-transitory computer-readable storage medium of claim 10 , wherein the distances comprise Euclidean distances.

15. The non-transitory computer-readable storage medium of claim 10 , wherein the combinations of features for the population are above a noise floor or threshold.

16. A method for providing a recommendation, comprising:

by an electronic device:

accessing one or more records associated with an individual based at least in part on an identifier of the individual;

assigning the individual to a predetermined segment based at least in part on values of features associated with the predetermined segment in a population that includes the individual and values of the features in the one or more records, wherein the features are identified based at least in part on count statistics of the features in pairwise combinations of all the features for the population, the count statistics of the features comprise a number of occurrences of the features in the combinations, and the predetermined segment is based at least in part on distances computed using values of the features; and

providing the recommendation for the individual, wherein the recommendation is based at least in part on one or more predefined attributes associated with one or more other individuals in the predetermined segment;

wherein the predetermined segments comprise first individuals associated with a first type of event and second individuals associated with a second type of event; and

wherein the recommendation is based at least in part on counterfactual analysis that determines factors that are predicted to change the individual from the first type of event to the second type of event.

17. The method of claim 16 , wherein the predetermined segment is based at least in part on an unsupervised-learning technique.

18. The method of claim 16 , wherein the one or more predefined attributes comprise one or more interventions of the one or more other individuals.

19. The method of claim 16 , wherein, after being identified based at least in part on the count statistics, the features are further identified based at least in part on inclusion of the features in a predictive model by a supervised-learning technique.

20. The method of claim 16 , wherein the combinations of features for the population are above a noise floor or threshold.