Methods and apparatus for artificial intelligence models and feature engineering to predict events

1. A method comprising:

determining, by a processor, a probability of occurrence of one or more features within a high dimensional patient dataset;

profiling, by the processor, the one or more features in accordance with their respective probability of occurrence;

executing, by the processor, a feature generation model to select at least one feature from the profiled features and a corresponding time window for the at least one feature;

executing, by the processor, a time search model to select at least one time interval from a set of time intervals that includes time intervals associated with the profiled features or the at least one feature;

executing, by the processor, a meta-learning model to calculate a fitness score based on the at least one feature and the at least one time interval;

using, by the processor, responsive to the at least one feature having the fitness score that satisfies a threshold, the at least one feature to predict an event associated with a patient; and

training, by the processor, a neural network using the profiled features, the corresponding time windows, and the fitness score that satisfies the threshold.

2. The method of claim 1 , further comprising:

executing, by the processor, a classifier to predict an importance score for the at least one feature.

3. The method of claim 1 , further comprising:

training, by the processor, a machine learning model to predict an importance score for the at least one feature.

4. The method of claim 1 , wherein the feature generation model comprises a feature store configured to store the at least one feature.

5. The method of claim 1 , wherein the meta-learning model iteratively calculates the fitness score for the at least one feature for each time interval within the set of time intervals.

6. The method of claim 1 , wherein the high dimensional patient dataset comprises data associated with one or more patient journey events comprising at least one of a doctor visit, a lab test, a pharmacy prescription, a hospitalization record, a diagnosis, a medication, or a procedure associated with the patient.

7. The method of claim 1 , wherein the feature generation model uses a recency aggregator indicating a time since last occurrence of the at least one feature within the high dimensional patient dataset.

8. The method of claim 1 , wherein the feature generation model uses a count aggregator indicating a number of occurrences of the at least one feature within the high dimensional patient dataset.

9. The method of claim 1 , wherein the feature generation model comprises a genetic algorithm.

10. A computer system comprising:

a server having a processor to communicate with a data repository storing a high dimensional patient dataset, the server configured to:

determine a probability of occurrence of one or more features within the high dimensional patient dataset;

profile the one or more features in accordance with their respective probability of occurrence;

execute a feature generation model to select at least one feature from the profiled features and a corresponding time window for the at least one feature;

execute a time search model to select at least one time interval from a set of time intervals that includes time intervals associated with the profiled features or the at least one feature;

execute a meta-learning model to calculate a fitness score based on the at least one feature and the at least one time interval;

use, responsive to the at least one feature having the fitness score that satisfies a threshold, the at least one feature to predict an event associated with a patient; and

train a neural network using the profiled features, the corresponding time windows, and the fitness score that satisfies the threshold.

11. The computer system of claim 10 , wherein the server is further configured to execute a classifier to predict an importance score for the at least one feature.

12. The computer system of claim 10 , wherein the server is further configured to train a machine learning model to predict an importance score for the at least one feature.

13. The computer system of claim 10 , wherein the feature generation model comprises a feature store configured to store the at least one feature.

14. The computer system of claim 10 , wherein the meta-learning model iteratively calculates the fitness score for the at least one feature for each time interval within the set of time intervals.

15. The computer system of claim 10 , wherein the high dimensional patient dataset comprises data associated with one or more patient journey events comprising at least one of a doctor visit, a lab test, a pharmacy prescription, a hospitalization record, a diagnosis, a medication, or a procedure associated with the patient.

16. The computer system of claim 10 , wherein the feature generation model uses a recency aggregator indicating a time since last occurrence of the at least one feature within the high dimensional patient dataset.

17. The computer system of claim 10 , wherein the feature generation model uses a count aggregator indicating a number of occurrences of the at least one feature within the high dimensional patient dataset.

18. The computer system of claim 10 , wherein the feature generation model comprises a genetic algorithm.

19. A computer system comprising:

a server comprising a processor and a non-transitory computer-readable medium containing instructions that when executed by the processor causes the processor to perform operations comprising:

determining a probability of occurrence of one or more features within a high dimensional patient dataset;

profiling the one or more features in accordance with their respective probability of occurrence;

executing a feature generation model to select at least one feature from the profiled features and a corresponding time window for the at least one feature;

executing a time search model to select at least one time interval from a set of time intervals that includes time intervals associated with the profiled features or the at least one feature;

executing a meta-learning model to calculate a fitness score based on the at least one feature and the at least one time interval;

using responsive to the at least one feature having the fitness score that satisfies a threshold, the at least one feature to predict an event associated with a patient; and

training a neural network using the profiled features, the corresponding time windows, and the fitness score that satisfies the threshold.

20. The computer system of claim 19 , wherein the instructions further cause the processor to execute a classifier to predict an importance score for the at least one feature.