Systems and methods for determining event probability

1. A system comprising:

one or more processors; and

computer-readable media storing first computer-executable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:

receiving, from a database and at a first time, health-related data of a person;

determining a time interval for splitting the health-related data into multiple records, wherein each of the multiple records corresponds to a second time that occurred before the first time;

generating, utilizing for a first instance of the time interval, a first record representing:

a first portion of the health-related data attributable to the first instance of the time interval; and

a first indication of whether the person was hospitalized during the first instance of the time interval;

generating, utilizing a second instance of the time interval, a second record representing:

a second portion of the health-related data attributable to the second instance of the time interval; and

a second indication of whether the person was hospitalized during the second instance of the time interval;

dissociating the first record from the second record such that the first record and the second record are indicated as being independent of each other;

storing the first record and the second record such that the first record and the second record are indicated as being dissociated;

generating a machine-learning model configured to determine a hospitalization probability;

generating a training dataset configured to train the machine-learning model, the training dataset including the dissociation of the first record and the second record, the health-related data, the first instance of the time interval, the second first instance of the time interval, the first record and, the second record;

training the machine-learning model utilizing the training dataset to identify trends between the dissociation of the first record and the second record, the health related data, the first instance of the time interval, the second first instance of the time interval, the first record, and the second record to determine if a hospitalization probability of a patient is improved;

determining, based at least in part on the trained machine-learning model, the hospitalization probability associated with people on which health-related data is available to the system; and

ranking the people based at least in part on the hospitalization probability and utilizing the trained machine-learning model.

2. The system of claim 1 , the operations further comprising:

identifying a hospitalization probability threshold;

identifying a portion of the people with hospitalization probabilities at or above the hospitalization probability threshold; and

generating an intervention recommendation for the portion of the people based at least in part on the hospitalization probabilities being at or above the hospitalization probability threshold.

3. The system of claim 1 , wherein generating the machine-learning model comprises generating the machine-learning model based at least in part on features associated with the health-related data that inform the hospitalization probability.

4. The system of claim 1 , the operations further comprising:

determining a difference between (1) the hospitalization probability determined by the machine-learning model with respect to a portion of the people and (2) hospitalization events occurring with respect to the portion of the people; and

wherein training the machine-learning model comprises training the machine-learning model based at least in part on the difference.

5. A method, comprising:

receiving, from a database and at a first time, data of a person;

determining a time interval for splitting the data into multiple records, wherein each of the multiple records corresponds to a second time that occurred before the first time;

generating, utilizing a first instance of the time interval, a first record including:

a first portion of the data attributable to the first instance of the time interval; and

a first indication of whether an event occurred with respect to the person during the first instance of the time interval;

generating, utilizing a second instance of the time interval, a second record including:

a second portion of the data attributable to the second instance of the time interval; and

a second indication of whether the event occurred with respect to the person during the second instance of the time interval;

dissociating the first record from the second record such that the first record and the second record are indicated as being independent of each other;

storing the first record and the second record such that the first record and the second record are indicated as being dissociated;

generating a machine-learning model configured to determine an event probability;

generating a training dataset configured to train the machine-learning model, the training dataset based at least in part on the dissociation of the first record and the second record, the health-related data, the first instance of the time interval, the second first instance of the time interval, the first record, and the second record;

training the machine-learning model utilizing the training dataset to identify trends between the dissociation of the first record and the second record, the health related data, the first instance of the time interval, the second first instance of the time interval, the first record, and the second record to determine if a hospitalization probability of a patient is improved;

determining, based at least in part on the trained machine-learning model, the event probability associated with a group of people; and

ranking, utilizing the trained machine-learning model, individual people in the group of people based at least in part on the event probability.

6. The method of claim 5 , further comprising:

identifying an event probability threshold;

identifying a portion of the group of people with event probabilities at or above the event probability threshold; and

generating a recommendation for the portion of the group of people based at least in part on the event probabilities being at or above the event probability threshold.

7. The method of claim 5 , wherein generating the machine-learning model comprises generating the machine-learning model based at least in part on features associated with the data that inform the event probability.

8. The method of claim 5 , further comprising:

determining a difference between (1) the event probability determined by the machine-learning model with respect to a portion of the group of people and (2) event occurrences with respect to the portion of the group of people; and

wherein training the machine-learning model comprises training the machine-learning model based at least in part on the difference.

9. The method of claim 5 , wherein the machine-learning model is associated with an area-under-the-resource-operation-curve value of at least 0.81.

10. The method of claim 5 , wherein generating the machine-learning model comprises generating the machine-learning model based at least in part on:

gradient boosting techniques; and

hyperparameter tuning of nodes associated with the event.

11. The method of claim 5 , wherein input features are associated with the data, and wherein generating the machine-learning model comprises generating the machine-learning model based at least in part on:

identifying a first group of the input features that, when utilized by the machine-learning model, increase an accuracy of the machine-learning model;

identifying a second group of the input features that, when utilized by the machine-learning model, decrease the accuracy of the machine-learning model or maintain the accuracy of the machine-learning model; and

refitting the machine-learning model to utilize the first group of the input features and refrain from utilizing the second group of the input features.

12. The method of claim 5 , further comprising:

identifying an event probability threshold;

identifying a portion of the group of people with event probabilities at or above the event probability threshold; and

causing the portion of the group of people to be accepted into a program based at least in part on the event probabilities being at or above the event probability threshold.

13. A system, comprising:

one or more processors; and

computer-readable media storing first computer-executable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:

receiving, from a database and at a first time, health-related data of a person;

determining a time interval for splitting the health-related data into multiple records, wherein each of the multiple records corresponds to a second time that occurred before the first time;

generating, utilizing a first instance of the time interval, a first record corresponding to the person, the first record including a first indication of whether an event occurred with respect to the person during a first instance of the time interval;

generating, utilizing a second instance of the time interval, a second record corresponding to the person, the second record including a second indication of whether the event occurred with respect to the person during a second instance of the time interval;

dissociating the first record from the second record such that the first record and the second record are indicated as being independent of each other;

storing the first record and the second record such that the first record and the second record are indicated as being dissociated;

generating a machine-learning model configured to determine an event probability;

generating a training dataset configured to train the machine-learning model, the training dataset based at least in part on the dissociation of the first record and the second record, the health-related data, the first instance of the time interval, the second first instance of the time interval, the first record, and the second record;

training the machine-learning model utilizing the training dataset to identify trends between the dissociation of the first record and the second record, the health related data, the first instance of the time interval, the second first instance of the time interval, the first record, and the second record to determine if a hospitalization probability of a patient is improved;

determining, based at least in part on the trained machine-learning model, the event probability associated with a group of people; and

ranking individual people in the group of people based at least in part on the event probability and utilizing the trained machine-learning-model.

14. The system of claim 13 , the operations further comprising:

identifying an event probability threshold;

identifying a portion of the group of people with event probabilities at or above the event probability threshold; and

generating a recommendation for the portion of the group of people based at least in part on the event probabilities being at or above the event probability threshold.

15. The system of claim 13 , wherein generating the machine-learning model comprises generating the machine-learning model based at least in part on one or more machine learning techniques indicating features associated with first data associated with the person that inform the event probability.

16. The system of claim 13 , the operations further comprising:

determining a difference between (1) the event probability determined by the machine-learning model with respect to a portion of the group of people and (2) event occurrences with respect to the portion of the group of people; and

wherein training the machine-learning model comprises training the machine-learning model based at least in part on the difference.

17. The system of claim 13 , wherein the machine-learning model is associated with an area-under-the-resource-operation-curve value of at least 0.81.

18. The system of claim 13 , wherein generating the machine-learning model comprises generating the machine-learning model based at least in part on:

gradient boosting techniques; and

hyperparameter tuning of nodes associated with the event.

19. The system of claim 13 , wherein input features are associated with first data of the person, and wherein generating the machine-learning model comprises generating the machine-learning model based at least in part on:

identifying a first group of the input features that, when utilized by the machine-learning model, increase an accuracy of the machine-learning model;

identifying a second group of the input features that, when utilized by the machine-learning model, decrease the accuracy of the machine-learning model or maintain the accuracy of the machine-learning model; and

refitting the machine-learning model to utilize the first group of the input features and refrain from utilizing the second group of the input features.

20. The system of claim 13 , the operations further comprising:

identifying an event probability threshold;

identifying a portion of the group of people with event probabilities at or above the event probability threshold; and

causing the portion of the group of people to be accepted into a program based at least in part on the event probabilities being at or above the event probability threshold.