Machine learning techniques for prospective event-based classification
Various embodiments of the present invention provide methods, apparatus, systems, computing devices, computing entities, and/or the like for performing predictive data analysis. Certain embodiments of the present invention utilize systems, methods, and computer program products that perform predictive data analysis by using at least one of prospective coverage score determination machine learning models and prospective event-based classification machine learning models.
1. A computer-implemented method comprising:
generating, by one or more processors and based at least in part on one or more predictive input events occurring during an observation period of a predictive timeframe associated with a predictive entity, a current coverage score for the observation period;
generating, by the one or more processors executing a prospective coverage score determination machine learning model comprising one or more neural network layers and trained based at least in part on historically observed coverage scores for one or more historical timeframes, a prospective coverage score for the predictive timeframe based at least in part on a group of prospective coverage model input features comprising an event distribution feature value associated with the one or more predictive input events and a predictive entity feature value associated with the predictive entity;
generating, by the one or more processors and based at least in part on a prospective coverage deviation measure for the prospective coverage score and a threshold coverage measure, a prospective coverage gap score for the predictive timeframe;
inputting, by the one or more processors and into a prospective event-based classification machine learning model, the current coverage score, the prospective coverage score, and the prospective coverage gap score as a group of classification input features, causing the prospective event-based classification machine learning model to generate a prospective event-based classification that is indicative of a prospective outcome associated with the predictive entity for the predictive timeframe, wherein the prospective event-based classification machine learning model is trained using a training set comprising a plurality of training entries each having a classification label; and
performing, by the one or more processors, one or more prediction-based actions based at least in part on the prospective event-based classification.
2. The computer-implemented method of claim 1 , wherein generating the current coverage score comprises:
determining a dual-event entity categorization indicator for the observation period that describes whether an event count for the one or more predictive input events satisfies a current computation threshold,
determining an observation exclusion indicator for the predictive entity based at least in part on whether the predictive entity satisfies at least one observation exclusion condition; and
in response to determining that the dual-event entity categorization indicator is a negative dual-event entity categorization indicator or that the observation exclusion indicator is an affirmative observation exclusion indicator, generating a null current coverage score.
3. The computer-implemented method of claim 1 , wherein generating the current coverage score comprises:
determining a dual-event entity categorization indicator for the observation period that describes whether an event count for the one or more predictive input events satisfies a current computation threshold,
determining an observation exclusion indicator for the predictive entity based at least in part on whether the predictive entity satisfies at least one observation exclusion condition; and
in response to determining that the dual-event entity categorization indicator is an affirmative dual-event entity categorization indicator and the observation exclusion indicator is a negative observation exclusion indicator, determining the current coverage score based at least in part on an observed ratio of an observed coverage count for the one or more predictive input events to an observation period count of the observation period.
4. The computer-implemented method of claim 1 , wherein the prospective coverage score determination machine learning model comprises an input filtering neural network layer for:
determining an observation exclusion indicator for the predictive entity based at least in part on whether the predictive entity satisfies at least one observation exclusion condition,
determining a single-event entity categorization indicator for the observation period that describes whether an event count for the one or more predictive input events satisfies a prospective computation threshold,
determining that the prospective coverage score is a null prospective coverage score in response to determining that the observation exclusion indicator is an affirmative observation exclusion indicator or that the single-event entity categorization indicator is a negative single-event entity categorization indicator, and
causing a feature extraction neural network layer of the prospective coverage score determination machine learning model to generate the group of prospective coverage model input features in response to determining that the observation exclusion indicator is the affirmative observation exclusion indicator and that the single-event entity categorization indicator is an affirmative single-event entity categorization indicator.
5. The computer-implemented method of claim 1 , wherein:
the prospective coverage score determination machine learning model comprises a regression neural network layer that is configured to determine the prospective coverage score based at least in part on the group of prospective coverage model input features, and
the regression neural network layer is trained based at least in part on one or more historically observed coverage scores for one or more historical timeframes having time durations that are equivalent to a time duration of the predictive timeframe.
6. The computer-implemented method of claim 1 , wherein the group of prospective coverage model input features further comprise an expected cross-event temporal gap measure for the predictive timeframe.
7. The computer-implemented method of claim 6 , wherein the group of classification input features comprise the expected cross-event temporal gap measure.
8. The computer-implemented method of claim 1 , wherein:
the event distribution feature value comprises a minimal prospective coverage score and a maximal prospective coverage score,
the minimal prospective coverage score reflects a predicted minimal coverage score for the predictive timeframe for minimal event coverage during a post-observation period of the predictive timeframe, and
the maximal prospective coverage score reflects a predicted maximal coverage score for the predictive timeframe for maximal event coverage during the post-observation period of the predictive timeframe.
9. A system comprising at least one processor and memory including program code, the memory and the program code configured to, with the at least one processor, cause the apparatus to at least:
generate, based at least in part on one or more predictive input events occurring during an observation period of a predictive timeframe associated with a predictive entity, a current coverage score for the observation period;
generate a prospective coverage score for the predictive timeframe using a prospective coverage score determination machine learning model comprising one or more neural network layers and trained based at least in part on historically observed coverage scores for one or more historical timeframes, based at least in part on a group of prospective coverage model input features comprising an event distribution feature value associated with the one or more predictive input events and a predictive entity feature value associated with the predictive entity;
generate, based at least in part on a prospective coverage deviation measure for the prospective coverage score and a threshold coverage measure, a prospective coverage gap score for the predictive timeframe;
input into a prospective event-based classification machine learning model, the current coverage score, the prospective coverage score, and the prospective coverage gap score as a group of classification input features, causing the prospective event-based classification machine learning model to generate a prospective event-based classification that is indicative of a prospective outcome associated with the predictive entity for the predictive timeframe, wherein the prospective event-based classification machine learning model is trained using a training set comprising a plurality of training entries each having a classification label; and
perform one or more prediction-based actions based at least in part on the prospective event-based classification.
10. The system of claim 9 , wherein to generate the current coverage score, the memory and the program code configured to, with the at least one processor, further cause the apparatus to:
determine a dual-event entity categorization indicator for the observation period that describes whether an event count for the one or more predictive input events satisfies a current computation threshold,
determine an observation exclusion indicator for the predictive entity based at least in part on whether the predictive entity satisfies at least one observation exclusion condition; and
in response to a determination that the dual-event entity categorization indicator is a negative dual-event entity categorization indicator or that the observation exclusion indicator is an affirmative observation exclusion indicator, generate a null current coverage score.
11. The system of claim 9 , wherein to generate the current coverage score, the memory and the program code configured to, with the at least one processor, further cause the apparatus to:
determine a dual-event entity categorization indicator for the observation period that describes whether an event count for the one or more predictive input events satisfies a current computation threshold,
determine an observation exclusion indicator for the predictive entity based at least in part on whether the predictive entity satisfies at least one observation exclusion condition; and
in response to a determination that the dual-event entity categorization indicator is an affirmative dual-event entity categorization indicator and the observation exclusion indicator is a negative observation exclusion indicator, determine the current coverage score based at least in part on an observed ratio of an observed coverage count for the one or more predictive input events to an observation period count of the observation period.
12. The system of claim 9 , wherein the prospective coverage score determination machine learning model comprises an input filtering neural network layer that is configured to:
determine an observation exclusion indicator for the predictive entity based at least in part on whether the predictive entity satisfies at least one observation exclusion condition,
determine a single-event entity categorization indicator for the observation period that describes whether an event count for the one or more predictive input events satisfies a prospective computation threshold,
determine that the prospective coverage score is a null prospective coverage score in response to a determination that the observation exclusion indicator is an affirmative observation exclusion indicator or that the single-event entity categorization indicator is a negative single-event entity categorization indicator, and
cause a feature extraction neural network layer of the prospective coverage score determination machine learning model to generate the group of prospective coverage model input features in response to a determination that the observation exclusion indicator is the affirmative observation exclusion indicator and that the single-event entity categorization indicator is an affirmative single-event entity categorization indicator.
13. The system of claim 9 , wherein:
the prospective coverage score determination machine learning model comprises a regression neural network layer that is configured to determine the prospective coverage score based at least in part on the group of prospective coverage model input features, and
the regression neural network layer is trained based at least in part on one or more historically observed coverage scores for one or more historical timeframes having time durations that are equivalent to a time duration of the predictive timeframe.
14. The system of claim 9 , wherein the group of prospective coverage model input features further comprise an expected cross-event temporal gap measure for the predictive timeframe.
15. The system of claim 14 , wherein the group of classification input features comprise the expected cross-event temporal gap measure.
16. The system of claim 9 , wherein:
the event distribution feature value comprises a minimal prospective coverage score and a maximal prospective coverage score,
the minimal prospective coverage score reflects a predicted minimal coverage score for the predictive timeframe for minimal event coverage during a post-observation period of the predictive timeframe, and
the maximal prospective coverage score reflects a predicted maximal coverage score for the predictive timeframe for maximal event coverage during the post-observation period of the predictive timeframe.
17. One or more non-transitory computer-readable storage media including instructions that, when executed by one or more processors, cause the one or more processors to:
generate, based at least in part on one or more predictive input events occurring during an observation period of a predictive timeframe associated with a predictive entity, a current coverage score for the observation period;
generate a prospective coverage score for the predictive timeframe using a prospective coverage score determination machine learning model comprising one or more neural network layers and trained based at least in part on historically observed coverage scores for one or more historical timeframes, based at least in part on a group of prospective coverage model input features comprising an event distribution feature value associated with the one or more predictive input events and a predictive entity feature value associated with the predictive entity;
generate, based at least in part on a prospective coverage deviation measure for the prospective coverage score and a threshold coverage measure, a prospective coverage gap score for the predictive timeframe;
input into a prospective event-based classification machine learning model, the current coverage score, the prospective coverage score, and the prospective coverage gap score as a group of classification input features, causing the prospective event-based classification machine learning model to generate a prospective event-based classification that is indicative of a prospective outcome associated with the predictive entity for the predictive timeframe, wherein the prospective event-based classification machine learning model is trained using a training set comprising a plurality of training entries each having a classification label; and
perform one or more prediction-based actions based at least in part on the prospective event-based classification.
18. The one or more non-transitory computer-readable storage media of claim 17 , wherein to generate the current coverage score, the instructions, when executed by one or more processors, further cause the one or more processors to:
determine a dual-event entity categorization indicator for the observation period that describes whether an event count for the one or more predictive input events satisfies a current computation threshold,
determine an observation exclusion indicator for the predictive entity based at least in part on whether the predictive entity satisfies at least one observation exclusion condition; and
in response to a determination that the dual-event entity categorization indicator is a negative dual-event entity categorization indicator or that the observation exclusion indicator is an affirmative observation exclusion indicator, generate a null current coverage score.
19. The one or more non-transitory computer-readable storage media of claim 17 , wherein to generate the current coverage score, the instructions, when executed by one or more processors, further cause the one or more processors to:
determine a dual-event entity categorization indicator for the observation period that describes whether an event count for the one or more predictive input events satisfies a current computation threshold,
determine an observation exclusion indicator for the predictive entity based at least in part on whether the predictive entity satisfies at least one observation exclusion condition; and
in response to a determination that the dual-event entity categorization indicator is an affirmative dual-event entity categorization indicator and the observation exclusion indicator is a negative observation exclusion indicator, determine the current coverage score based at least in part on an observed ratio of an observed coverage count for the one or more predictive input events to an observation period count of the observation period.
20. The one or more non-transitory computer-readable storage media of claim 17 , wherein the prospective coverage score determination machine learning model comprises an input filtering neural network layer that is configured to:
determine an observation exclusion indicator for the predictive entity based at least in part on whether the predictive entity satisfies at least one observation exclusion condition,
determine a single-event entity categorization indicator for the observation period that describes whether an event count for the one or more predictive input events satisfies a prospective computation threshold,
determine that the prospective coverage score is a null prospective coverage score in response to a determination that the observation exclusion indicator is an affirmative observation exclusion indicator or that the single-event entity categorization indicator is a negative single-event entity categorization indicator, and
cause a feature extraction neural network layer of the prospective coverage score determination machine learning model to generate the group of prospective coverage model input features in response to a determination that the observation exclusion indicator is the affirmative observation exclusion indicator and that the single-event entity categorization indicator is an affirmative single-event entity categorization indicator.