Supervised machine learning-based modeling of sensitivities to potential disruptions
Computer-implemented systems, methods and products for modeling sensitivities to potential disruptions by observing performances of entities in a first sub-population and a second sub-population using a machine learning model comprising a set of predictors and a binary indicator variable associated with a first entity subjected to a first event associated with the first sub-population, the machine learning model trained to predict an expected performance for the first entity based on at least one of a known attribute associated with the first entity in relation to the first event and a value of the binary indicator variable associated with the first event.
1. An improved computer-implemented treatment system comprising:
one or more processors coupled to one or more data storage devices to process or store data associated with known attributes for a plurality of entities subjected to a plurality of known events;
a machine learning model implemented as a data structure stored in the one or more data storage devices, the machine learning model trained to observe performances of the plurality of entities in a first sub-population and a second sub-population,
the machine learning model associated with a set of predictors and at least a binary indicator variable corresponding to a first entity subjected to a first event experienced by members of the first sub-population,
the machine learning model trained to predict an expected performance for the first entity based on at least one of (1) a known attribute associated with the first entity in relation to the first event and (2) a value of the binary indicator variable associated with the first event,
members of the first sub-population and the second sub-population being distinguishable based on whether one or more events have been experienced by the members of the first sub-population and the second sub-population, the members of the first sub-population experiencing the first event and the members of the second sub-population experiencing a second event,
wherein an expected performance of the first entity subjected to the first event is determined based on a sensitivity value determined for the first entity by quantifying an expected performance of a member of the first sub-population subjected to the first event according to a sensitivity index generated based on different permutations of potential outcomes for the plurality of known events according to a weighted combination of outputs of multiple sensitivity index models focused on types of disruption or predicted outcomes, the sensitivity index models being generated based on disruption type or target outcome, and
a treatment being administered to the first entity, responsive to the determined expected performance, wherein the sensitivity index predicts how sensitive a predicted outcome for the first entity is.
2. The system of claim 1 , wherein the first sub-population and the second sub-populations optimally capture a binary test of a type of outcome for the sensitivity index based on different events with different time periods associated with different known attributes.
3. The system of claim 2 , wherein the different known attributes are contemporaneous to at least one of (1) the different events causing a binary disruption in the first entity's status quo, (2) an absence of the first event, and (3) the second event being less disruptive than the first event.
4. The system of claim 3 , wherein the different known attributes have values that are non-randomized, such that the first entity experiencing the first event and a second entity experiencing the second event exhibit materially different distributions of said values.
5. The system of claim 1 , wherein the sensitivity index is generated based on at least a first disruption type or a first target outcome.
6. The system of claim 1 , wherein a known attribute is determined based on historical data associated with a population of entities having been subjected to an event.
7. The system of claim 4 , wherein the population is divided into the first sub-population and the second sub-population based on a propensity score calculated for pairs of matched entities in the population.
8. The system of claim 1 , wherein the first event comprises a first medical treatment and the second event comprises a second medical treatment not including the first medical treatment.
9. The system of claim 1 , wherein the historical data comprises one or more of age, weight, body mass index, ethnic background, socio-economic factors, and health condition.
10. The system of claim 8 , wherein the sensitivity index predicts how sensitive a predicted health outcome for the first entity is depending on a potential future disruption in the first medical treatment.
11. A computer-implemented method executed by one or more processors, the method comprising:
using one or more processors coupled to one or more data storage devices to process or store data associated with known attributes for a plurality of entities subjected to a plurality of known events,
a machine learning model implemented as a data structure stored in the one or more data storage devices, the machine learning model trained to observe performances of the plurality of entities in a first sub-population and a second sub-population,
the machine learning model associated with a set of predictors and at least a binary indicator variable corresponding to a first entity subjected to a first event experience by members of the first sub-population,
the machine learning model trained to predict an expected performance for the first entity based on at least one of (1) a known attribute associated with the first entity in relation to the first event and (2) a value of the binary indicator variable associated with the first event,
members of the first sub-population and the second sub-population being distinguishable based on whether one or more events have been experienced by the members of the first sub-population and the second sub-population, the members of the first sub-population experiencing the first event and the members of the second sub-population experiencing a second event,
determining, by the one or more processors, an expected performance of the first entity subjected to the first event is determined based on a sensitivity value determined for the first entity by quantifying an expected performance of a member of the first sub-population subjected to the first event according to a sensitivity index generated based on different permutations of potential outcomes for the plurality of known events according to a weighted combination of outputs of multiple sensitivity index models focused on types of disruption or predicted outcomes, the sensitivity index models being generated based on disruption type or target outcome, and
administering a treatment to the first entity, responsive to the determined expected performance, wherein the sensitivity index predicts how sensitive a predicted outcome for the first entity is.
12. The method of claim 11 , wherein the first sub-population and the second sub-populations optimally capture a binary test of a type of outcome for the sensitivity index based on different events with different time periods associated with different known attributes.
13. The method of claim 12 , wherein the different known attributes are contemporaneous to at least one of (1) the different events causing a binary disruption in the first entity's status quo, (2) an absence of the first event, and (3) the second event being less disruptive than the first event.
14. The method of claim 13 , wherein the different known attributes have values that are non-randomized, such that the first entity experiencing the first event and a second entity experiencing the second event exhibit materially different distributions of said values.
15. The method of claim 11 , wherein the sensitivity index is generated based on at least a first disruption type or a first target outcome.
16. The method of claim 11 , wherein a known attribute is determined based on historical data associated with a population of entities having been subjected to an event.
17. The method of claim 16 , wherein the population is divided into the first sub-population and the second sub-population based on a propensity score calculated for pairs of matched entities in the population, the first event comprises a first medical treatment and the second event comprises a second medical treatment not including the first medical treatment, the historical data comprises one or more of age, weight, body mass index, ethnic background, socio-economic factors, and health condition, and the sensitivity index predicts how sensitive a predicted health outcome for the first entity is depending on a potential future disruption in the first medical treatment.
18. A non-transitory data storage medium capable of storing instructions that when executed by one or more processors cause a system to:
process or store data associated with known attributes for a plurality of entities subjected to a plurality of known events,
a machine learning model implemented as a data structure stored in one or more data storage devices, the machine learning model trained to observe performances of the plurality of entities in a first sub-population and a second sub-population,
the machine learning model associated with a set of predictors and at least a binary indicator variable corresponding to a first entity subjected to a first event experience by members of the first sub-population,
the machine learning model trained to predict an expected performance for the first entity based on at least one of (1) a known attribute associated with the first entity in relation to the first event and (2) a value of the binary indicator variable associated with the first event,
members of the first sub-population and the second sub-population being distinguishable based on whether one or more events have been experienced by the members of the first sub-population and the second sub-population, the members of the first sub-population experiencing the first event and the members of the second sub-population experiencing a second event,
determine an expected performance of the first entity subjected to the first event is determined based on a sensitivity value determined for the first entity by quantifying an expected performance of a member of the first sub-population subjected to the first event according to a sensitivity index generated based on different permutations of potential outcomes for the plurality of known events according to a weighted combination of outputs of multiple sensitivity index models focused on types of disruption or predicted outcomes, the sensitivity index models being generated based on disruption type or target outcome, and
a treatment being applied to the first entity, responsive to the determined expected performance, wherein the sensitivity index predicts how sensitive a predicted outcome for the first entity is.
19. The data storage medium of claim 18 , wherein the first sub-population and the second sub-populations optimally capture a binary test of a type of outcome for the sensitivity index based on different events with different time periods associated with different known attributes and the different known attributes are contemporaneous to at least one of (1) the different events causing a binary disruption in the first entity's status quo, (2) an absence of the first event, and (3) the second event being less disruptive than the first event.
20. The data storage medium of claim 19 , wherein the different known attributes have values that are non-randomized, such that the first entity experiencing the first event and a second entity experiencing the second event exhibit materially different distributions of said values, the sensitivity index is generated based on at least a first disruption type or a first target outcome, a known attribute is determined based on historical data associated with a population of entities having been subjected to an event, and the population of entities is divided into the first sub-population and the second sub-population based on a propensity score calculated for pairs of matched entities in the population of entities.