Machine learning models for automated selection of executable sequences
A computer system includes processor hardware configured to execute instructions from memory hardware. The instructions include training a machine learning model to generate an entity expiration likelihood output, obtaining a set of multiple database entities, and processing, by the machine learning model, feature vector inputs associated with each database entry to generate an entity expiration likelihood output. The instructions include determining a subset of the database entities having the highest entity expiration likelihood outputs, and, for each database entity in the subset, determining output impact scores for parameters of the feature vector input associated with the database entity, generating a feature list based on the determined output impact scores, and automatically selecting an executable sequence according to the entity expiration likelihood output associated with the database entity. The feature list is specific to the database entity and includes one or more of the parameters having the highest output impact scores.
1 . A computer system comprising:
memory hardware configured to store a machine learning model for automated bias correction, historical feature vector inputs, and computer-executable instructions for an artificial neural network, wherein the historical feature vector inputs include historical data structures specific to multiple historical database entities; and
artificial neural network processor hardware configured to execute the instructions on a plurality of weighted nodes, wherein the instructions include:
training the machine learning model with the historical feature vector inputs to the artificial neural network to generate an entity expiration likelihood output, wherein the training includes, for each expired entity of the multiple historical database entities:
determining an expiration date of the expired entity; and
minimizing seasonal bias in the historical data structures by generating a random offset time value for the expired entity,
wherein generating the random offset time value includes randomly selecting a day within a time period prior to the expiration date of the historical database entity, and
wherein the historical feature vector input for the expired entity includes only historical data structures that are dated prior to the expiration date of the expired entity minus the random offset time value;
obtaining a set of multiple database entities;
for each database entity in the set of multiple database entities:
obtaining structured input data specific to the database entity;
generating a feature vector input according to the structured input data;
processing, by the machine learning model, the feature vector input to generate the entity expiration likelihood output from the artificial neural network; and
storing the entity expiration likelihood output in association with the database entity;
determining a subset of the multiple database entities having the highest entity expiration likelihood outputs; and
for each database entity in the subset:
determining output impact scores for parameters of the feature vector input associated with the database entity, each output impact score indicative of an effect of the parameter on the entity expiration likelihood output for the database entity;
generating a feature list based on the determined output impact scores, wherein the feature list is specific to the database entity and includes one or more of the parameters having the highest output impact scores; and
automatically selecting an executable sequence according to the entity expiration likelihood output associated with the database entity to perform automated bias correction.
2 . The system of claim 1 wherein:
training the machine learning model includes generating a random offset time value for each non-expired entity of the multiple historical database entities;
the historical feature vector input for each non-expired entity includes only historical data structures that are dated prior to an end date of the historical data structures specific to the non-expired entity minus the random offset time value; and
a distribution of the random offset time values for the expired entities is equal to a distribution of the random offset time values for the non-expired entities.
3 . The system of claim 1 wherein generating the feature list includes, for each of the one or more parameters, removing the parameter from the feature list in response to determining that the parameter is negatively correlated with the entity expiration likelihood output for the database entity.
4 . The system of claim 1 wherein the instructions further include:
generating an archive report including the entity expiration likelihood output and feature list for each database entity in the subset; and
storing the archive report in an archive database on a periodic basis.
5 . The system of claim 1 wherein automatically selecting the executable sequence includes automatically scheduling a palliative care intervention for the database entity.
6 . The system of claim 5 wherein the palliative care intervention includes at least one of a text message intervention, an email intervention, an automated phone call intervention, and a live phone call intervention.
7 . The system of claim 1 wherein automatically selecting the executable sequence includes automatically scheduling the database entity to a palliative care case management database.
8 . The system of claim 1 wherein training the machine learning model includes:
comparing multiple entity expiration likelihood outputs of the machine learning model to the historical data structures;
determining whether an accuracy of the comparison is greater than or equal to a specified accuracy threshold;
adjusting parameters of the machine learning model or selecting a different machine learning model type for retraining the machine learning model, in response to the accuracy of the comparison being less than the specified accuracy threshold; and
saving the machine learning model for use in generating entity expiration likelihood outputs, in response to the accuracy of the comparison being greater than or equal to the specified accuracy threshold.
9 . The computer system of claim 1 wherein training the machine learning model includes:
training multiple machine learning model types simultaneously or in succession;
identifying one of the multiple machine learning model types having a highest output accuracy compared to others of the machine learning model types; and
saving the identified machine learning model type having the highest output accuracy for use in generating entity expiration likelihood outputs.
10 . The system of claim 1 wherein training the machine learning model includes:
separating portions of the historical feature vector inputs into structured training data and structured test data;
training the machine learning model using the structured training data;
testing the trained machine learning model using the structured test data;
evaluating results of testing the trained machine learning model; and
saving the machine learning model for use in generating entity expiration likelihood outputs, in response to an accuracy of the evaluated results being greater than or equal to a specified accuracy threshold.
11 . The system of claim 1 wherein training the machine learning model includes training a light gradient boosted tree model.
12 . The system of claim 1 wherein generating the feature vector input includes generating the feature vector input according to at least one of structured claim data specific to the database entity, structured demographic data specific to the database entity, structured lab test data specific to the database entity, structured event data specific to the database entity, structured social depravation index (SDI) score data specific to the database entity, structured pharmacy data specific to the database entity, and structured patient data specific to the database entity.
13 . The system of claim 1 wherein:
the time period is a six to twelve month time period prior to the expiration date of the historical database entity; and
the historical feature vector input for the expired entity includes only historical data structures that are dated within a twelve month time period prior to the expiration date of the expired entity minus the random offset time value.
14 . A computerized method for automated selection of executable sequences, the method comprising:
training a machine learning model with an artificial neural network and historical feature vector inputs to generate an entity expiration likelihood output for automated bias correction, wherein the historical feature vector inputs include historical data structures specific to multiple historical database entities, and wherein the training uses weighted nodes in the artificial neural network and includes, for each expired entity of the multiple historical database entities:
determining an expiration date of the expired entity; and
minimizing seasonal bias in the historical data structures by generating a random offset time value for the expired entity,
wherein generating the random offset time value includes randomly selecting a day within a time period prior to the expiration date of the historical database entity, and
wherein the historical feature vector input for the expired entity includes only historical data structures that are dated prior to the expiration date of the expired entity minus the random offset time value;
obtaining a set of multiple database entities;
for each database entity in the set of multiple database entities:
obtaining structured input data specific to the database entity;
generating a feature vector input according to the structured input data;
processing, by the machine learning model, the feature vector input to generate the entity expiration likelihood output; and
storing the entity expiration likelihood output in association with the database entity;
determining a subset of the multiple database entities having the highest entity expiration likelihood outputs; and
for each database entity in the subset:
determining output impact scores for parameters of the feature vector input associated with the database entity, each output impact score indicative of an effect of the parameter on the entity expiration likelihood output for the database entity;
generating a feature list based on the determined output impact scores, wherein the feature list is specific to the database entity and includes one or more of the parameters having the highest output impact scores; and
automatically selecting an executable sequence according to the entity expiration likelihood output associated with the database entity.
15 . The method of claim 14 wherein:
training the machine learning model includes generating a random offset time value for each non-expired entity of the multiple historical database entities;
the historical feature vector input for each non-expired entity includes only historical data structures that are dated prior to an end date of the historical data structures specific to the non-expired entity minus the random offset time value; and
a distribution of the random offset time values for the expired entities is equal to a distribution of the random offset time values for the non-expired entities.
16 . The method of claim 14 wherein generating the feature list includes, for each of the one or more parameters, removing the parameter from the feature list in response to determining that the parameter is negatively correlated with the entity expiration likelihood output for the database entity.
17 . The method of claim 14 further comprising:
generating an archive report including the entity expiration likelihood output and feature list for each database entity in the subset; and
storing the archive report in an archive database on a periodic basis.
18 . The method of claim 14 wherein automatically selecting the executable sequence includes automatically scheduling a palliative care intervention for the database entity.
19 . The method of claim 18 wherein the palliative care intervention includes at least one of a text message intervention, an email intervention, an automated phone call intervention, and a live phone call intervention.
20 . The method of claim 14 wherein automatically selecting the executable sequence includes automatically scheduling the database entity to a palliative care case management database.