AI based optimized decision making for epidemiological modeling

Patent metadata

Field	Value
Publication number	US-12424335-B2
Application number	US-202117355971-A
Country	US
Kind code	B2
Filing date	Jun 23, 2021
Priority date	Jul 8, 2020
Publication date	Sep 23, 2025
Grant date	Sep 23, 2025

Abstract

The present invention relates to an ESP decision optimization system for epidemiological modeling. ESP based modeling approach is used to predict how non-pharmaceutical interventions (NPIs) affect a given pandemic, and then automatically discover effective NPI strategies as control measures. The ESP decision optimization system comprises of a data-driven predictor, a supervised machine learning model, trained with historical data on how given actions in given contexts led to specific outcomes. The Predictor is then used as a surrogate in order to evolve prescriptor, i.e. neural networks that implement decision policies (i.e. NPIs) resulting in best possible outcomes. Using the data-driven LSTM model as the Predictor, a Prescriptor is evolved in a multi-objective setting to minimize the pandemic impact.

First claim

The invention claimed is:

1. A system for automatic discovery of non-pharmaceutical intervention (NPI) strategies to optimize one or more objectives related to an epidemiological event, comprising:
a continuously trainable predictor neural network model, P d (C, A)=O, implemented on a processor, the predictor neural network model being configured to receive input data, the input data including context information (C) and actions (A) performed in a given context, and
predict an outcome (O) based on the input data, wherein the outcome includes data for the one or more objectives;

a database for storing updated input data, wherein the updated input data is used to continuously train the predictor neural network model over time;
a prescriptor neural network model, P s (C)=A, implemented on a processor, the prescriptor neural network model being configured to
receive context information as input data, wherein the context information includes epidemiological event data; and
output actions that optimize the one or more objectives as outcomes corresponding to the context, wherein the output actions include changes to the implementation of one more non-pharmaceutical interventions (NPIs);

wherein the prescriptor neural network model is evolved over multiple generations using the predictor neural network model as a surrogate.

2. The system of claim 1 , wherein the predictor neural network model is a Long Short-Term Memory (LSTM) model.

3. The system of claim 1 , wherein the context information (C) and actions (A) data is time series data.

4. The system of claim 3 , wherein the predictor neural network model is decomposed into separate factors for context and actions.

5. The system of claim 4 , wherein each of the separate factors is a Long Short-Term Memory (LSTM) model.

6. The system of claim 1 , wherein the one or more objectives related to an epidemiological event are selected from number of cases, number of hospitalizations and number of deaths.

7. The system of claim 1 , wherein the context information including epidemiological event data is selected from the group consisting of: number of confirmed cases, number of hospitalizations, number of deaths, and number of recovered patients, per country, region, and day.

8. The system of claim 1 , wherein the one or more non-pharmaceutical interventions (NPIs) actions are selected from the group consisting of: school closing; workplace closing; cancellation of public events; restrictions on gatherings; public transportation closing; stay at home requirements; restrictions on internal movement and restrictions on international travel.

9. The system of claim 8 , wherein each of the one or more non-pharmaceutical interventions (NPIs) actions can be implemented at different levels of stringency.

10. A computer-implemented process for automatic discovery of non-pharmaceutical intervention (NPI) strategies to optimize one or more objectives related to an epidemiological event, comprising:
training a predictor neural network model, P d (C, A)=O, implemented on a processor, the predictor neural network model being configured to receive input training data, the input historical training data sets (C, A, O) including context information (C), actions (A) performed in a given context, and outcomes (O) resulting from action performed in the given context;
evolving a prescriptor neural network model, P s (C)=A, implemented on a processor, wherein the prescriptor neural network model is evolved over multiple generations using the trained predictor neural network model as a surrogate, the prescriptor neural network model being configured to
receive context information as input data, wherein the context information includes epidemiological event data; and
output actions that optimize the one or more objectives as outcomes corresponding to the received context information, wherein the output actions include changes to the implementation of one more non-pharmaceutical interventions (NPIs).

11. The computer-implemented process of claim 10 , wherein the predictor neural network model is trained with supervised methods.

12. The computer-implemented process of claim 10 , wherein the predictor model is a Long Short-Term Memory (LSTM) model.

13. The computer-implemented process of claim 10 , wherein the context information (C) and actions (A) data is time series data.

14. The computer-implemented process of claim 13 , wherein the predictor neural network model is decomposed into separate factors for context and actions.

15. The computer-implemented process of claim 14 , wherein each of the separate factors is a Long Short-Term Memory (LSTM) model.

16. The computer-implemented process of claim 10 , wherein the one or more objectives related to an epidemiological event are selected from number of cases, number of hospitalizations and number of deaths.

17. The computer-implemented process of claim 16 , wherein the context information including epidemiological event data is selected from the group consisting of: number of confirmed cases, number of hospitalizations, number of deaths, and number of recovered patients, per country, region, and day.

18. The computer-implemented process of claim 17 , wherein the one or more non-pharmaceutical interventions (NPIs) actions are selected from the group consisting of: school closing; workplace closing; cancellation of public events; restrictions on gatherings; public transportation closing; stay at home requirements; restrictions on internal movement and restrictions on international travel.

19. The computer-implemented process of claim 18 , wherein evolving a prescriptor neural network model, P s (C)=A includes:
establishing an initial population of candidate prescriptor neural network models, wherein each candidate prescriptor neural network model includes prescribed actions for a given context, the prescribed actions including recommended changes to the implementation of one or more non-pharmaceutical interventions (NPIs);
selecting a subset of candidate prescriptor neural network models from the initial population; and
evaluating, using the trained predictor neural network model as a surrogate, each candidate prescriptor neural network model in the subset in accordance with an evaluation of two NPI-related factors impacted by the prescribed actions, the two NPI-related factors including an expected number of cases according to the prescribed NPIs and the total stringency of the prescribed NPIs.

20. At least one non-transitory computer-readable medium storing instructions that, when executed by a computer, perform a process for automatic discovery of non-pharmaceutical intervention (NPI) strategies to optimize one or more objectives related to an epidemiological event, comprising:
training a predictor neural network model, P d (C, A)=O, the predictor model being configured to receive input training data, the input historical training data sets (C, A, O) including context information (C), actions (A) performed in a given context, and outcomes (O) resulting from action performed in the given context;
evolving a prescriptor neural network model, P s (C)=A, wherein the prescriptor neural network model is evolved over multiple generations using the trained predictor neural network model as a surrogate, the prescriptor neural network model being configured to
receive context information as input data, wherein the context information includes epidemiological event data, and
output actions that optimize the one or more objectives as outcomes correspondin

Assignees

• Cognizant Tech Solutions U S Corporation

Inventors

• Meyerson Elliot
• Francon Olivier

Classifications

• G06N3/048

  Activation functions · CPC title

• G06N3/045

  Combinations of networks · CPC title

• G06N20/10

  using kernel methods, e.g. support vector machines [SVM] · CPC title

• G06N5/01

  Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound · CPC title

• G06N3/086

  using evolutionary algorithms, e.g. genetic algorithms or genetic programming · CPC title