Systems and methods for probabilistic forecasting of extremes

Patent metadata

Field	Value
Publication number	US-12412111-B2
Application number	US-202217843129-A
Country	US
Kind code	B2
Filing date	Jun 17, 2022
Priority date	Jun 17, 2022
Publication date	Sep 9, 2025
Grant date	Sep 9, 2025

Abstract

A computer-implemented method for producing probabilistic forecasts of extreme values. The method comprises obtaining input data comprising a plurality of signals of interest and a plurality of covariates associated therewith, each covariate of the plurality of covariates having an associated data type. The method further comprises performing a first forecast based on the input data. Performing the first forecast comprises: obtaining one or more trained machine learning models, each trained machine learning model of the one or more trained machine learning models having been trained to map one or more covariates of a respective data type to one or more surrogate covariates; mapping, using the one or more trained machine learning models and the input data, the plurality of covariates to one or more surrogate covariates, the one or more surrogate covariates corresponding to a compressed representation of the input data; fitting a statistical model of extremes to the plurality of signals of interest and the one or more surrogate covariates thereby generating a fitted statistical model of extremes, the statistical model of extremes being defined according to a predetermined distribution having a plurality of parameters; and obtaining a probabilistic forecast of future extreme values based on the fitted statistical model of extremes for one or more future time steps. The method further comprises causing control of a controllable system based at least in part on the probabilistic forecast of future extremes.

First claim

What is claimed is:

1. A computer-implemented method for producing probabilistic forecasts of extreme values, the method comprising:
(a) obtaining, from one or more sensors of a controllable system, input data comprising a plurality of signals of interest and a plurality of covariates associated therewith, each covariate of the plurality of covariates having an associated data type;
(b) performing a first forecast based on the input data, wherein performing the first forecast comprises:
(c) obtaining one or more trained machine learning models, each trained machine learning model of the one or more trained machine learning models having been trained to map one or more covariates of a respective data type to one or more surrogate covariates;
(d) mapping, using the one or more trained machine learning models and the input data, the plurality of covariates to one or more surrogate covariates, the one or more surrogate covariates corresponding to a compressed representation of the input data;
(e) fitting a statistical model of extremes to the plurality of signals of interest and the one or more surrogate covariates thereby generating a fitted statistical model of extremes, the statistical model of extremes being defined according to a predetermined distribution having a plurality of parameters; and
(f) obtaining a probabilistic forecast of future extreme values based on the fitted statistical model of extremes for one or more future time steps; and
(g) causing control of the controllable system based at least in part on the probabilistic forecast of future extremes.

2. The computer-implemented method of claim 1 wherein the controllable system comprises a glucose monitoring system such that the probabilistic forecast of future extreme values corresponds to a predicted high or low blood glucose level at the one or more future time steps.

3. The computer-implemented method of claim 2 wherein the one or more sensors comprise a blood glucose level sensor such that the plurality of signals of interest comprise blood glucose levels obtained from the blood glucose level sensor.

4. The computer-implemented method of claim 1 wherein the controllable system comprises an environmental monitoring system such that the probabilistic forecast of future extreme values corresponds to a predicted high concentration of one or more atmospheric pollutants at the one or more future time steps.

5. The computer-implemented method of claim 4 wherein the one or more sensors comprise an atmospheric pollutant sensor such that the plurality of signals of interest comprise atmospheric pollutant levels obtained from the atmospheric pollutant sensor.

6. The computer-implemented method of claim 1 wherein fitting the statistical model of extremes comprises:
determining the plurality of parameters by estimating the latent functional dependency of the plurality of parameters upon the one or more surrogate covariates.

7. The computer-implemented method of claim 6 wherein the latent functional dependency of the plurality of parameters upon the one or more surrogate covariates is estimated using one or more machine learning models.

8. The computer-implemented method of claim 7 wherein the one or more machine learning models include an artificial neural network having been trained to estimate the plurality of parameters from the plurality of signals of interest and the one or more surrogate covariates.

9. The computer-implemented method of claim 7 wherein the one or more machine learning models is a regression model comprising a plurality of regression parameters, wherein the one or more surrogate covariates correspond to explanatory variables of the regression model and the plurality of parameters correspond to dependent variables of the regression model.

10. The computer-implemented method of claim 9 wherein a joint posterior density of the plurality of regression parameters is estimated using a Markov Chain Monte Carlo, MCMC, estimator, wherein the MCMC estimator obtains the joint posterior density of the plurality of regression parameters by combining a joint prior density of the plurality of regression parameters with the likelihood of the plurality of regression parameters based on the plurality of observed signals of interest and the one or more surrogate covariates.

11. The computer-implemented method of claim 10 further comprising, after performing the first forecast:
obtaining further input data; and
performing a second forecast based on the further input data;
wherein the joint prior density of the plurality of regression parameters used during the second forecast corresponds to the joint posterior density of the plurality of regression parameters estimated during the first forecast.

12. The computer-implemented method of claim 1 wherein the predetermined distribution of the statistical model of extremes corresponds to a multivariate generalised Pareto distribution specified using one or more dependence functions such that the plurality of parameters incorporate parameters of the dependence function.

13. A system for producing probabilistic forecasts of extreme values, the system comprising:
a tracking device comprising at least one sensor;
an output unit configured to provide an output to a user; and
a control device communicatively coupled to the tracking device and the output unit, the control device comprising:
a forecasting unit configured to carry out the steps claim 1 ; and
an alerting unit configured to:
analyse the probabilistic forecast of future extreme values produced by the forecasting unit; and
in accordance with a determination that the probabilistic forecast of future extreme values is indicative of a predicted extremal state occurring at the one or more future time steps, cause an alert to be issued, via the output unit, to notify the user of the predicted extremal state.

14. The system of claim 13 wherein the at least one sensor of the tracking device is configured to measure a blood glucose level of a user of the tracking device such that the predicted extremal state corresponds to a predicted high or low blood glucose level of the user of the tracking device at the one or more future time steps.

15. The system of claim 13 wherein the at least one sensor of the tracking device is configured to measure a concentration of one or more atmospheric pollutants such that the predicted extremal state corresponds to a predicted high concentration of one or more atmospheric pollutants at the one or more future time steps.

16. A non-transitory computer-readable medium comprising instructions which, when executed by one or more processors, cause the one or more processors to carry out the steps of:
(a) obtaining input data comprising a plurality of signals of interest and a plurality of covariates associated therewith, each covariate of the plurality of covariates having an associated data type;
(b) performing a first forecast based on the input data, wherein performing the first forecast comprises:
(c) mapping, using one or more trained machine learning models, the input data to a plurality of parameters of a statistical model of extremes, wherein the one or more trained machine learning models comprise a trained neural network comprising a plurality of dense layers including a final layer which outputs the scale parameter of the GPD, the shape parameter of the GPD, and an exceedance probability, and wherein the statistic model of extremes is defined by a generalised Pareto distribution, GPD, such that the one or more trained machine learning models map the input data to a scale parameter and

Assignees

• Univ Leeds Innovations Ltd

Inventors

• Bogachev Leonid
• Gyarmati-Szabó János

Classifications

• G06F17/18

  for evaluating statistical data {, e.g. average values, frequency distributions, probability functions, regression analysis (forecasting specially adapted for a specific administrative, business or logistic context G06Q10/04)} · CPC title

• G06N3/047

  Probabilistic or stochastic networks · CPC title

• G05B13/048

  using a predictor · CPC title

• G05B13/027 Primary

  using neural networks only · CPC title

• G06N7/01 Primary

  Probabilistic graphical models, e.g. probabilistic networks · CPC title