# US 12,009,100 B2 · Predictive monitoring of the glucose-insulin endocrine metabolic regulatory system > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 12,009,100 B2 - Document ID: US12009100B2 - Status: Granted - Kind: B2 - Application number: 16/985,759 - Filed: 2020-08-05 - Published: 2022-02-10 - Granted: 2024-06-11 - Earliest claimed: 2020-08-05 - Assignees: UnitedHealth Group Incorporated - Applicants: UnitedHealth Group Incorporated - Inventors: Rachel Lauren Jennings; Jonathan Michael Rolfs; Alex Taub Bacon - Web page: https://patentgrep.com/patent/US12009100B2 ## CPC - G16H 50/20: for computer-aided diagnosis, e.g. based on medical expert systems - G06N 3/0442: characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] - G06N 3/0455: Auto-encoder networks; Encoder-decoder networks - G06N 3/0464: Convolutional networks [CNN, ConvNet] - G06N 3/08: Learning methods - G06N 3/09: Supervised learning - G06N 20/00: Machine learning - G16H 10/20: for electronic clinical trials or questionnaires - G16H 20/10: relating to drugs or medications, e.g. for ensuring correct administration to patients - G16H 20/30: relating to physical therapies or activities, e.g. physiotherapy, acupressure or exercising - G16H 20/90: relating to alternative medicines, e.g. homeopathy or oriental medicines - G16H 50/70: for mining of medical data, e.g. analysing previous cases of other patients - G06N 3/044: Recurrent networks, e.g. Hopfield networks - G06N 3/045: Combinations of networks - G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA - G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS ## Abstract There is a need for more effective and efficient predictive data analysis, such as more effective and efficient data analysis solutions for performing predictive monitoring of the glucose-insulin endocrine metabolic regulatory system. Certain embodiments utilize systems, methods, and computer program products that perform predictive data analysis by utilizing at least one of glucose surge excursion detections, steady-state glucose-insulin machine learning models, and parameter space refinement machine learning models. ## Claims ### Claim 1 A computer-implemented method comprising: receiving, by one or more processors, a model utility data object associated with a non-optimized machine learning model, wherein (i) the model utility data object maps a parameter value combination of a plurality of parameter value combinations defining a parameter space associated with the non-optimized machine learning model to a utility measurement of one or more utility measurements and (ii) the parameter value combination of the plurality of parameter value combinations comprises a parameter value for an optimizable parameter of one or more optimizable parameters associated with the non-optimized machine learning model; identifying, by the one or more processors and using a parameter space refinement machine learning model, a reduced parameter space from the model utility data object, wherein the reduced parameter space comprises a subrange of parameter values from the parameter space; and training, using an expediated local optimization technique, the non-optimized machine learning model to generate an optimized machine learning model configured to predict an inferred measurement from one or more recorded values by: determining an estimated optimum utility measurement from an intra-region utility measurement subset comprising a plurality of intra-region utility measurements for a plurality of parameter values within the reduced parameter space, and updating the non-optimized machine learning model based at least in part on the estimated optimum utility measurement. ### Claim 2 (depends on 1) The computer-implemented method of claim 1, wherein the parameter space refinement machine learning model is a convolutional neural network model that is configured to process an image representation of the model utility data object to determine an optimum image region of the image representation; and the reduced parameter space is determined based at least in part on the optimum image region. ### Claim 3 (depends on 1) The computer-implemented method of claim 1, wherein the parameter space refinement machine learning model is a recurrent neural network model configured to process the one or more utility measurements in a sequential manner to generate a final hidden state for an ultimate timestep of the recurrent neural network model; the final hidden state is used to generate an encoded representation of the reduced parameter space; and the reduced parameter space is determined based at least in part on the encoded representation. ### Claim 4 (depends on 1) The computer-implemented method of claim 1, wherein: the estimated optimum utility measurement is based on an intra-region optimum utility measurement; the intra-region optimum utility measurement is an element of the intra-region utility measurement subset; the intra-region utility measurement subset is a subset of the one or more utility measurements that are associated with the parameter value combination; and the intra-region optimum utility measurement is an intra-region global optimum utility measurement for the intra-region utility measurement subset. ### Claim 5 (depends on 4) The computer-implemented method of claim 4, wherein the intra-region global optimum utility measurement is determined using a holistic traversal of the intra-region utility measurement subset. ### Claim 6 (depends on 1) The computer-implemented method of claim 1, wherein: the estimated optimum utility measurement is based on an intra-region optimum utility measurement; the intra-region optimum utility measurement is an element of the intra-region utility measurement subset; the intra-region utility measurement subset is a subset of the one or more utility measurements that are associated with the parameter value combination; and the intra-region optimum utility measurement is an intra-region local optimum utility measurement for the intra-region utility measurement subset. ### Claim 7 (depends on 6) The computer-implemented method of claim 6, wherein the intra-region local optimum utility measurement is determined using a gradient-descent-based traversal of the intra-region utility measurement subset. ### Claim 8 A computing system comprising one or more processors and memory including a computer program code, the memory and the computer program code configured to, with the one or more processors, cause the computing system to: receive a model utility data object associated with a non-optimized machine learning model, wherein (i) the model utility data object maps a parameter value combination of a plurality of parameter value combinations defining a parameter space associated with the non-optimized machine learning model to a utility measurement of one or more utility measurements; and (ii) the parameter value combination of the plurality of parameter value combinations comprises a parameter value for an optimizable parameter of one or more optimizable parameters associated with the non-optimized machine learning model; identify, using a parameter space refinement machine learning model, a reduced parameter space from the model utility data object, wherein the reduced parameter space comprises a subrange of parameter values from the parameter space; and train, using an expediated local optimization technique, the non-optimized machine learning model to generate an optimized machine learning model configured to predict an inferred measurement from one or more recorded values by: determining an estimated optimum utility measurement from an intra-region utility measurement subset comprising a plurality of intra-region utility measurements for a plurality of parameter value within the reduced parameter space, and updating the non-optimized machine learning model based at least in part on the estimated optimum utility measurement. ### Claim 9 (depends on 8) The computing system of claim 8, wherein: the parameter space refinement machine learning model is a convolutional neural network model that is configured to process an image representation of the model utility data object to determine an optimum image region of the image representation; and the reduced parameter space is determined based at least in part on the optimum image region. ### Claim 10 (depends on 8) The computing system of claim 8, wherein: the parameter space refinement machine learning model is a recurrent neural network model configured to process the one or more utility measurements in a sequential manner to generate a final hidden state for an ultimate timestep of the recurrent neural network model; the final hidden state is used to generate an encoded representation of the reduced parameter space; and the reduced parameter space is determined based at least in part on the encoded representation. ### Claim 11 (depends on 8) The computing system of claim 8, wherein the estimated optimum utility measurement is based on an intra-region optimum utility measurement of the intra-region utility measurement subset of the one or more utility measurements that are each associated with the parameter value combination and the intra-region optimum utility measurement is an intra-region global optimum utility measurement for the intra-region utility measurement subset. ### Claim 12 (depends on 11) The computing system of claim 11, wherein the intra-region global optimum utility measurement is determined using a holistic traversal of the intra-region utility measurement subset. ### Claim 13 (depends on 11) The computing system of claim 11, wherein the intra-region optimum utility measurement is an intra-region local optimum utility measurement for the intra-region utility measurement subset. ### Claim 14 (depends on 13) The computing system of claim 13, wherein the estimated optimum utility measurement is based on an intra-region optimum utility measurement of the intra-region utility measurement subset of the one or more utility measurements that are each associated with the parameter value combination and the intra-region local optimum utility measurement is determined using a gradient-descent-based traversal of the intra-region utility measurement subset. ### Claim 15 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: receive a model utility data object associated with a non-optimized machine learning model, wherein (i) the model utility data object maps a parameter value combination of a plurality of parameter value combinations defining a parameter space associated with the non-optimized machine learning model to a utility measurement of one or more utility measurements; and (ii) the parameter value combination of the plurality of parameter value combinations comprises a parameter value for an optimizable parameter of one or more optimizable parameters associated with the non-optimized machine learning model; identify, using a parameter space refinement machine learning model, a reduced parameter space from the model utility data object, wherein the reduced parameter space comprises a subrange of parameter values from the parameter space; and train, using an expediated local optimization technique, the non-optimized machine learning model to generate an optimized machine learning model configured to predict an inferred measurement from one or more recorded values by: determining an estimated optimum utility measurement from an intra-region utility measurement subset comprising a plurality of intra-region utility measurements for a plurality of parameter value within the reduced parameter space, and updating the non-optimized machine learning model based at least in part on the estimated optimum utility measurement. ### Claim 16 (depends on 15) The one or more non-transitory computer-readable storage media of claim 15, wherein the parameter space refinement machine learning model is a convolutional neural network model that is configured to process an image representation of the model utility data object to determine an optimum image region of the image representation; and the reduced parameter space is determined based at least in part on the optimum image region. ### Claim 17 (depends on 15) The one or more non-transitory computer-readable storage media of claim 15, wherein: the parameter space refinement machine learning model is a recurrent neural network model configured to process the one or more utility measurements in a sequential manner to generate a final hidden state for an ultimate timestep of the recurrent neural network model; the final hidden state is used to generate an encoded representation of the reduced parameter space; and the reduced parameter space is determined based at least in part on the encoded representation. ### Claim 18 (depends on 15) The one or more non-transitory computer-readable storage media of claim 15, wherein the estimated optimum utility measurement is based on an intra-region optimum utility measurement of the intra-region utility measurement subset of the one or more utility measurements that are each associated with the parameter value combination and the intra-region optimum utility measurement is an intra-region global optimum utility measurement for the intra-region utility measurement subset. ### Claim 19 (depends on 18) The one or more non-transitory computer-readable storage media of claim 18, wherein the intra-region global optimum utility measurement is determined using a holistic traversal of the intra-region utility measurement subset. ### Claim 20 (depends on 15) The one or more non-transitory computer-readable storage media of claim 15, wherein the estimated optimum utility measurement is based on an intra-region optimum utility measurement of the intra-region utility measurement subset of the one or more utility measurements that are each associated with the parameter value combination and the intra-region optimum utility measurement is an intra-region local optimum utility measurement for the intra-region utility measurement subset.