Machine learning-based therapy device resupply predictions

Abstract

Techniques for improved machine learning are provided. Forecasted acquisitions data indicating predicted future acquisitions of one or more respiratory therapy systems is determined. An active devices prediction is generated based on processing at least a subset of the forecasted acquisitions data using a first trained machine learning model trained based on historical acquisitions of one or more respiratory therapy systems. A resupply effectiveness prediction is generated using a second trained machine learning model trained based on historical consumption of one or more consumables for the one or more respiratory therapy systems. A consumables prediction is generated based on the active devices prediction and the resupply effectiveness prediction.

Description

MACHINE LEARNING-BASED THERAPY DEVICE RESUPPLY PREDICTIONSCROSS-REFERENCE TO RELATED APPLICATION

[0001] This Application claims the benefit of and priority to U.S. Provisional Patent Application No. 63 / 573,760, filed on April 3, 2024, the entire contents of which are incorporated herein by reference.TECHNICAL FILED

[0002] The present disclosure relates generally to machine learning, and more particularly, to use of machine learning to predict respiratory therapy resupply behavior.

[0003] Many individuals suffer from sleep-related and / or respiratory -related disorders such as, for example, Periodic Limb Movement Disorder (PLMD), Restless Leg Syndrome (RLS), Sleep-Disordered Breathing (SDB) such as Obstructive Sleep Apnea (OSA) and Central Sleep Apnea (CSA), Cheyne-Stokes Respiration (CSR). respiratory insufficiency, Obesity Hyperventilation Syndrome (OHS), Chronic Obstructive Pulmonary Disease (COPD), Neuromuscular Disease (NMD), and chest wall disorders. These disorders are often treated using respiratory therapy systems.

[0004] Each respiratory therapy system generally has a respiratory therapy device connected to a user interface (e.g., a mask) via a conduit and optionally a connector. The user wears the user interface and is supplied a flow of pressurized air from the respiratory therapy device via the conduit. The user interface generally is a specific category and type of user interface for the user, such as direct or indirect connections for the category of user interface, and full face mask, a partial face mask, nasal mask, or nasal pillows for the Npe of user interface. In addition to the specific category and type, the user interface generally is a specific model made by a specific manufacturer, e.g., AirFit™ F20 manufactured by ResMed.

[0005] In many respiratory therapy regimes, users are expected to replace various components according to specified guidelines, such as periodically or after a defined amount of use. However, such resupply behavior is dynamic and difficult (or impossible) to predict, as, even with clearly defined guidelines, individual users can vary substantially in their preferences.

[0006] Improved systems and techniques to predict resupply behavior to improve therapy (and related systems) are needed.SUMMARY

[0007] According to some implementations of the present disclosure, a method includes: determining forecasted acquisitions data indicating predicted future acquisitions of one or morerespiratory therapy systems: generating an active devices prediction based on processing at least a subset of the forecasted acquisitions data using a first trained machine learning model trained based on historical acquisitions of one or more respiratory therapy systems; generating a resupply effectiveness prediction using a second trained machine learning model trained based on historical consumption of one or more consumables for the one or more respiratory therapy systems; and generating a consumables prediction based on the active devices prediction and the resupply effectiveness prediction.

[0008] According to some implementations of the present disclosure, a method includes: accessing historical acquisitions data indicating prior acquisitions of one or more respiratory therapy systems; accessing historical active device data associated with the one or more respiratory therapy systems, the historical active device data corresponding to respiratory therapy systems that were in an active state during one or more prior windows of time; accessing resupply effectiveness data associated with the one or more respiratory therapy systems, the resupply effectiveness data corresponding to consumption of one or more consumables for the one or more respiratory therapy systems; updating one or more parameters of a first machine learning model, based on the historical acquisitions data and the historical active device data, to predict future active devices for one or more future windows of time; and updating one or more parameters of a second machine learning model, based on the resupply effectiveness data, to predict future resupply effectiveness.

[0009] According to some implementations of the present disclosure, a system includes a control system and a memory. The control system includes one or more processors. The memory has stored thereon machine readable instructions. The control system is coupled to the memory', and any one of the methods disclosed herein is implemented when the machine executable instructions in the memory are executed by at least one of the one or more processors of the control system.

[0010] Other aspects provide processing systems configured to perform the aforementioned method as well as those described herein; non-transitory, computer-readable media comprising instructions that, when executed by one or more processors of a processing system, cause the processing system to perform the aforementioned methods as well as those described herein; a computer program product embodied on a computer-readable storage medium comprising code for performing the aforementioned methods as well as those further described herein; and a processing system comprising means for performing the aforementioned methods as well as those further described herein.

[0011] The above summary is not intended to represent each implementation or everyaspect of the present disclosure. Additional features and benefits of the present disclosure are apparent from the detailed description and figures set forth below.BRIEF DESCRIPTION OF THE DRAWINGS

[0012] So that the manner in which the above-recited features of the present disclosure can be understood in detail, a more particular description, briefly summarized above, may be had by reference to aspects, some of which are illustrated in the appended drawings. It is to be noted, however, that the appended drawings illustrate only certain typical aspects of this disclosure and are therefore not to be considered limiting of its scope, for the description may admit to other equally effective embodiments.

[0013] FIG. 1 depicts an example environment for training machine learning models and predicting therapy resupply, according to some implementations of the present disclosure.

[0014] FIG. 2 depicts an example workflow for training machine learning models to predict therapy resupply actions, according to some implementations of the present disclosure.

[0015] FIG. 3 is a flow diagram depicting an example method for predicting resupply actions using machine learning, according to some embodiments of the present disclosure.

[0016] FIG. 4 is a flow diagram depicting an example method for training machine learning models to predict resupply behavior, according to some embodiments of the present disclosure.

[0017] FIG. 5 is a flow diagram depicting an example method for predicting resupply behavior using machine learning, according to some embodiments of the present disclosure.

[0018] FIG. 6 is a flow diagram depicting an example method for training machine learning models to predict resupply, according to some embodiments of the present disclosure.

[0019] FIG. 7 is a flow diagram depicting an example method for predicting resupply using machine learning models, according to some embodiments of the present disclosure.

[0020] FIG. 8 depicts an example computing device configured to perform various aspects of the present disclosure, according to some embodiments disclosed herein.

[0021] While the present disclosure is susceptible to various modifications and alternative forms, specific implementations and embodiments thereof have been shown by way of example in the drawings and will herein be described in detail. It should be understood, however, that it is not intended to limit the present disclosure to the particular forms disclosed, but on the contrary, the present disclosure is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the present disclosure as defined by the appended claims.DETAILED DESCRIPTION

[0022] Embodiments of the present disclosure generally provide techniques for using machine learning to predict resupply behavior, such as for respirator}7therapy systems.

[0023] Generally, respiratory therapy refers to the use of a flow generator and user interface to deliver air and / or oxygen to a user, such as during sleep. For example, respiratory7therapy may include use of continuous positive airway pressure (CPAP) devices, bi-level positive airway pressure (BiPAP) devices, and the like. Respiratory therapy can significantly improve the lives of users who engage in it. In some cases, one or more components of the respiratory therapy system may be consumable (e.g., using the respiratory7therapy system consumes or otherwise reduces the life of the component(s)). In many respiratory therapy systems, users are instructed to replace some components according to a maintenance or replacement schedule (e.g., periodically, after a defined amount of use, and the like). For example, in some cases, users may replace components such as the conduit (also referred to in some aspects as a tube) used to connect the user interface (e.g., mask) and the respiratory7therapy device (e.g., the flow generator) to provide airflow to the interface, the humidifier tub or tank (e.g., the container that holds water used to humidify the generated airflow), and the like.

[0024] Replacing such components according to manufacturer suggestion may7improve therapy results, such as by ensuring that the components are not degraded (e.g., the conduit has no crushed portions, kinks, holes, or other wear-related damage that may affect airflow) and / or to ensure that the components are clean and sanitary (e.g., to ensure no adverse illness results from an infected or damaged water chamber). In some embodiments of the present disclosure, replacing such consumable components may collectively be referred to as "resupply."

[0025] In order for users to resupply adequately (e.g., when desired and / or in accordance with manufacturer or other recommendation), sufficient spare parts must of course be available. To ensure adequate part manufacturing and distribution, it may be beneficial to know (or predict) spare part usage substantially in advance of the actual need for such parts. However, as user bases shift, demand changes, and usage fluctuates over time, it becomes difficult (or impossible) to reliably predict future need. This is particularly true when new systems are developed or deployed (e.g., when new flow generator models are developed), as there is simply no data available for such devices.

[0026] In some embodiments of the present disclosure, techniques and architectures are provided to accurately forecast future spare parts usage for devices, such as respiratory therapy systems, in order to ensure adequate manufacturing and distribution, as well as to improve therapy outcomes for users. In some embodiments, a resupply effectiveness measure is definedto correlate active device usage and spare part consumption in order to facilitate improved machine learning. For example, by utilizing such a resupply effectiveness measure, embodiments of the present disclosure enable machine learning models to more accurately and reliably leam to identify trends and patterns in consumption and usage, as compared to conventional approaches. This improved data processing can result in substantially improved model performance, which may reduce computational expense of the systems (e.g., reducing the complexity of the model(s), thereby reducing the resources needed to train and / or inference using such models), as well as improving component manufacturing and distribution (e.g., reducing or preventing scarcity as well as over-abundance of such components).Example Environment for Training Machine Learning Models and Predicting Therapy Resupply

[0027] FIG. 1 depicts an example environment 100 for training machine learning models and predicting therapy resupply, according to some implementations of the present disclosure.

[0028] In the illustrated example, a machine learning system 115 accesses a set of sales data 105 and activity data 110 and trains one or more machine learning models 125. As used herein, “accessing” data may generally include receiving, retrieving, requesting, collecting, measuring, generating, obtaining, or otherwise gaining access to the data. For example, the machine learning system 115 may access the sales data 105 and / or activity data 110 from one or more storage repositories, one or more other systems that generate such data, and the like. The machine learning system 115 is generally representative of any computing system, and may be implemented using hardware, software, or a combination of hardware and softw are. For example, the machine learning system 115 may be implemented as a physical server, as a cloud-based application, and the like.

[0029] In some embodiments, the sales data 105 (also referred to in some aspects as acquisition data) generally includes information relating to historical and / or future user acquisitions for one or more components or systems, such as respiratory' therapy devices, spare parts for such devices, and the like. For example, the sales data 105 may include historical acquisitions of one or more respiratory therapy systems (e.g., indicating the number of such systems sold during one or more prior times, such as the number per month, per quarter, per year, and the like). Similarly, the sales data 105 may include forecasted or predicted sales data for one or more respiratory7therapy system (e.g., indicating the number of such systems that are expected to be sold for one or more future times, such as the next month, the next quarter, the next year, and the like). In some aspects the sales data 105 includes information for multiplemodels or versions of respiratory therapy system. For example, the sales data 105 may include historical sales data for a first model and forecasted sales data for a second (newer) model (e.g., if the second model has no historical data yet).

[0030] In some aspects, the sales data 105 may further include information related to historical sales or consumption of spare parts for such systems, such as the number of humidifier tanks, conduits, or other components that were sold or consumed during one or more prior times (e.g., the number per month, per quarter per year, and the like). In some aspects, if the sales data 105 includes data for multiple models or versions, the spare parts data may correspond to historical spare part usage (rather than future usage).

[0031] In some embodiments, the activity data 110 generally includes information relating to active devices or systems for one or more periods of time. For example, the activity data 110 may indicate the number of respiratory therapy systems (for one or more models or versions) that were in active use during one or more prior weeks, months, quarters, years, and the like. In some aspects, the activity' data 110 corresponds to or is generated based on therapydata. For example, if the respiratory therapy system(s) report their usage (e.g., for tracking by healthcare professionals, by the users themselves, and the like), the activity data 110 may be generated by evaluating this therapy' usage data to identify the number of devices that were in an active state.

[0032] In some aspects, devices may be referred to as “active'’ if usage of the device satisfies one or more defined criteria. For example, a respiratory therapy system may be referred to as “active” in a given period (e.g., a month) if the device is used in an active state (e.g., powered on and providing therapeutic airflow) for at least a minimum duration during the period. This may include, for example, a minimum total or aggregate usage during the period, a minimum continuous usage (e.g., for at least one continuous hour during the period), a minimum average usage (e.g., for at least an hour per day during the period), and the like. For example, the activity data 110 may indicate the total number of respiratory therapy systems that are considered “active” (based on usage) for each of one or more prior months.

[0033] In some aspects, as discussed below in more detail, the machine learning system 115 may train one or more machine learning models (e.g.. of the machine learning models 125) based on the sales data 105 (e.g., historical sales for a respiratory therapy system) and the activity data 110 (e.g., historical activity or usage of the respiratory' therapy systems) to predict the number of active devices for one or more periods of time based on the sales data for one or more periods of time. For example, based on historical or predicted sales of the respiratory therapy system during one or more months or quarters, the model may predict the number ofrespiratory therapy systems that will be active during one or more months or quarters.

[0034] In some embodiments, as discussed below in more detail, the machine learning system 115 may train one or more machine learning models (e.g., of the machine learning models 125) based on the sales data 105 (e.g., historical spare part sales) and the activity data 110 (e.g., historical activity7or usage of the respiratory therapy systems) to predict the spare part resupply effectiveness for one or more periods of time. For example, based on historical or predicted resupply effectiveness (which may be defined based on active devices and device sales) during one or more months or quarters, the model may predict the future resupply effectiveness during one or more months or quarters.

[0035] In some embodiments, the “resupply effectiveness” is defined as s ai eswhere Etis the resupplv effectiveness for time period t, Sparestis thenumber of spare parts of a given type (e.g., conduits, humidifier tanks, and the like) that are consumed (e.g., sold to or purchased by7users) during the time period t, and AdiveD evice stis the number of devices (which are able to use the spare part(s)) that were considered “active” during the time period t. For example, if one hundred replacement conduits were sold during a given month and one thousand respiratory therapy systems that can use such replacement conduit were in the “active” state during the given month, the resupply effectiveness for the given month may be 0.1. In some aspects, in addition to or instead of considering spare part usage and active device usage during the same time period t, different periods may be used. For example, the system may use the number of active devices from one period (e.g.. period t) and the number of spare parts used for a different period (e.g., period t + 1) to define the resupply effectiveness measure.

[0036] In some embodiments, a separate resupply effectiveness measure may be defined for each type of spare part (e.g., each type of conduit, each type of humidifier tank, and the like) and / or for each type of respiratory7therapy system (or other system that uses the spare parts). For example, a first resupply effectiveness may be determined for a given ty pe of humidifier tank that fits a given model of respiratory7therapy sy stem, while a second resupply effectiveness may be defined for a second type of humidifier tank that fits a different model, and a third resupply effectiveness may be defined for a replacement conduit that fits either model (e.g., based on the total number of active devices of either model).

[0037] In the illustrated environment 100, these machine learning models 125 may then be deployed for inferencing. As used herein, “deploying” the machine learning models 125 may generally include any operations used to prepare or provide the models for runtime use (eitherlocally by the machine learning system 115, or by one or more other systems). For example, the machine learning system 115 may store and / or transmit the learned parameters and architecture, may instantiate a model instance using the learned parameters, and the like.

[0038] As illustrated, sales data 120 (also referred to in some aspects as acquisition data) can then be processed using the machine learning model (s) 125 to generate spares predictions 130 (also referred to in some aspects as consumables predictions). In some embodiments, the sales data 120 generally includes information relating to historical and / or future sales for one or more components or systems, such as respiratory therapy devices. For example, the sales data 120 may include historical sales data of one or more respiratory7therapy systems (e.g., indicating the number of such systems sold for one or more prior times, such as the number per month, per quarter, per year, and the like). Similarly, the sales data 120 may include forecasted or predicted sales data for one or more respiratory therapy system (e.g., indicating the number of such systems that are expected to be sold for one or more future times, such as the next month, the next quarter, the next year, and the like). In some embodiments, the sales data 105 (used to train the machine learning models 125) corresponds to historical sales of a first version or model of respiratory therapy system, and the sales data 120 (processed by the models during runtime) corresponds to forecasted or predicted sales of a second version or model of the respiratory7therapy system (e.g., for a new model that is about to be released).

[0039] Generally, the sales data 120 may be generated or accessed from a variety of sources. For example, in some embodiments, the sales data 120 corresponds to forecasted sales of the system(s) or device(s) for one or more future periods of time. In some embodiments, the sales data 120 may be generated by subject matter experts based on a variety7of information and operations, such as market research, surveys, and the like.

[0040] In some embodiments, the machine learning model(s) 125 process the sales data 120 (e.g., forecasted respiratory therapy system sales for one or more periods of time) to predict the number of respiratory therapy' systems that will be in the “active” state during the one or more future periods of time. In some embodiments, the machine learning model (s) 125 can further be used to predict future resupply effectiveness measure(s) for one or more consumables during one or more future periods of time. In the illustrated example, the system can then aggregate the predicted resupply effectiveness measure(s) and the predicted active device(s) for one or more periods of time to generate a prediction of the number of spare parts that will be consumed (e.g., purchased) during the future period(s) of time. For example, given a predicted resupply effectiveness Etfor a given spare part during a future time period t and a predicted number of active devices ActiveDevicest, which can use the given spare part, duringthe future time period t, the system may compute the spares predictions 130 as Sparest= Et* ActiveD evicest, where Sparestis the number of spare parts (of the given type) that are predicted to be used or consumed during the future time period t.

[0041] In some embodiments, in addition to or instead of using predicted resupply effectiveness and predicted active device numbers for the same time period, the system(s) may use active device predictions from different periods (e.g., from one or more periods prior to and / or including time t) to predict spare part consumption during the period t.

[0042] In some embodiments, the machine learning model(s) 125 may be trained and / or used to predict future spare part consumption on a per-spare part basis, on a per-region basis, and the like. For example, the machine learning system 115 may train a separate machine learning model 125 to predict resupply effectiveness for each type of spare part, may train a separate machine learning model 125 to predict the number of active devices for each type or model of respiratory' therapy system, may train a separate machine learning model 125 to predict the number of active devices for each specific region (e.g., each country), and the like.Example Workflow for Training Machine Learning Models to Predict Therapy Resupply Actions

[0043] FIG. 2 depicts an example workflow 200 for training machine learning models to predict therapy resupply actions, according to some implementations of the present disclosure. In some embodiments, the workflow 200 is performed by a machine learning system, such as the machine learning system 1 15 of FIG. 1.

[0044] In the illustrated workflow 200, historical sales data 205 and / or activity' data 215 are used to train an active device machine learning model 225. In some embodiments, the historical sales data 205 (also referred to in some aspects as historical acquisition data) corresponds to all or a part of the sales data 105 of FIG. 1. For example, in some embodiments, the historical sales data 205 generally includes information relating to historical sales for one or more components or systems, such as respiratory' therapy devices. For example, the sales data 205 may include historical sales data of one or more respiratory therapy systems (e.g., indicating the number of such systems sold for one or more prior times, such as the number per month, per quarter, per year, and the like). Although not depicted in the illustrated example, in some embodiments, the system may similarly use forecasted or predicted sales data for one or more respiratory therapy system to train the active device machine learning model 225.

[0045] Although not depicted in the illustrated example, in some aspects, the historical sales data 205 may also include future or forecasted sales data (or the forecasted data may beused as a separate input). For example, the model may be trained based on both forecasted sales for a given period as well as actual historical sales for the given period (e.g., to allow the model to leam to compensate for disparities in the prediction(s) and reality).

[0046] In some embodiments, the activity data 215 corresponds to all or a part of the activity data 110 of FIG. 1. For example, in some embodiments, the activity data 215 generally includes information relating to active devices or systems for one or more periods of time. For example, the activity data 215 may indicate the number of respiratory therapy systems (for one or more models or versions) that were in active use during one or more prior weeks, months, quarters, years, and the like. In some aspects, the activity data 215 corresponds to or is generated based on therapy data, as discussed above. In some aspects, as discussed above, devices may be referred to or defined as "active” if usage of the device satisfies one or more defined criteria. For example, a respiratory therapy system may be referred to as “active” in a given period (e.g., a month) if the device is used in an active state (e.g., powered on and providing therapeutic airflow) for at least a minimum duration during the period.

[0047] In some embodiments, the machine learning system trains the active device machine learning model 225 using the historical sales data 205 and the activity data 215 as training exemplars or samples. For example, some or all of the historical sales data 205 may be used as the input portion of each exemplar (e.g., as the data used as input to the model) while some or all of the activity data 215 is used as the label or target output of exemplar (e.g.. used as the target output that the model should generate).

[0048] Generally, the particular operations used to train the active device machine learning model 225 may vary depending on the particular architecture of the model. For example, if the model is a regression model (e.g., a linear regression model, a neural network, a random forest model, and the like), the machine learning system may process the input portion of an exemplar (e.g., the historical sales data for one or more periods of time) as input to the model to generate a prediction. This prediction may then be compared against the label portion of the exemplar (e.g., the activity data 215 for one or more periods of time) to compute a loss, and the loss may be used to refine or update the value(s) of one or more parameter(s) of the active device machine learning model 225 (e.g., using backpropagation). This process may be repeated using any number of exemplars and for any number of iterations or epochs. Similarly, the training process can generally be performed using individual exemplars (e.g., using stochastic gradient descent) and / or using batches of exemplars (e.g., using batch gradient descent), depending on the particular implementation.

[0049] In some embodiments, the particular features (from the historical sales data 205)used as the input data may vary depending on the particular implementation. For example, in some embodiments, the input features may include the device sales from the past month, from the past two months, from the month prior to the past month (e.g., if the past month is month m, the features may include the sales from the prior month m — 1), from the past quarter, from the past two quarters and / or from the quarter prior to the past quarter, from the past year, from the past two years and / or the year prior to the past year, and the like. In some embodiments, the input data features may be determined based at least in part on the expected lifetime of the spare part(s) that are being evaluated. For example, if a component is expected to be replaced every' six months, the machine learning system may train the model to predict activity' data 215 for at least six months after the purchase period (e.g., if the input sales data corresponds to month m, the active device machine learning model 225 may be trained to predict activity data based on the input data for activity' data 215 at least through month m + 6).

[0050] As discussed above, based on this training, the active device machine learning model 225 leams to predict the number of devices that will be active during one or more time periods based on processing sales data indicating the historical and / or forecasted future sales of the device during one or more time periods.

[0051] In the illustrated workflow 200, historical spares sales data 220 and / or activity' data 215 are also used to train a resupply machine learning model 230. In some embodiments, the historical spares sales data 220 (also referred to in some aspects as historical consumables data) corresponds to all or apart of the sales data 105 of FIG. 1. For example, in some embodiments, the historical spares sales data 220 may7include information related to historical sales or consumption of spare parts for such systems, such as the number of humidifier tanks, conduits, or other components that were sold or consumed during one or more prior times (e.g., the number per month, per quarter per year, and the like).

[0052] In some embodiments, the machine learning system trains the resupply machine learning model 225 using the historical spares sales data 220 and the activity' data 215 as training exemplars or samples. In some embodiments, the machine learning system uses the activity data 215 and historical spares sales data 220 to generate one or more resupply effectiveness measures, as discussed above. For example, for each month reflected in the activity7data 215 and historical spares sales data 220, the machine learning system may compute a resupply effectiveness for the month (e g., by dividing the number of active devices for the month by the number of spares sold or consumed during the month). These resupply effectiveness measures may then be used to tram the resupply machine learning model 230.

[0053] Generally, the particular operations used to train the resupply machine learning model 230 may vary depending on the particular architecture of the model. For example, in some embodiments, the resupply machine learning model 230 is a statistical analysis model, such as an autoregressive integrated moving average (ARIMA) model, which can be trained to forecast time series data. In some embodiments, a sequence of resupply effectiveness measures (e.g., one generated for each month) may be used as a time series of values to train the resupply machine learning model 230 to predict future values in the time series (e.g., resupply effectiveness for one or more future periods). For example, one or more parameters of the resupply machine learning model 230 may be refined or updated to fit the curve or trend(s) in the resupply effectiveness training data, allowing the model to leam to account for fluctuations caused by a variety of causes such as the season, how much time has elapsed since the respiratory therapy device was released, how much time has elapsed since a given type of spare part became available, and the like.

[0054] As discussed above, based on this training, the resupply machine learning model 230 learns to predict the future resupply effectiveness for one or more spare parts or other consumables during one or more future time periods based on the historical trends reflected in the activity data 215 and historical spares sales data 220. As discussed above, the active device machine learning model 225 and the resupply machine learning model 230 may then be deployed for inferencing.Example Workflow for Predicting Resupply Actions using Machine Learning

[0055] FIG. 3 depicts an example workflow 300 for predicting resupply actions using machine learning, according to some embodiments of the present disclosure. In some embodiments, the workflow 300 is performed by a machine learning system, such as the machine learning system 115 of FIG. 1 and / or the machine learning system discussed above with reference to FIG. 2. In some embodiments, the workflow 300 is performed by one or more other systems, such as dedicated inferencing systems.

[0056] In the illustrated workflow, forecasted sales data 310 (also referred to in some aspects as forecasted acquisitions data) is used as input to the active device machine learning model 225 to generate an active devices prediction 325. In some embodiments, the forecasted sales data 310 corresponds to all or a part of the sales data 120 of FIG. 1.

[0057] For example, in some embodiments, the forecasted sales data 310 generally includes information relating to future sales for one or more components or systems, such as respiratory therapy devices. For example, the forecasted sales data 310 may include the forecasted orpredicted sales data for one or more respiratory therapy system (e.g., indicating the number of such systems that are expected to be sold for one or more future times, such as the next month, the next quarter, the next year, and the like).

[0058] Although not depicted in the illustrated example, in some aspects, the forecasted sales data 310 may also include historical sales data (or the historical data may be used as a separate input). For example, the model may be evaluate both forecasted sales (e.g., for periods that have not yet occurred) as well as historical sales (e.g., for periods that have already occurred) to allow the model to make predictions based on a combination of real and forecasted data.

[0059] In some embodiments, as discussed above, the sales used to train the active device machine learning model 225 corresponds to a first version or model of respiratory therapy system, while the forecasted sales data 310 processed during runtime corresponds to sales of a second version or model of the respiratory therapy system (e.g., for a new model that is about to be released).

[0060] In the illustrated example, the active devices prediction 325 may include one or more predictions for one or more future time periods. For example, the active devices prediction 325 may predict the number of devices that will be active (e g., used above a designated threshold, such as for at least one continuous hour per month) during one or more future time periods (e.g., one or more months).

[0061] In the illustrated workflow 300, the resupply machine learning model 230 is used to generate one or more resupply effectiveness predictions 330. Although not depicted in the illustrated example, in some embodiments, the resupply machine learning model 230 processes historical resupply effectiveness measures (e.g., generated based on the historical activity data 215 and / or the historical spares data 220, each of FIG. 2) to predict the resupply effectiveness measure(s) for one or more future time periods. For example, the resupply effectiveness prediction(s) 330 may predict the resupply effectiveness measures for one or more spare parts during one or more future time periods (e.g., one or more months). In some embodiments, the time period(s) reflected in one or more of the active devices prediction(s) 325 correspond to or match the time period(s) reflected in the resupply effectiveness prediction(s) 330.

[0062] In the illustrated workflow 300, the active devices prediction(s) 325 and the resupply effectiveness prediction(s) 330 are then accessed by an aggregation component 335, which processes them to generate one or more spare parts prediction(s) 340. In some embodiments, for each time period (e.g.. for each month), the aggregation component 335 multiplies the corresponding active devices prediction 325 with the corresponding resupplyeffectiveness prediction 330 to generate a spare parts prediction 340 for the given time period. As discussed above, the spare parts prediction 340 generally indicates the number of spare part(s) of the given type that are expected or predicted to be consumed (e.g., purchased by users) during the given time period.

[0063] As discussed above, the machine learning system (or other systems) may take a variety of actions based on the spare parts predictions 340. In some embodiments, the system may facilitate reconfiguration of one or more manufacturing operations and / or one or more distribution operations of the spare parts based on the prediction(s). For example, the spare parts prediction 340 may be used by a user (e.g., a manager of a manufacturing operation) to reconfigure aspects such as the number of spare parts they produce in a given time, the mix of spare parts (e.g., the number of humidifier tanks for each model of respiratory therapy system), and the like. Similarly, a manager of a distribution operation may use the spare parts prediction 340 to adjust when spare parts are shipped to retail environments, which region(s) the parts are shipped to in a given time, and the like.

[0064] In some aspects, facilitating reconfiguration of the manufacturing and / or distribution can include providing suggestions or recommendations (e.g., recommending that future production and / or distribution be reconfigured to align with the predictions). In some embodiments, facilitating reconfiguration may include implementing one or more reconfigurations automatically (e.g., automatically adjusting manufacturing targets or instructions, or modifying one or more automated manufacturing processes, such as to set the duty cycle of a manufacturing machine, adjust the speed or feed rate of such machines, and the like).Example Method for Training Machine Learning Models to Predict Resupply Behavior

[0065] FIG. 4 is a flow diagram depicting an example method 400 for training machine learning models to predict resupply behavior, according to some embodiments of the present disclosure. In some embodiments, the method 400 is performed by a machine learning system, such as the machine learning system 115 of FIG. 1, the machine learning system discussed above with reference to FIG. 2, and / or the machine learning system discussed above with reference to FIG. 3.

[0066] At block 405, the machine learning system accesses historical sales data (e.g., the sales data 105 of FIG. 1 and / or the historical sales data 205 of FIG. 2). In some embodiments, as discussed above, the historical sales data generally includes information relating to prior sales or use of one or more devices or systems, such as respiratory therapy systems, for whichspare parts or other consumables are available or needed. For example, the historical data may indicate the number of each type or model of respiratory therapy system that was purchased during one or more time periods (e.g., per month, per quarter, per year, and the like).

[0067] At block 410, the machine learning system accesses active device data (e.g., the activity data 110 of FIG. 1 and / or the activity data 215 of FIG. 2). In some embodiments, as discussed above, the active device data generally includes information relating to the usage or activity of the one or more devices or systems corresponding to the sales data. For example, the active device data may indicate the number of each type or model of respiratory therapy system that was active during one or more time periods (e.g., per month, per quarter, per year, and the like). In some embodiments, as discussed above, systems are classified as “active’' for a given period or window of time in response to determining that the system has been in active use (e.g., providing therapeutic airflow) for at least a minimum amount of time during the given period of time. For example, the machine learning system may determine that a total of N respiratory therapy systems were “active” in a given month based on determining that the N systems were each turned on and in the active state for at least one continuous hour during the period.

[0068] At block 415, the machine learning system trains an active device prediction model (e.g., the active device machine learning model 225 of FIG. 2) based on the historical sales data (accessed at block 405) and the active device data (accessed at block 410). As discussed above, the particular operations used to train the machine learning model may vary depending on the particular architecture and implementation. In some embodiments, to train the model, the machine learning system uses one or more features from the historical sales data as input to the model to generate a predicted active device measure (e.g., a predicted number of active devices for one or more future periods of time), and compares this prediction against the actual active device data to generate a loss. This loss may be used to refine the parameter(s) of the model in order to generate more accurate predictions.

[0069] Although not included in the illustrated example, in some embodiments, the machine learning system may further access and train the model based on forecasted sales data, as discussed above. For example, by using the forecasted sales for a given period, along with the actual sales for the given period, as input to the model, the model may learn to adjust for potential inaccuracies in the forecasts. This may enable improved prediction accurate during runtime (e.g., when forecasts alone are used as input).

[0070] At block 420. the machine learning system accesses historical spare parts sales data (e.g., the sales data 105 of FIG. 1 and / or the historical spares sales data 220 of FIG. 2). Insome embodiments, as discussed above, the historical spare parts data generally includes information relating to the sale, purchase, or use / consumption of one or more consumable components (e.g., spare parts) of the one or more devices. For example, the spare parts sales data may indicate the number of each type or model of component that was consumed or purchased by users during one or more time periods (e.g., per month, per quarter, per year, and the like). In some embodiments, as discussed above, the spare parts information is delineated based on the specific part (e.g.. where the data indicates the number of conduits sold per month, the number of humidifier tubs sold per month, and the like).

[0071] At block 425, the machine learning system trains a resupply effectiveness prediction model (e.g., the resupply machine learning model 230 of FIG. 2) based on the active device data (accessed at block 410) an the historical spare parts sales data (accessed at block 420). As discussed above, the particular operations used to train the machine learning model may vary depending on the particular architecture and implementation. In some embodiments, to train the model, the machine learning system uses the spare parts information and the active device information to generate one or more resupply effectiveness measures for one or more periods (e.g.. by dividing the number of spare parts sales in a given period by the number of active devices in the given period). By generating a sequence of such resupply effectiveness measures (e.g., over multiple periods of time), the machine learning system can use this sequence as a time series to train a time series prediction model that leams to predict future values of the resupply effectiveness measure at future periods of time.

[0072] At block 430, the machine learning system determines whether one or more training termination criteria are met. Generally, the termination criteria may include a wide variety of considerations, depending on the particular implementation. For example, the machine learning system may determine whether any additional training samples or iterations remain, whether a defined amount of time and / or computing resources have been spent training, whether the model(s) have reached a desired level of accuracy, and the like. If the criteria are not met, the method 400 returns to block 405. If the criteria are met, the method 400 continues to block 435.

[0073] At block 435, the machine learning system deploys the machine learning model(s) for inferencing. As discussed above, deploying the models may include a variety of operations, including instantiating the models locally to process data, storing the learned model parameters for future use, transmitting the learned parameters to one or more other inferencing systems, and the like.

[0074] Although the illustrated example depicts a sequential process for conceptual clarity(e.g., training the models using each individual exemplar based on stochastic gradient descent), in some aspects, the machine learning system may use batch gradient descent or other techniques to train the models. Additionally, although the illustrated example depicts a distinct training phase followed by model deployment, in some embodiments, the machine learning system may further refine, fine-tune, or update one or more model(s) after deployment. For example, the machine learning system may periodically retrain the models (e.g., each month), or may retrain the models as new data becomes available. This can allow the models to continue to leam and adapt to changing circumstances.Example Method for Predicting Resupply Behavior using Machine Learning Models

[0075] FIG. 5 is a flow diagram depicting an example method 500 for predicting resupply behavior using machine learning models, according to some embodiments of the present disclosure. In some embodiments, the method 500 is performed by a machine learning system, such as the machine learning system 115 of FIG. 1, the machine learning system discussed above with reference to FIG. 2, the machine learning system discussed above with reference to FIG. 3, and / or the machine learning system discussed above with reference to FIG. 4. In some embodiments, the method 400 is performed by one or more other systems, such as dedicated inferencing systems.

[0076] At block 505, the machine learning system accesses forecasted sales data (e.g., the sales data 120 of FIG. 1 and / or the forecasted sales data 310 of FIG. 3) for one or more devices or systems. In some embodiments, as discussed above, the forecasted sales data generally includes information relating to expected or predicted future sales or use of one or more devices or systems, such as respiratory therapy systems, for which spare parts or other consumables are available or needed. For example, the forecasted data may indicate the number of each type or model of respiratory therapy system that are expected to be purchased during one or more future time periods (e.g.. per month, per quarter, per year, and the like).

[0077] At block 510, the machine learning system generates an active devices prediction (e.g., the active devices prediction 325 of FIG. 3) based on processing the forecasted sales data using a machine learning model (e.g., the active device machine learning model 225 of FIG. 3). In some embodiments, as discussed above, the machine learning system may perform one or more preprocessing or feature extraction operations on the forecasted data prior to providing it as input to the model. For example, the machine learning system may delineate the forecasts into time periods, such as by month or quarter, and provide these period-specific forecast as separate inputs to the model. In some embodiments, as discussed above, the active devicesprediction indicates the number of devices (e.g., respiratory therapy systems) that are predicted to be in an active state (e.g., used for at least a minimum amount of time) during one or more future periods of time.

[0078] At block 515, the machine learning system generates a predicted resupply effectiveness measure (e.g., the resupply effectiveness prediction 330 of FIG. 3) for one or more consumable components (e.g., spare parts) of the system(s) using a machine learning model (e.g., the resupply machine learning model 230 of FIG. 2). For example, as discussed above, the machine learning system may use a time-series model that predicts future resupply effectiveness measures for one or more future windows of time based on the historic trends of resupply effectiveness for one or more prior windows of time. In some embodiments, as discussed above, the machine learning system generates a separate resupply effectiveness measure for each type or model of consumable component that is being evaluated (e.g., a first prediction for conduits, a second prediction for humidifier tanks, and the like).

[0079] At block 520. the machine learning system generates a spare parts prediction (e.g., the spare parts prediction 340 of FIG. 3) based on the active devices prediction (generated at block 510) and the resupply effectiveness prediction (generated at block 515). For example, in some embodiments, the machine learning system multiplies the spare parts prediction with the active devices prediction, as discussed above. In some embodiments, other operations may be used to generate the spare parts prediction. For example, the machine learning system may process the active devices prediction and the resupply effectiveness prediction using one or more machine learning models or other algorithms to predict the spare parts usage. As discussed above, the spare parts prediction generally indicates the predicted number of consumable components that are predicted to be used, consumed, and / or purchased during one or more future periods of time (e.g., the next month, the next quarter, and so on).

[0080] At block 525, the machine learning system optionally reconfigures one or more spare parts operations based on the spare parts prediction(s). In some embodiments, as discussed above, the machine learning system may facilitate reconfiguration of one or more manufacturing operations and / or one or more distribution operations of the spare parts based on the prediction(s). For example, the machine learning system may output the prediction(s) to one or more users, allowing the user(s) to reconfigure aspects such as the number of spare parts they produce in a given time, the mix of spare parts (e.g., the number of humidifier tanks for each model of respiratory therapy system), the distribution of the spare parts, and the like.

[0081] In some aspects, as discussed above, facilitating reconfiguration of the manufacturing and / or distribution can include providing suggestions or recommendations (e.g.,recommending that future production and / or distribution be reconfigured to align with the predictions). In some embodiments, facilitating reconfiguration may include implementing one or more reconfigurations automatically (e.g., automatically adjusting manufacturing targets or instructions, or modifying one or more automated manufacturing processes, such as to set the duty cycle of a manufacturing machine, adjust the speed or feed rate of such machines, and the like).Example Method for Training Machine Learning Models to Predict Resupply

[0082] FIG. 6 is a flow diagram depicting an example method 600 for training machine learning models to predict resupply, according to some embodiments of the present disclosure. In some embodiments, the method 600 is performed by a machine learning system, such as the machine learning system 115 of FIG. 1, the machine learning system discussed above with reference to FIG. 2. the machine learning system discussed above with reference to FIG. 3, the machine learning system discussed above with reference to FIG. 4, and / or the machine learning system discussed above with reference to FIG. 5.

[0083] At block 605, historical acquisitions data (e.g., the sales data 105 of FIG. 1 and / or the historical sales data 205 of FIG. 2) indicating prior acquisitions of one or more respiratory therapy systems is accessed.

[0084] At block 610, historical active device data (e.g., the activity' data 110 of FIG. 1 and / or the activity data215 of FIG. 2) associated with the one or more respiratory' therapysystems is accessed, the historical active device data corresponding to respiratory- therapy systems that were in an active state during one or more prior windows of time.

[0085] At block 615, resupply effectiveness data (e.g., determined based on the sales data 105 and activity data 110, each of FIG. 1, and / or determined based on the activity' data 215 and historical spares sales data 220, each of FIG. 2) associated with the one or more respiratory therapy systems is accessed, the resupply effectiveness data corresponding to consumption of one or more consumables for the one or more respiratory therapy systems.

[0086] At block 620, one or more parameters of a first machine learning model (e.g., the active device machine learning model 225 of FIG. 2) are updated, based on the historical acquisitions data and the historical active device data, to predict future active devices for one or more future windows of time.

[0087] At block 625, one or more parameters of a second machine learning model (e.g., the resupply machine learning model 230 of FIG. 2) are updated, based on the resupply effectiveness data, to predict future resupply effectiveness.Example Method for Predicting Resupply using Machine Learning

[0088] FIG. 7 is a flow diagram depicting an example method 700 for predicting resupply using machine learning, according to some embodiments of the present disclosure. In some embodiments, the method 700 is performed by a machine learning system, such as the machine learning system 115 of FIG. 1, the machine learning system discussed above with reference to FIG. 2, the machine learning system discussed above with reference to FIG. 3, the machine learning system discussed above with reference to FIG. 4, the machine learning system discussed above with reference to FIG. 5, and / or the machine learning system discussed above with reference to FIG. 6. In some embodiments, the method 700 is performed by one or more other systems, such as dedicated inferencing systems.

[0089] At block 705, forecasted acquisitions data (e.g., the forecasted sales data 310 of FIG. 3) indicating predicted future acquisitions of one or more respiratory therapy systems is determined.

[0090] At block 710, an active devices prediction (e.g., the active devices prediction 325 of FIG. 3) is generated based on processing at least a subset of the forecasted acquisitions data using a first trained machine learning model (e.g., the active device machine learning model 225 of FIG. 3) trained based on historical acquisitions (e.g., the historical sales data 205 of FIG. 2) of one or more respiratory therapy systems.

[0091] At block 715, a resupply effectiveness prediction (e.g., the resupply effectiveness prediction 330 of FIG. 3) is generated using a second trained machine learning model (e.g., the resupply machine learning model 230 of FIG. 3) trained based on historical consumption (e.g.. the historical spares sales data 220 of FIG. 2) of one or more consumables for the one or more respiratory therapy systems.

[0092] At block 720, a consumables prediction (e.g., the spare parts prediction 340 of FIG. 3) is generated based on the active devices prediction and the resupply effectiveness prediction.Example Processing System for Usage Prediction Machine Learning

[0093] FIG. 8 depicts an example computing device 800 configured to perform various aspects of the present disclosure, according to some embodiments disclosed herein. Although depicted as a physical device, in embodiments, the computing device 800 may be implemented using virtual device(s), and / or across a number of devices (e.g., in a cloud environment). In one embodiment, the computing device 800 corresponds to a machine learning system, such as the machine learning system of FIG. 1, the machine learning system discussed above withreference to FIG. 2, the machine learning system discussed above with reference to FIG. 3, the machine learning system discussed above with reference to FIG. 4, the machine learning system discussed above with reference to FIG. 5, the machine learning system discussed above with reference to FIG. 6, and / or the machine learning system discussed above with reference to FIG. 7

[0094] As illustrated, the computing device 800 includes a CPU 805, memory’ 810, storage 815, a network interface 825, and one or more input / output (I / O) interfaces 820. In the illustrated embodiment, the CPU 805 retrieves and executes programming instructions stored in memory 810, as well as stores and retrieves application data residing in storage 815. The CPU 805 is generally representative of a single CPU and / or GPU, multiple CPUs and / or GPUs, a single CPU and / or GPU having multiple processing cores, and the like. The memory 810 is generally included to be representative of a random access memory. Storage 815 may be any combination of disk drives, flash-based storage devices, and the like, and may include fixed and / or removable storage devices, such as fixed disk drives, removable memory' cards, caches, optical storage, network attached storage (NAS), or storage area networks (SAN).

[0095] In some embodiments, I / O devices 835 (such as keyboards, monitors, etc.) are connected via the I / O interface(s) 820. Further, via the network interface 825, the computing device 800 can be communicatively coupled with one or more other devices and components (e.g., via a network, which may include the Internet, local network(s), and the like). As illustrated, the CPU 805, memory 810. storage 815, network interface(s) 825, and I / O interface(s) 820 are communicatively coupled by one or more buses 830.

[0096] In the illustrated embodiment, the memory' 810 includes a training component 850, an inferencing component 855, an aggregation component 860, and a reconfiguration component 865, which may perform one or more embodiments discussed above. Although depicted as discrete components for conceptual clarity, in embodiments, the operations of the depicted components (and others not illustrated) may be combined or distributed across any number of components. Further, although depicted as softw are residing in memory' 810, in embodiments, the operations of the depicted components (and others not illustrated) may be implemented using hardware, software, or a combination of hardware and software.

[0097] In some embodiments, the training component 850 can be used to generate, train, and / or update machine learning models (e.g., the active device machine learning model 225 and / or the resupply machine learning model 230, each of FIG. 2), as discussed above. For example, the training component 850 may use training data (e.g., the historical sales data 205, activity7data 215, and / or historical spares sales data 220, each of FIG. 2) to update or generatetrained machine learning models, such that the models learn to predict future active device numbers and / or resupply effectiveness for consumables, as discussed above..

[0098] In some embodiments, the inferencing component 855 can be used to predict future device activity and / or resupply effectiveness using trained models, as discussed above. For example, the inferencing component 855 may use trained machine learning models (e.g., the active device machine learning model 225 and / or the resupply machine learning model 230, each of FIG. 2) to generate predicted device usage (e.g., the number of devices that will be active in the future) and / or resupply effectiveness for one or more future periods of time. In some embodiments, as discussed above, the inferencing component 855 may further aggregate or combine such predictions to generate spare part usage predictions, as discussed above.

[0099] In some embodiments, the aggregation component 860 can be used to aggregate predictions regarding device activity and / or resupply effectiveness to predict future consumable consumption, as discussed above. For example, the aggregation component 860 may, for each of one or more future windows of time, multiply the corresponding predicted active devices with the corresponding predicted resupply effectiveness, as discussed above.

[0100] In some embodiments, the reconfiguration component 865 can be used to reconfigure (or facilitate reconfiguration of) operations relating to consumable manufacturing, distribution, marketing, and the like, as discussed above. For example, the reconfiguration component 865 may provide recommendations or suggestions to users, may adjust manufacturing targets based on predicted demand, may modify manufacturing operations (e.g., increasing or reducing throughput on one or more manufacturing machines), and the like.

[0101] In the illustrated example, the storage 815 includes sales data 870 (which may correspond to the sales data 105 and / or the sales data 120, each of FIG. 1, the historical sales data 205 of FIG. 2, and / or the forecasted sales data 310 of FIG. 3). The storage 815 also includes activity data 875 (which may correspond to the activity data 110 of FIG. 1 and / or the activity' data 215 of FIG. 2) and machine learning model(s) 880 (which may correspond to the active device machine learning model 225 and / or the resupply machine learning model 230, each of FIGS. 2-3). Although depicted as residing in storage 815, the depicted data may be stored in any suitable location, including memory 810.

[0102] Generally, the depicted components (and others not depicted) in memory 810 may evaluate and / or use the depicted data (and others not depicted) in storage 815 to provide therapy data-based detection of user interface swaps or changes and / or image-based identification / classification of user interfaces, as discussed above.Example Clauses

[0103] Clause 1: A method, comprising: determining forecasted acquisitions data indicating predicted future acquisitions of one or more respiratory therapy systems; generating an active devices prediction based on processing at least a subset of the forecasted acquisitions data using a first trained machine learning model trained based on historical acquisitions of one or more respiratory therapy systems; generating a resupply effectiveness prediction using a second trained machine learning model trained based on historical consumption of one or more consumables for the one or more respiratory therapy systems; and generating a consumables prediction based on the active devices prediction and the resupply effectiveness prediction.

[0104] Clause 2: A method according to Clause 1, further comprising facilitating reconfiguration of manufacturing operations for one or more consumables for one or more respiratory therapy systems based on the consumables prediction.

[0105] Clause 3: A method according to Clause 1 or 2, wherein the consumables prediction indicates at least one of: (i) a predicted number of a humidifier tub for a respiratory therapy systems that will be consumed, or (ii) a predicted number of a conduit for a respiratory therapy systems that will be consumed.

[0106] Clause 4: A method according to any of Clauses 1-3, wherein: the historical acquisitions correspond to a first version of a respiratory therapy system, the forecasted acquisitions data corresponds to a second version of the respiratory therapy system, and the active devices prediction corresponds to the second version of the respiratory therapy systems.

[0107] Clause 5: A method according to any of Clauses 1-4, wherein: the historical acquisitions correspond to a first version of a respiratory therapy system, the forecasted acquisitions data corresponds to the first version of the respiratory therapy system, and the active devices prediction corresponds to the first version of the respiratory therapy system.Clause 6: A method according to any of Clauses 1-5, wherein generating the consumables prediction comprises multiplying the active devices prediction with the resupply effectiveness prediction.

[0108] Clause 7: A method according to any of Clauses 1-6, wherein the first and second trained machine learning models were trained to predict active devices based on active device data indicating a number of active respiratory therapy systems.

[0109] Clause 8: A method according to any of Clauses 1-7, wherein the number of active respiratory therapy systems corresponds to a number of respiratory therapy systems that were in an active state for at least a minimum duration of time during a defined window of time.

[0110] Clause 9: A system, comprising: a memory comprising computer-executableinstructions; and one or more processors configured to execute the computer-executable instructions and cause the processing system to perform a method in accordance with any one of Clauses 1-8.[OHl] Clause 10: A system, comprising means for performing a method in accordance with any one of Clauses 1-8.

[0112] Clause 11 : A non-transitory computer-readable medium comprising computerexecutable instructions that, when executed by one or more processors of a processing system, cause the processing system to perform a method in accordance with any one of Clauses 1-8.

[0113] Clause 12: A computer program product embodied on a computer-readable storage medium comprising code for performing a method in accordance with any one of Clauses 1-8.Additional Considerations

[0114] One or more elements or aspects or steps, or any portion(s) thereof, from one or more of any of claims below can be combined with one or more elements or aspects or steps, or any portion(s) thereof, from one or more of any of the other claims below or combinations thereof, to form one or more additional implementations and / or claims of the present disclosure.

[0115] While the present disclosure has been described with reference to one or more particular embodiments or implementations, those skilled in the art will recognize that many changes may be made thereto without departing from the spirit and scope of the present disclosure. Each of these implementations and obvious variations thereof is contemplated as falling within the spirit and scope of the present disclosure. It is also contemplated that additional implementations according to aspects of the present disclosure may combine any number of features from any of the implementations described herein.

[0116] The preceding description is provided to enable any person skilled in the art to practice the various embodiments described herein. The examples discussed herein are not limiting of the scope, applicability, or embodiments set forth in the claims. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments. For example, changes may be made in the function and arrangement of elements discussed without departing from the scope of the disclosure. Various examples may omit, substitute, or add various procedures or components as appropriate. For instance, the methods described may be performed in an order different from that described, and various steps may be added, omitted, or combined. Also, features described with respect to some examples may be combined in some other examples. For example, an apparatus may be implemented or a method may be practiced usingany number of the aspects set forth herein. In addition, the scope of the disclosure is intended to cover such an apparatus or method that is practiced using other structure, functionality, or structure and functionality in addition to, or other than, the various aspects of the disclosure set forth herein. It should be understood that any aspect of the disclosure disclosed herein may be embodied by one or more elements of a claim.

[0117] As used herein, the word ‘“exemplary” means “serving as an example, instance, or illustration.” Any aspect described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other aspects.

[0118] As used herein, a phrase referring to “at least one of’ a list of items refers to any combination of those items, including single members. As an example, “at least one of a, b, or c” is intended to cover a, b, c. a-b, a-c, b-c, and a-b-c. as well as any combination with multiples of the same element (e.g., a-a, a-a-a, a-a-b, a-a-c, a-b-b, a c c, b-b, b-b-b, b-b-c, c-c, and c-c-c or any other ordering of a, b, and c).

[0119] As used herein, the term “determining” encompasses a wide variety of actions. For example, “determining” may include calculating, computing, processing, deriving, investigating, looking up (e.g., looking up in a table, a database or another data structure), ascertaining and the like. Also, “determining” may include receiving (e.g., receiving information), accessing (e.g., accessing data in a memory ) and the like. Also, “determining” may include resolving, selecting, choosing, establishing and the like.

[0120] The methods disclosed herein comprise one or more steps or actions for achieving the methods. The method steps and / or actions may be interchanged with one another without departing from the scope of the claims. In other words, unless a specific order of steps or actions is specified, the order and / or use of specific steps and / or actions may be modified without departing from the scope of the claims. Further, the various operations of methods described above may be performed by any suitable means capable of performing the corresponding functions. The means may include various hardware and / or software component(s) and / or module(s), including, but not limited to a circuit, an application specific integrated circuit (ASIC), or processor. Generally, where there are operations illustrated in figures, those operations may have corresponding counterpart means-plus-function components with similar numbering.

[0121] Embodiments of the invention may be provided to end users through a cloud computing infrastructure. Cloud computing generally refers to the provision of scalable computing resources as a service over a network. More formally, cloud computing may be defined as a computing capability that provides an abstraction between the computing resourceand its underlying technical architecture (e.g., servers, storage, networks), enabling convenient, on-demand network access to a shared pool of configurable computing resources that can be rapidly provisioned and released with minimal management effort or sendee provider interaction. Thus, cloud computing allows a user to access virtual computing resources (e.g., storage, data, applications, and even complete virtualized computing systems) in “the cloud,” without regard for the underlying physical systems (or locations of those systems) used to provide the computing resources.

[0122] Typically, cloud computing resources are provided to a user on a pay-per-use basis, where users are charged only for the computing resources actually used (e.g., an amount of storage space consumed by a user or a number of virtualized systems instantiated by the user). A user can access any of the resources that reside in the cloud at any time, and from anywhere across the Internet. In context of the present invention, a user may access applications or systems (e.g., the training system and / or inferencing system) or related data available in the cloud. For example, the training system and / or inferencing system could execute on a computing system in the cloud and train and use machine learning models to predict user interface changes and / or to classify user interfaces. In such a case, the training system and / or inferencing system could receive and process the therapy data, and store the models and predictions at a storage location in the cloud. Doing so allows a user to access this information from any computing system attached to a network connected to the cloud (e.g., the Internet).

[0123] The following claims are not intended to be limited to the embodiments shown herein, but are to be accorded the full scope consistent with the language of the claims. Within a claim, reference to an element in the singular is not intended to mean “one and only one” unless specifically so stated, but rather “one or more.” Unless specifically stated otherwise, the term “some” refers to one or more. No claim element is to be construed under the provisions of 35 U.S.C. § 112(f) unless the element is expressly recited using the phrase “means for” or, in the case of a method claim, the element is recited using the phrase “step for.” All structural and functional equivalents to the elements of the various aspects described throughout this disclosure that are known or later come to be known to those of ordinary skill in the art are expressly incorporated herein by reference and are intended to be encompassed by the claims. Moreover, nothing disclosed herein is intended to be dedicated to the public regardless of whether such disclosure is explicitly recited in the claims.

Claims

CLAIMSWHAT IS CLAIMED IS:

1. A method, comprising: determining forecasted acquisitions data indicating predicted future acquisitions of one or more respiratory’ therapy systems; generating an active devices prediction based on processing at least a subset of the forecasted acquisitions data using a first trained machine learning model trained based on historical acquisitions of one or more respiratory' therapy systems; generating a resupply effectiveness prediction using a second trained machine learning model trained based on historical consumption of one or more consumables for the one or more respiratory therapy systems; and generating a consumables prediction based on the active devices prediction and the resupply effectiveness prediction.

2. The method of claim 1, further comprising facilitating reconfiguration of manufacturing operations for one or more consumables for one or more respiratory therapy systems based on the consumables prediction.

3. The method of claim 1, wherein the consumables prediction indicates at least one of:(i) a predicted number of a humidifier tub for a respiratory therapy systems that will be consumed, or(ii) a predicted number of a conduit for a respiratory therapy systems that will be consumed.

4. The method of claim 1, yvherein: the historical acquisitions correspond to a first version of a respiratory therapy system, the forecasted acquisitions data corresponds to a second version of the respiratory’ therapy system, and the active devices prediction corresponds to the second version of the respiratory therapy systems.

5. The method of claim 1, wherein: the historical acquisitions correspond to a first version of a respiratory therapy system,the forecasted acquisitions data corresponds to the first version of the respiratory therapy system, and the active devices prediction corresponds to the first version of the respiratory therapy system.

6. The method of claim 1 , wherein generating the consumables prediction comprises multiplying the active devices prediction with the resupply effectiveness prediction.

7. The method of claim 1, wherein the first and second trained machine learning models were trained to predict active devices based on active device data indicating a number of active respiratory’ therapy systems.

8. The method of claim 7, wherein the number of active respiratory therapy systems corresponds to a number of respiratory' therapy systems that were in an active state for at least a minimum duration of time during a defined window of time.

9. A method, comprising: accessing historical acquisitions data indicating prior acquisitions of one or more respiratory therapy systems; accessing historical active device data associated with the one or more respiratory therapy systems, the historical active device data corresponding to respiratory therapy systems that were in an active state during one or more prior windows of time; accessing resupply effectiveness data associated with the one or more respiratory therapy systems, the resupply effectiveness data corresponding to consumption of one or more consumables for the one or more respiratory therapy systems; updating one or more parameters of a first machine learning model, based on the historical acquisitions data and the historical active device data, to predict future active devices for one or more future windows of time; and updating one or more parameters of a second machine learning model, based on the resupply effectiveness data, to predict future resupply effectiveness.

10. The method of claim 9, wherein the resupply effectiveness data corresponds to at least one of:(i) consumption of a humidifier tub for a respiratory therapy systems, or(ii) consumption of a conduit for a respiratory therapy systems.

11. The method of claim 9, wherein the resupply effectiveness data was generated based on multiplying the historical active device data with consumables data associated with the one or more respiratory' therapy systems.

12. The method of claim 9, wherein the historical acquisitions data indicates a number of the one or more respiratory- therapy systems that yvere acquired during one or more prior yvindoyvs of time.

13. The method of claim 9, wherein the historical active device data indicates a number active respiratory therapy systems during one or more prior windoyvs of time.

14. The method of claim 13, wherein the number of active respiratory therapy systems corresponds to a number of respiratory’ therapy systems that were in an active state for at least a minimum duration of time during the one or more prior windows of time.

15. A system, comprising: one or more memories comprising computer-executable instructions; and one or more processors configured to execute the computer-executable instructions and cause the system to perform an operation comprising: determining forecasted acquisitions data indicating predicted future acquisitions of one or more respiratory therapy systems: generating an active devices prediction based on processing at least a subset of the forecasted acquisitions data using a first trained machine learning model trained based on historical acquisitions of one or more respiratory' therapy systems; generating a resupply effectiveness prediction using a second trained machine learning model trained based on historical consumption of one or more consumables for the one or more respiratory therapy systems; and generating a consumables prediction based on the active devices prediction and the resupply effectiveness prediction.

16. The system of claim 15, the operation further comprising facilitating reconfiguration of manufacturing operations for one or more consumables for one or more respiratory therapy systems based on the consumables prediction.

17. The system of claim 15, wherein the consumables prediction indicates at least one of:(i) a predicted number of a humidifier tub for a respiratory’ therapy systems that will be consumed, or(ii) a predicted number of a conduit for a respiratory therapy systems that will be consumed.

18. The system of claim 15, wherein generating the consumables prediction comprises multiplying the active devices prediction with the resupply effectiveness prediction.

19. The system of claim 15, wherein the first and second trained machine learning models were trained to predict active devices based on active device data indicating a number of active respiratory' therapy systems.

20. The system of claim 19, wherein the number of active respiratory' therapy systems corresponds to a number of respiratory therapy systems that were in an active state for at least a minimum duration of time during a defined window of time.

Images