Parameter prediction method, system, terminal and medium based on biological homeostasis

Abstract

This application discloses a method, system, terminal, and medium for parameter prediction based on biological homeostasis, relating to the field of artificial intelligence technology. The method includes: acquiring state parameter data of a first life system, including parameter names, parameter values, and parameter value validity identifiers; performing scale compression processing on the parameter values ​​according to a preset scale compression method corresponding to the first life system state parameters to obtain target parameter values; and predicting parameter prediction results using a trained prediction model based on the parameter names, target parameter values, and parameter value validity identifiers of the first life system state parameters. The predicted results include parameter values ​​corresponding to each parameter in a preset parameter set, where the parameters include multiple second life system state parameters. The prediction model is trained to predict parameters based on the biological homeostasis relationships between parameters. This allows for the prediction of parameter values ​​for other parameters based on some parameters, thereby improving the parameter prediction effect.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a parameter prediction method, system, terminal, and medium based on biological homeostasis. Background Technology

[0002] With the development of science and technology, especially the development of large language models and neural networks, the application of various parameter prediction schemes is becoming more and more widespread.

[0003] In existing technologies, parameter prediction schemes are typically applied to the processing and prediction of time-series data. For example, they predict the parameter value of a corresponding parameter at the next time step based on the parameter value at a historical time step. The problem with existing technologies is that they cannot predict the parameter values ​​of other parameters based on only some parameters, which affects the accuracy of parameter prediction.

[0004] Therefore, the relevant technologies still need to be improved and developed. Summary of the Invention

[0005] The main objective of this application is to provide a parameter prediction method, system, terminal, and medium based on biological homeostasis, aiming to solve the technical problem that parameter prediction schemes in related technologies can usually only predict the parameter value of the corresponding parameter at the next moment based on the parameter value at a historical moment, and cannot predict the parameter value of other parameters based on some parameters, thus affecting the parameter prediction effect.

[0006] To achieve the above objectives, a first aspect of this application provides a parameter prediction method based on biological homeostasis, wherein the method includes: Obtain first life system state parameter data, wherein the first life system state parameter data includes parameter names, parameter values ​​and parameter value validity identifiers corresponding to multiple preset first life system state parameters; For each of the aforementioned first life system state parameters, the parameter values ​​of the aforementioned first life system state parameters are scaled according to the preset scale compression method corresponding to the aforementioned first life system state parameters, so as to map the aforementioned parameter values ​​to a preset target parameter value range and obtain the target parameter values; Based on the parameter names, target parameter values, and parameter value validity identifiers corresponding to the aforementioned first life system state parameters, the parameter prediction results are obtained through the trained prediction model. The aforementioned parameter prediction results include the parameter values ​​corresponding to each parameter in the preset parameter set. The parameters in the preset parameter set include multiple preset second life system state parameters. The aforementioned prediction model is trained to predict parameters based on the biological homeostatic relationship between the parameters.

[0007] Optionally, based on the parameter names, target parameter values, and parameter value validity identifiers corresponding to the aforementioned first life system state parameters, the above-mentioned parameter prediction results are obtained through a trained prediction model, including: Input the parameter names, target parameter values, and parameter value validity identifiers corresponding to the above-mentioned first life system state parameters into the above-mentioned trained prediction model; In the above prediction model, the parameter names, target parameter values ​​and parameter value validity identifiers corresponding to the above first life system state parameters are encoded to obtain the embedded numerical representations corresponding to each of the above first life system state parameters. The above-mentioned embedded numerical representations are semantically mapped to obtain the parameter representation vector sequence corresponding to the above-mentioned first life system state parameter data; Obtain the training validity identifier corresponding to each of the above-mentioned first life system state parameters, and determine the parameter validity identifier corresponding to each of the above-mentioned first life system state parameters based on the training validity identifier and the parameter value validity identifier. Based on the above parameter validity identifier, the above target parameter value, and the above parameter representation vector sequence, obtain the steady-state variable vector corresponding to the above first life system state parameter data; Based on the above steady-state variable vector, the predicted parameters are obtained.

[0008] Optionally, the above-mentioned encoding process of the parameter name, target parameter value, and parameter value validity identifier corresponding to the first life system state parameter to obtain the embedded numerical representation corresponding to each of the first life system state parameters includes: For each of the aforementioned first life system state parameters, a semantic embedded value is generated based on the parameter name, a parameter value embedded value is generated based on the target parameter value, and a value validity embedded value is generated based on the parameter value validity identifier. The semantic embedded value, the parameter value embedded value, and the value validity embedded value are then fused to obtain the embedded value representation corresponding to the aforementioned first life system state parameter.

[0009] Optionally, the above prediction model includes a parameter semantic mapping module, which includes at least one pre-specified target hidden layer; The above semantic mapping processing of the embedded numerical representation to obtain the parameter representation vector sequence corresponding to the first life system state parameter data includes: The above embedded numerical representation is input into the above parameter semantic mapping module for semantic mapping processing; The outputs of each of the aforementioned target hidden layers are obtained and used as the parameter representation vector sequence.

[0010] Optionally, determining the parameter validity identifier corresponding to each of the first life system state parameters based on the training validity identifier and the parameter value validity identifier includes: For each of the above-mentioned first life system state parameters, if both the above-mentioned training validity identifier and the above-mentioned parameter value validity identifier are valid, then the parameter validity identifier corresponding to the above-mentioned first life system state parameter is determined to be valid; otherwise, the parameter validity identifier corresponding to the above-mentioned first life system state parameter is determined to be invalid. The above-mentioned parameter value validity identifier is used to indicate whether the parameter value of the above-mentioned first life system state parameter is valid; the above-mentioned training validity identifier is used to indicate whether the training process of the above-mentioned prediction model is affected by the above-mentioned first life system state parameter; the above-mentioned parameter validity identifier is used to indicate whether the above-mentioned first life system state parameter is used when making predictions through the above-mentioned prediction model.

[0011] Optionally, obtaining the steady-state variable vector corresponding to the first life system state parameter data based on the above-mentioned parameter validity identifier, the above-mentioned target parameter value, and the above-mentioned parameter representation vector sequence includes: For each of the aforementioned first life system state parameters, the parameter weights corresponding to the aforementioned first life system state parameters are determined based on the aforementioned parameter validity identifiers and the aforementioned target parameter values. The above parameter representation vector sequence is weighted and fused according to the above parameter weights to obtain the steady-state variable vector corresponding to the above first life system state parameter data.

[0012] Optionally, the parameters in the above-mentioned preset parameter set may also include the above-mentioned first life system state parameters; The above parameter prediction results include the parameter name, parameter value, and parameter value source indicator for each parameter in the preset parameter set.

[0013] A second aspect of this application provides a parameter prediction system based on biological homeostasis, wherein the system comprises: The data acquisition module is used to acquire the first life system state parameter data, wherein the first life system state parameter data includes the parameter name, parameter value and parameter value validity identifier corresponding to multiple preset first life system state parameters; The data processing module is used to perform scale compression processing on the parameter values ​​of the first life system state parameters according to the preset scale compression method corresponding to the first life system state parameters, so as to map the parameter values ​​to a preset target parameter value range and obtain the target parameter values. The prediction module is used to predict parameter prediction results based on the parameter names, target parameter values, and parameter value validity identifiers corresponding to the first life system state parameters, using a trained prediction model. The parameter prediction results include the parameter values ​​corresponding to each parameter in a preset parameter set. The parameters in the preset parameter set include multiple preset second life system state parameters. The prediction model is trained to predict parameters based on the biological homeostatic relationship between parameters.

[0014] A third aspect of this application provides a terminal, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of any of the above-mentioned parameter prediction methods based on biological homeostasis.

[0015] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the above-described biological homeostasis-based parameter prediction methods.

[0016] As can be seen from the above, the present application provides a parameter prediction method, system, terminal, and medium based on biological homeostasis. Specifically, the method includes: acquiring state parameter data of a first life system, wherein the state parameter data of the first life system includes parameter names, parameter values, and parameter value validity identifiers corresponding to multiple preset state parameters of the first life system; for each of the first life system state parameters, performing scale compression processing on the parameter values ​​of the first life system state parameters according to a preset scale compression method corresponding to the first life system state parameters, so as to map the parameter values ​​to a preset target parameter value range to obtain target parameter values; and predicting parameter prediction results using a trained prediction model based on the parameter names, target parameter values, and parameter value validity identifiers corresponding to the first life system state parameters, wherein the parameter prediction results include parameter values ​​corresponding to each parameter in a preset parameter set, the parameters in the preset parameter set include multiple preset state parameters of a second life system, and the prediction model is trained to predict parameters based on the biological homeostasis relationship between parameters.

[0017] Thus, when predicting the state parameters of a living system, the first set of state parameter data for the first living system is obtained. The parameter values ​​in this data are then scaled and compressed to map to a preset target parameter value range, yielding the target parameter values. Next, based on the parameter names, target parameter values, and parameter validity identifiers corresponding to the first living system state parameters, a trained prediction model capable of predicting parameters based on biological homeostatic relationships between parameters is used to make predictions and obtain the prediction results. These prediction results include the parameter values ​​corresponding to the second living system state parameters. Therefore, this application's solution, considering biological homeostatic relationships, enables the prediction of data corresponding to the second living system state parameters based on data corresponding to the first living system state parameters. This means it can predict the parameter values ​​of other parameters based on some parameters, thereby improving the parameter prediction effect. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating a parameter prediction method based on biological homeostasis provided in an embodiment of this application; Figure 2 This is a schematic diagram of the constituent modules of a parameter prediction system based on biological homeostasis provided in an embodiment of this application; Figure 3 This is a block diagram illustrating the internal structure of a terminal provided in an embodiment of this application. Detailed Implementation

[0020] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of this application with unnecessary detail.

[0021] It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0022] It should also be understood that the terminology used in this application specification is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this application specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0023] It should also be further understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0024] As used in this specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to classification." Similarly, the phrases "if determined" or "if classified to [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once classified to [the described condition or event]," or "in response to classification to [the described condition or event]."

[0025] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0026] Many specific details are set forth in the following description in order to provide a full understanding of this application. However, this application may also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the spirit of this application. Therefore, this application is not limited to the specific embodiments disclosed below.

[0027] Currently, various parameter prediction schemes are being used more and more widely. However, parameter prediction schemes are usually applied to the processing and prediction of time series data. For example, they predict the parameter value of the corresponding parameter at the next time based on the parameter value at a historical time. They cannot predict the parameter value of other parameters based on some parameters, which affects the parameter prediction effect.

[0028] Biological homeostasis reflects an organism's ability to maintain functional stability under multi-scale, multi-system coordinated regulation, and there are homeostatic correlations among the system state parameters of living organisms. However, current technologies lack parameter prediction schemes based on biological homeostatic relationships.

[0029] To address at least one of the aforementioned technical problems, this application proposes a parameter prediction method, system, terminal, and medium based on biological homeostasis. Specifically, the method includes: acquiring state parameter data of a first life system, wherein the first life system state parameter data includes parameter names, parameter values, and parameter value validity identifiers corresponding to multiple preset first life system state parameters; for each of the first life system state parameters, performing scale compression processing on the parameter values ​​of the first life system state parameters according to a preset scale compression method corresponding to the first life system state parameters, so as to map the parameter values ​​to a preset target parameter value range to obtain target parameter values; and predicting parameter prediction results using a trained prediction model based on the parameter names, target parameter values, and parameter value validity identifiers corresponding to the first life system state parameters, wherein the parameter prediction results include parameter values ​​corresponding to each parameter in a preset parameter set, the parameters in the preset parameter set include multiple preset second life system state parameters, and the prediction model is trained to predict parameters based on the biological homeostasis relationship between parameters.

[0030] Thus, when predicting the state parameters of a living system, the first set of state parameter data for the first living system is obtained. The parameter values ​​in this data are then scaled and compressed to map to a preset target parameter value range, yielding the target parameter values. Next, based on the parameter names, target parameter values, and parameter validity identifiers corresponding to the first living system state parameters, a trained prediction model capable of predicting parameters based on biological homeostatic relationships between parameters is used to make predictions and obtain the prediction results. These prediction results include the parameter values ​​corresponding to the second living system state parameters. Therefore, this application's solution, considering biological homeostatic relationships, enables the prediction of data corresponding to the second living system state parameters based on data corresponding to the first living system state parameters. This means it can predict the parameter values ​​of other parameters based on some parameters, thereby improving the parameter prediction effect.

[0031] like Figure 1 As shown in the embodiments of this application, a parameter prediction method based on biological homeostasis is provided. Specifically, the method includes the following steps: Step S100: Obtain first life system state parameter data, wherein the first life system state parameter data includes parameter names, parameter values ​​and parameter value validity identifiers corresponding to multiple preset first life system state parameters; Step S200: For each of the above-mentioned first life system state parameters, according to the preset scale compression method corresponding to the above-mentioned first life system state parameters, the parameter values ​​of the above-mentioned first life system state parameters are scale compressed to map the above-mentioned parameter values ​​to a preset target parameter value range to obtain target parameter values; Step S300: Based on the parameter name, target parameter value, and parameter value validity identifier corresponding to the first life system state parameter, the parameter prediction result is obtained through the trained prediction model. The parameter prediction result includes the parameter value corresponding to each parameter in the preset parameter set. The parameters in the preset parameter set include multiple preset second life system state parameters. The prediction model is trained to predict parameters based on the biological homeostatic relationship between parameters.

[0032] The aforementioned first life system state parameters are pre-specified life system state parameters that need to be used in the prediction process, i.e., life system state parameters that need to be collected. Life system state parameters refer to indicators used to assess the physiological functions and health status of cells, animals, and other organisms. For example, they may include one or more of the following indicators: Cell-level indicators may include gene damage degree and morphological stability; animal (or organism)-level indicators may include the proportion of stereotyped behaviors, anxiety level, pulse, heart rate, respiratory rate, blood pressure, and blood oxygen saturation. In practical applications, other indicators may also be included, which are not specifically limited here.

[0033] It should be noted that during the collection of primary life system status parameter data, some primary life system status parameters may not have their corresponding parameter values ​​collected, or the collected parameter values ​​may be abnormal (e.g., exceeding a reasonable preset parameter value range). Therefore, a parameter value validity indicator is used to indicate the validity of the primary life system status parameter values. For parameters whose collected values ​​conform to the preset parameter value range, the parameter value validity indicator is valid; otherwise, it is invalid. The aforementioned preset parameter value range can be set separately for each parameter according to actual needs, and is not specifically limited here.

[0034] Furthermore, the first life system state parameters whose parameter values ​​are identified as valid can be used in subsequent processing. To reduce computational complexity, their parameter values ​​are preprocessed. Specifically, based on the preset scale compression method corresponding to the aforementioned first life system state parameters, the parameter values ​​of the aforementioned first life system state parameters are scale-compressed to map the parameter values ​​to a preset target parameter value range, thereby obtaining the target parameter values.

[0035] It should be noted that the value ranges of different first-life system state parameters may vary significantly. Therefore, different scaling compression methods can be set for different first-life system state parameters to achieve better compression. The aforementioned target parameter value range can be set and adjusted according to actual needs. It should be further noted that different first-life system state parameters can correspond to different target parameter value ranges, or they can correspond to the same target parameter value range. In this embodiment, different first-life system state parameters correspond to the same target parameter value range, and the aforementioned target parameter value range is from -1 to 1, but this is not a specific limitation. Through scaling compression, the dimensional differences between different types of parameters can be eliminated, bringing the parameter values ​​into a uniform numerical range, which facilitates subsequent model processing and feature extraction.

[0036] In this embodiment, after parameter value compression, processing is performed based on a trained prediction model to achieve parameter prediction. The prediction model can be a pre-trained large language model.

[0037] Specifically, based on the parameter names, target parameter values, and parameter value validity identifiers corresponding to the aforementioned first life system state parameters, the trained prediction model predicts the parameter prediction results, including: Input the parameter names, target parameter values, and parameter value validity identifiers corresponding to the above-mentioned first life system state parameters into the above-mentioned trained prediction model; In the above prediction model, the parameter names, target parameter values ​​and parameter value validity identifiers corresponding to the above first life system state parameters are encoded to obtain the embedded numerical representations corresponding to each of the above first life system state parameters. The above-mentioned embedded numerical representations are semantically mapped to obtain the parameter representation vector sequence corresponding to the above-mentioned first life system state parameter data; Obtain the training validity identifier corresponding to each of the above-mentioned first life system state parameters, and determine the parameter validity identifier corresponding to each of the above-mentioned first life system state parameters based on the training validity identifier and the parameter value validity identifier. Based on the above parameter validity identifier, the above target parameter value, and the above parameter representation vector sequence, obtain the steady-state variable vector corresponding to the above first life system state parameter data; Based on the above steady-state variable vector, the predicted parameters are obtained.

[0038] In some application scenarios, the aforementioned prediction model may include a living system state parameter input representation module, a parameter semantic mapping module, a steady-state variable construction module, a steady-state perception aggregation module, and a steady-state variable system prediction module, each performing different functions. These five functional modules are used to process and model the living system state parameters step by step.

[0039] Specifically, the life system state parameter input representation module is used to uniformly encode and structurally represent multi-source, multi-modal life system state parameters collected from cells, animals, and / or humans, forming standardized inputs for subsequent processing. The parameter semantic mapping module is used to semantically model the life system state parameters, mapping the original parameters to a high-dimensional latent representation space with semantic consistency; the functionality of this module can be implemented based on an artificial intelligence large language model or semantic mapper. The steady-state variable construction module is used to extract and construct steady-state variable vectors representing the steady-state characteristics of the life system from the semantically mapped latent representation, thus characterizing the intrinsic steady-state structure of the life system. The steady-state perception aggregation module is used to perform steady-state perception and aggregation on the steady-state variable vectors, forming a unified representation of life steady-state. The steady-state variable system prediction module is used to predict and output the state parameters of a second life system based on the aggregated steady-state variable vectors.

[0040] In some application scenarios, the aforementioned prediction model can also include a steady-state optimization module to control the model training process and optimize the overall prediction model built upon the five functional modules, thereby achieving accurate identification, quantitative characterization, and state prediction of life homeostasis. Specifically, the steady-state optimization module optimizes the prediction model, learning to adjust homeostatic variables and model parameters. It should be noted that the aforementioned first life system state parameters and second life system state parameters respectively include different life system state parameters, and there are biological homeostatic correlations between the corresponding life system state parameters. The prediction model also uses life system state parameters with biological homeostatic correlations for training during the training process to learn these correlations.

[0041] It should be further noted that the modules and their corresponding functions described above are for illustrative purposes only and are not intended to limit the scope of the application. The specific steps executed by each module can be found in the detailed processing procedures described below, and are not specifically limited here.

[0042] Specifically, the above-mentioned encoding process is performed on the parameter names, target parameter values, and parameter value validity identifiers corresponding to the aforementioned first life system state parameters to obtain the embedded numerical representations corresponding to each of the aforementioned first life system state parameters, including: For each of the aforementioned first life system state parameters, a semantic embedded value is generated based on the parameter name, a parameter value embedded value is generated based on the target parameter value, and a value validity embedded value is generated based on the parameter value validity identifier. The semantic embedded value, the parameter value embedded value, and the value validity embedded value are then fused to obtain the embedded value representation corresponding to the aforementioned first life system state parameter.

[0043] In this embodiment of the application, the aforementioned first life system state parameter data is derived from any one or more life systems in cells, animals, or the human body. The aforementioned first life system state parameter data includes parameter names, parameter values, and parameter value validity identifiers corresponding to a plurality of preset first life system state parameters.

[0044] For example, at the cellular level, the first life system state parameter can be a gene-related parameter, with parameter names such as TP53 and MYC, and parameter values ​​representing the expression level of the corresponding gene. The validity indicator of the parameter value is used to indicate whether the expression level of the corresponding gene has been effectively measured (1 for valid and 0 for invalid). At the human body level, the first life system state parameter can be a physiological laboratory parameter, with parameter names such as blood glucose, blood pressure, and blood lipids, and parameter values ​​representing the measured values ​​of the corresponding laboratory indicators. The validity indicator of the parameter value is used to indicate whether the corresponding measured value is valid (e.g., whether it is within a reasonable measurement range, whether there is a measurement error, etc.).

[0045] In some application scenarios, the aforementioned first life system state parameter data are first represented as semantic representation unit vectors, numerical representation unit vectors, and missing representation unit vectors.

[0046] For example, at the cellular level, gene names (such as TP53, MYC) are represented as semantic representation unit vectors, and the corresponding gene expression levels are represented as numerical representation unit vectors; at the animal level, physiological indicator types (such as heart rate, activity intensity, feeding status) are represented as semantic representation unit vectors, and the corresponding measurement values ​​are represented as numerical representation unit vectors; at the human level, questionnaire items, laboratory indicator names, or subjective symptom descriptions are represented as semantic representation unit vectors, and blood test values, physiological measurement values, or scale scores are represented as numerical representation unit vectors; for uncollected or missing life system state parameters, they are uniformly represented as missing representation unit vectors.

[0047] Vector mapping techniques are used to map semantic representation unit vectors to obtain corresponding semantic embedding values. For example, a learnable neural network embedding layer can be used to map the semantic representation unit vectors corresponding to gene names, physiological indicator names, or questionnaire items into continuous vector representations of fixed dimensions to characterize the potential relationships between different semantic units.

[0048] Numerical mapping techniques are used to map numerical representation unit vectors to obtain corresponding parameter values ​​embedded in numerical values. For example, nonlinear mapping of gene expression levels, physiological measurements, or laboratory indicator values ​​can be performed through learnable multi-layer numerical neural networks to eliminate dimensional differences and enhance the expressive power of numerical features.

[0049] Vector mapping techniques are used to map missing representation unit vectors to obtain corresponding valid embedding values. For example, a uniform vector representation can be assigned to missing states through a learnable neural network embedding layer, enabling the model to explicitly perceive the missing state parameters of the living system, rather than simply ignoring the missing information.

[0050] Furthermore, the semantic embedding values, parameter value embedding values, and validity embedding values ​​mentioned above are fused to construct a unified embedding value representation, thereby obtaining the input representation of the life system state parameters. For example, by setting learnable weights, the semantic embedding values, parameter value embedding values, and validity embedding values ​​can be weighted and fused to generate a unified parameter representation vector for subsequent steady-state modeling.

[0051] Furthermore, the above prediction model includes a parameter semantic mapping module, which includes at least one pre-specified target hidden layer; The above semantic mapping processing of the embedded numerical representation to obtain the parameter representation vector sequence corresponding to the first life system state parameter data includes: The above embedded numerical representation is input into the above parameter semantic mapping module for semantic mapping processing; The outputs of each of the aforementioned target hidden layers are obtained and used as the parameter representation vector sequence.

[0052] Specifically, the input representation of the life system state parameters (i.e., the embedded numerical representation mentioned above) is input into the semantic mapper of the parameter semantic mapping module for processing. Among them, the row number semantic mapper is used to perform semantic modeling of the life system state parameters; for example, the semantic mapper can use a pre-trained open-source large language model as the base model.

[0053] During pre-training, selective training control is applied to the network structures in the semantic mapper, so that some network structures keep their parameters fixed while others participate in model training. For example, the attention module in the semantic mapper can be frozen, and only some of its hidden layers (i.e., the target hidden layer) can be unfrozen to participate in model training, so as to achieve adaptation to the state parameters of the living system while maintaining the original semantic representation ability.

[0054] Different numerical precision settings are adopted for different network structures that participate in training and those that do not, in order to balance model stability and computational efficiency. For example, attention modules that do not participate in training can be set to a lower numerical precision representation, while hidden layer modules that participate in training can be set to a higher numerical precision representation.

[0055] After the model training is completed, the output of the target hidden layer participating in the training is extracted from the semantic mapper and used as the semantic mapping result of the state parameters of the living system for subsequent steady-state variable construction and analysis.

[0056] It should be noted that the number of target hidden layers and the specific target hidden layers selected can be set and adjusted according to actual needs, and no specific restrictions are imposed here.

[0057] It should be further explained that the output of each target hidden layer is treated as a sequence of parameter representation vectors. For example, if three target hidden layers are specified in advance, three sequences of parameter representation vectors are obtained. Each sequence of parameter representation vectors includes multiple parameter representation vectors that correspond one-to-one with the state parameters of the first life system.

[0058] Furthermore, the determination of the parameter validity identifier corresponding to each of the aforementioned first life system state parameters based on the aforementioned training validity identifier and the aforementioned parameter value validity identifier includes: For each of the above-mentioned first life system state parameters, if both the above-mentioned training validity identifier and the above-mentioned parameter value validity identifier are valid, then the parameter validity identifier corresponding to the above-mentioned first life system state parameter is determined to be valid; otherwise, the parameter validity identifier corresponding to the above-mentioned first life system state parameter is determined to be invalid. The above-mentioned parameter value validity identifier is used to indicate whether the parameter value of the above-mentioned first life system state parameter is valid; the above-mentioned training validity identifier is used to indicate whether the training process of the above-mentioned prediction model is affected by the above-mentioned first life system state parameter; the above-mentioned parameter validity identifier is used to indicate whether the above-mentioned first life system state parameter is used when making predictions through the above-mentioned prediction model.

[0059] Specifically, the validity indicator of the above parameter values ​​is used to reflect whether the corresponding first life system state parameters are missing during the original data acquisition or data processing process.

[0060] For example, when the expression level of a certain gene is not measured, a certain physiological indicator of an animal is not recorded, or a certain laboratory indicator of a human is not collected, the validity flag of the corresponding parameter value is set to invalid (e.g., a value of 0); when a valid observation value exists for the parameter, it is set to valid (e.g., a value of 1). The validity flag of the parameter value can be automatically generated by missing value flags, null value flags, threshold judgments, or consistency verification rules from the data source.

[0061] Training validity identifiers are used to artificially mask or selectively control the visibility of parameters during model training to achieve robust modeling, generalization enhancement, or specific task constraints. For example, during training, some observed parameters can be randomly set to invisible to simulate some observation conditions in a real-world scenario; parameters can be masked by system, modality, or hierarchy to train the model to construct steady-state variables and predict the states of other systems even when only a certain system or a small number of system parameters are given; or specific types of parameters can be masked according to a preset strategy to improve the model's robustness to noisy or redundant parameters. Training validity identifiers can be generated using random, hierarchical, system-based, or task-based strategies.

[0062] The parameter validity identifier can be obtained by combining the above-mentioned training validity identifier and the above-mentioned parameter value validity identifier. The combination rule is used to simultaneously reflect "whether the data exists" and "whether it is allowed to be used in training". For example, the final validity mask (i.e., parameter validity identifier) ​​can be obtained by product fusion: when the parameter exists in the data and is allowed to be seen during the training phase, the final parameter validity identifier is valid; when the parameter itself is missing or is manually masked during training, the final parameter validity identifier is invalid.

[0063] In this embodiment of the application, obtaining the steady-state variable vector corresponding to the first life system state parameter data based on the parameter validity identifier, the target parameter value, and the parameter representation vector sequence includes: For each of the aforementioned first life system state parameters, the parameter weights corresponding to the aforementioned first life system state parameters are determined based on the aforementioned parameter validity identifiers and the aforementioned target parameter values. The above parameter representation vector sequence is weighted and fused according to the above parameter weights to obtain the steady-state variable vector corresponding to the above first life system state parameter data.

[0064] Specifically, for parameter validity indicators, 0 can represent invalidity and 1 can represent validity. Then, for each of the above parameter representation vector sequences, the absolute value of the product of the parameter validity indicator and the parameter representation vector corresponding to that parameter in the sequence is taken as the parameter weight. Furthermore, the parameter weights can be normalized to ensure that the sum of the parameter weights of all valid parameters corresponding to a parameter representation vector sequence is a preset value (e.g., 1), thus forming a weight distribution that reflects the relative importance of each parameter.

[0065] Based on the obtained parameter weights, the parameter representation vector sequence is weighted and fused to obtain one or more steady-state variable vectors, which are used to characterize the intrinsic state structure of the living system in different steady-state dimensions. For example, the parameter representation vectors can be weighted and summed / or weighted combined according to the weight distribution to obtain the corresponding steady-state variable vectors. Each steady-state variable vector is used to characterize the comprehensive state features of the living system in a certain steady-state dimension.

[0066] It should be noted that one or more weighted fusion methods can be pre-set to obtain one or more corresponding steady-state variable vectors. One weighted fusion method corresponds to one steady-state variable vector. For example, if the obtained parameter representation vector sequence consists of three elements, one pre-set weighted fusion method is to weightedly fuse all parameter representation vector sequences to obtain one steady-state variable vector; another pre-set weighted fusion method is to weightedly fuse the latter two (corresponding to the last two target hidden layers) parameter representation vector sequences to obtain one steady-state variable vector. The weighted fusion method can be set and adjusted according to actual needs, and no specific limitations are made here.

[0067] Furthermore, based on the aforementioned steady-state variable vector, the predicted parameters are obtained. Specifically, the steady-state variable vector is input into the prediction layer of the prediction model, and the steady-state variable vector is normalized to eliminate scale differences between different steady-state dimensions. Then, through linear mapping and nonlinear transformation, high-order steady-state features for system-level prediction are extracted. Finally, based on the aforementioned high-order steady-state features, a system-level prediction mapping is constructed to map the steady-state representation to the target prediction space, generating parameter values ​​corresponding to each parameter in the preset parameter set.

[0068] In this embodiment of the application, the parameters in the aforementioned preset parameter set also include the aforementioned first life system state parameters; the aforementioned parameter prediction results include the parameter name, parameter value, and parameter value source indication flag corresponding to each parameter in the preset parameter set. The source indication flag for the measured first life system state parameters is "measured," and the source indication flag for the predicted second life system state parameters and the predicted values ​​of the first life system state parameters is "predicted," making it easier for users to distinguish between measured data and predicted data.

[0069] In some application scenarios, multiple homeostatic variable vectors are used as input to the homeostatic perception aggregation module to describe the state characteristics of a living system in different homeostatic dimensions. For example, for a sample of a living system, it can receive multiple homeostatic variable vectors constructed by the previous module, where each homeostatic variable vector represents a different homeostatic dimension such as metabolic homeostasis, immune homeostasis, or behavioral homeostasis.

[0070] The steady-state perception aggregation module constructs a steady-state coupling matrix based on the interaction relationships between steady-state variable vectors. This matrix characterizes the strength and direction of the associations between different steady-state variables. For example, a steady-state coupling matrix describing the coupling relationships between steady-state variables can be formed by calculating the correlation, similarity, or mapping relationships between different steady-state variable vectors.

[0071] In this model, each value of the steady-state coupling matrix represents the correlation between elements at corresponding positions in the two steady-state variable vectors. For example, for the first and second steady-state variable vectors, the element in the first row and second column of the steady-state coupling matrix represents the correlation between the first element in the first steady-state variable vector and the second element in the second steady-state variable vector. The steady-state coupling matrix is ​​calculated based on the correlation between the steady-state variable vectors.

[0072] It should be noted that the steady-state coupling matrix is ​​not used as label data during model training. During training, stability constraints are used to ensure that the relationship between steady-state variable vectors remains in a steady state during optimization.

[0073] Imposing linearization constraints on the steady-state coupling matrix allows the relationships between steady-state variables to be approximated as linear within a local range, thus characterizing the system behavior near the steady state. For example, the nonlinear relationships between steady-state variables can be linearly approximated near the current steady state, resulting in a linearized steady-state coupling matrix that reflects the steady-state correlation characteristics.

[0074] A maximum spectral radius constraint is imposed on the linearized steady-state coupling matrix to ensure that its spectral radius does not exceed a preset threshold (which can be pre-set and adjusted according to actual needs). This ensures the stability of the coupling relationship between steady-state variables. For example, by constraining the amplitude of the maximum eigenvalue of the steady-state coupling matrix to be less than a preset value (which can be pre-set and adjusted according to actual needs), it is ensured that the interaction between steady-state variables does not lead to system state divergence, thus guaranteeing that what is being characterized is a steady-state correlation, rather than an unsteady or unstable correlation.

[0075] Based on a steady-state coupling matrix that satisfies stability constraints, steady-state variable vectors are aggregated using steady-state perception to obtain an aggregated representation that reflects the overall steady-state structure of the living system. For example, the steady-state coupling matrix can be used to perform weighted combinations or structured aggregations of each steady-state variable vector to generate a comprehensive steady-state representation for subsequent steady-state prediction or steady-state analysis.

[0076] The steady-state variable system prediction module receives a comprehensive steady-state representation output from the steady-state perception aggregation module as input for systematic prediction, characterizing the overall steady-state structure of the living system at the current moment or under given conditions. For example, it can receive an aggregated vector representing the overall steady-state state of a biological sample, used to predict the state distribution of that sample across multiple physiological systems.

[0077] Normalization is applied to the overall steady-state representation to eliminate scale differences between different steady-state dimensions and improve the stability of the prediction process. For example, normalization operations can be used to standardize the aggregated steady-state representation, ensuring that different steady-state variables have similar numerical ranges during the prediction process.

[0078] Linear mapping and nonlinear transformation are applied to the normalized comprehensive steady-state representation to extract high-order steady-state features for system-level prediction. For example, a set of learnable linear transformations combined with nonlinear activation functions can be used to transform the comprehensive steady-state representation to enhance the nonlinear expressive power between steady-state variables.

[0079] Based on the aforementioned higher-order steady-state characteristics, a system-level prediction mapping is constructed to map the steady-state representation to the target prediction space, generating prediction results for multiple life system state parameters. For example, the steady-state representation can be mapped to a prediction vector for multiple physiological indicators, behavioral indicators, or cell state variables, where each dimension corresponds to a life system state parameter to be predicted.

[0080] It outputs predictions of steady-state variables to describe the parameter distribution of a living system under current steady-state conditions. For example, it can output predicted values ​​for multiple system parameters of a human sample, such as those in the cardiovascular, metabolic, and immune systems, or predict the activity levels of different functional genes in cells.

[0081] In some application scenarios, the training process for the above prediction model includes: Collecting training data: Collecting state parameter data of life systems at different life levels (e.g., cells, animals, or humans) and different species as training data. The training data includes multiple samples, each containing parameter names, parameter values, parameter value validity identifiers, and training validity identifiers corresponding to multiple life system state parameters. It also includes the true values ​​of each parameter in the preset parameter set corresponding to each sample, used for model training and validation. Data preprocessing: Preprocessing the training data, including data cleaning, outlier removal, missing value labeling, and scaling the parameter values ​​according to the method in step S200 above, to obtain preprocessed training data. Model initialization: Initializing each module of the prediction model. The parameter semantic mapping module of this model uses a pre-trained large language model as the base model to initialize the model parameters. Model Training: The preprocessed training data is input into the initialized prediction model. The model processes the data and performs forward propagation to obtain prediction results. The loss value (e.g., mean squared error loss) between the prediction results and the true values ​​is calculated. Based on the loss value, the backpropagation algorithm is used to update the model parameters until the model converges (the loss value is less than a preset threshold or a preset number of training epochs are reached). Model Validation and Optimization: The converged model is validated using validation data. The model's hyperparameters (e.g., learning rate, number of hidden layers, preset target parameter value range, etc.) are adjusted to optimize model performance, ultimately obtaining the trained prediction model.

[0082] Specifically, differentiated learning parameters are set for different functional modules of the prediction model. The semantic mapper uses a relatively small learning rate, while the other functional modules use relatively large learning rates. This aims to accelerate the convergence speed of steady-state related modules while maintaining the stability of the pre-trained semantic representations. For example, during model training, the semantic mapper in the parameter semantic mapping module can be set to a smaller learning rate to avoid disrupting its existing semantic structure; simultaneously, the steady-state variable construction module, the steady-state perception aggregation module, and the steady-state variable system prediction module are set to larger learning rates to facilitate the rapid learning of steady-state variables and system-level relationships.

[0083] The model optimization process is conducted separately based on life system data from different species and individuals, enabling the steady-state model to simultaneously learn both species-wide common steady-state structures and individual-specific steady-state characteristics. For example, during model training, the data sources for different species can be distinguished, allowing the model to share some parameters while introducing adaptive parameters for different species or individuals, thereby avoiding the mixed modeling of steady-state characteristics from different life systems.

[0084] During model optimization, parameters corresponding to missing values ​​are not included in model parameter updates to prevent missing information from interfering with steady-state model optimization. For example, when the expression level of a certain cell gene is not measured or a certain human physiological indicator is missing, a missing value mask (i.e., a parameter value validity identifier) ​​can be used to prevent the corresponding prediction error from participating in loss calculation, thereby ensuring that the model is optimized only based on valid observation data.

[0085] During model optimization, the original values ​​of the life system's state parameters are scaled and used as training input to maintain consistency between the steady-state modeling results and the actual measurements of the life system. For example, the original gene expression levels, physiological measurements, or laboratory index values ​​can be directly used for model training, rather than relying solely on normalized intermediate representations. This improves the interpretability of steady-state prediction results in practical applications and facilitates the use of mean squared error to optimize the prediction results.

[0086] Introducing system masking information (i.e., training validity indicators) during model training allows for structured constraints on different life systems or functional modules, enabling the model to learn system-level homeostatic relationships. For example, at the cellular level, gene parameters can be partitioned using system masks based on gene functional blocks or signaling pathways; at the animal and human levels, physiological parameters can be grouped using system masks according to the definitions of physiological systems such as the cardiovascular, metabolic, and immune systems, thereby guiding the model to perform homeostatic optimization at the system level.

[0087] This application provides a parameter prediction method based on biological homeostasis. Specifically, the method includes: acquiring state parameter data of a first life system, wherein the first life system state parameter data includes parameter names, parameter values, and parameter value validity identifiers corresponding to multiple preset first life system state parameters; for each of the first life system state parameters, performing scale compression processing on the parameter values ​​of the first life system state parameters according to a preset scale compression method corresponding to the first life system state parameters, so as to map the parameter values ​​to a preset target parameter value range to obtain target parameter values; and predicting parameter prediction results using a trained prediction model based on the parameter names, target parameter values, and parameter value validity identifiers corresponding to the first life system state parameters, wherein the parameter prediction results include parameter values ​​corresponding to each parameter in a preset parameter set, the parameters in the preset parameter set include multiple preset second life system state parameters, and the prediction model is trained to predict parameters based on the biological homeostasis relationship between parameters.

[0088] Thus, when predicting the state parameters of a living system, the first set of state parameter data for the first living system is obtained. The parameter values ​​in this data are then scaled and compressed to map to a preset target parameter value range, yielding the target parameter values. Next, based on the parameter names, target parameter values, and parameter validity identifiers corresponding to the first living system state parameters, a trained prediction model capable of predicting parameters based on biological homeostatic relationships between parameters is used to make predictions and obtain the prediction results. These prediction results include the parameter values ​​corresponding to the second living system state parameters. Therefore, this application's solution, considering biological homeostatic relationships, enables the prediction of data corresponding to the second living system state parameters based on data corresponding to the first living system state parameters. This means it can predict the parameter values ​​of other parameters based on some parameters, thereby improving the parameter prediction effect.

[0089] like Figure 2 As shown, corresponding to the above-mentioned parameter prediction method based on biological homeostasis, this application embodiment also provides a parameter prediction system based on biological homeostasis, the above-mentioned parameter prediction system based on biological homeostasis includes: The data acquisition module 210 is used to acquire first life system state parameter data, wherein the first life system state parameter data includes parameter names, parameter values ​​and parameter value validity identifiers corresponding to multiple preset first life system state parameters; The data processing module 220 is used to perform scale compression processing on the parameter values ​​of the first life system state parameters according to the preset scale compression method corresponding to the first life system state parameters, so as to map the parameter values ​​to a preset target parameter value range and obtain the target parameter values. The prediction module 230 is used to predict the parameter prediction results based on the parameter names, target parameter values ​​and parameter value validity identifiers corresponding to the first life system state parameters, through a trained prediction model. The parameter prediction results include the parameter values ​​corresponding to each parameter in a preset parameter set. The parameters in the preset parameter set include multiple preset second life system state parameters. The prediction model is trained to predict parameters based on the biological homeostatic relationship between parameters.

[0090] Thus, when predicting the state parameters of a living system, the first set of state parameter data for the first living system is obtained. The parameter values ​​in this data are then scaled and compressed to map to a preset target parameter value range, yielding the target parameter values. Next, based on the parameter names, target parameter values, and parameter validity identifiers corresponding to the first living system state parameters, a trained prediction model capable of predicting parameters based on biological homeostatic relationships between parameters is used to make predictions and obtain the prediction results. These prediction results include the parameter values ​​corresponding to the second living system state parameters. Therefore, this application's solution, considering biological homeostatic relationships, enables the prediction of data corresponding to the second living system state parameters based on data corresponding to the first living system state parameters. This means it can predict the parameter values ​​of other parameters based on some parameters, thereby improving the parameter prediction effect.

[0091] It should be noted that the specific structure and implementation of the above-mentioned biological homeostasis-based parameter prediction system and its various modules or units can be referred to the corresponding descriptions in the above method embodiments, and will not be repeated here.

[0092] It should be noted that the division of the modules in the above-mentioned biological homeostasis-based parameter prediction system is not unique and is not intended as a specific limitation.

[0093] Based on the above embodiments, this application also provides a terminal, the principle block diagram of which can be as follows: Figure 3 As shown. The terminal includes a processor, memory, network interface, and display screen connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements the steps of any of the aforementioned biological homeostasis-based parameter prediction methods.

[0094] Those skilled in the art will understand that Figure 3 The block diagram shown is only a partial structural diagram related to the solution of this application and does not constitute a limitation on the terminal on which the solution of this application is applied. The specific terminal may include more or fewer components than shown in the figure, or combine some components, or have different component arrangements.

[0095] In one embodiment, a terminal is provided, the terminal including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of any of the biological homeostasis-based parameter prediction methods provided in the embodiments of this application.

[0096] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of any of the biological homeostasis-based parameter prediction methods provided in this application.

[0097] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0098] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the above device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above device can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0099] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0100] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0101] In the embodiments provided in this application, it should be understood that the disclosed systems / terminal devices and methods can be implemented in other ways. For example, the system / terminal device embodiments described above are merely illustrative. For instance, the division of modules or units described above is merely a logical functional division, and in actual implementation, it can be divided in other ways. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0102] If the integrated modules / units described above are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, and software distribution media, etc. It should be noted that the content included in the computer-readable storage medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction.

[0103] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions are not in essence a departure from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A parameter prediction method based on biological homeostasis, characterized in that, The method includes: Acquire first life system state parameter data, wherein the first life system state parameter data includes parameter names, parameter values ​​and parameter value validity identifiers corresponding to multiple preset first life system state parameters; For each of the first life system state parameters, the parameter values ​​of the first life system state parameters are scaled according to the preset scale compression method corresponding to the first life system state parameters, so as to map the parameter values ​​to a preset target parameter value range and obtain the target parameter values. Based on the parameter name, target parameter value, and parameter value validity identifier corresponding to the first life system state parameter, the parameter prediction result is obtained through the trained prediction model. The parameter prediction result includes the parameter value corresponding to each parameter in the preset parameter set. The parameters in the preset parameter set include multiple preset second life system state parameters. The prediction model is trained to predict parameters based on the biological homeostatic relationship between parameters.

2. The parameter prediction method based on biological homeostasis according to claim 1, characterized in that, The step of predicting parameter prediction results based on the parameter names, target parameter values, and parameter value validity identifiers corresponding to the first life system state parameters, using a trained prediction model, includes: Input the parameter name, target parameter value, and parameter value validity identifier corresponding to the first life system state parameter into the trained prediction model; In the prediction model, the parameter name, target parameter value and parameter value validity identifier corresponding to the first life system state parameter are encoded to obtain the embedded numerical representation corresponding to each first life system state parameter; The embedded numerical representation is semantically mapped to obtain a sequence of parameter representation vectors corresponding to the state parameter data of the first life system. Obtain the training validity identifier corresponding to each of the first life system state parameters, and determine the parameter validity identifier corresponding to each of the first life system state parameters based on the training validity identifier and the parameter value validity identifier; Based on the parameter validity identifier, the target parameter value, and the parameter representation vector sequence, obtain the steady-state variable vector corresponding to the state parameter data of the first life system; Based on the steady-state variable vector, the parameter prediction results are obtained.

3. The parameter prediction method based on biological homeostasis according to claim 2, characterized in that, The process of encoding the parameter name, target parameter value, and parameter value validity identifier corresponding to the first life system state parameter to obtain the embedded numerical representation corresponding to each first life system state parameter includes: For each of the first life system state parameters, a semantic embedded value is generated based on the parameter name, a parameter value embedded value is generated based on the target parameter value, and a value validity embedded value is generated based on the parameter value validity identifier. The semantic embedded value, the parameter value embedded value, and the value validity embedded value are fused to obtain the embedded value representation corresponding to the first life system state parameter.

4. The parameter prediction method based on biological homeostasis according to claim 2, characterized in that, The prediction model includes a parameter semantic mapping module, which includes at least one pre-specified target hidden layer. The step of performing semantic mapping processing on the embedded numerical representation to obtain the parameter representation vector sequence corresponding to the first life system state parameter data includes: The embedded numerical representation is input into the parameter semantic mapping module for semantic mapping processing; The output of each target hidden layer is obtained and used as the parameter representation vector sequence.

5. The parameter prediction method based on biological homeostasis according to claim 2, characterized in that, The step of determining the parameter validity identifier corresponding to each of the first life system state parameters based on the training validity identifier and the parameter value validity identifier includes: For each of the first life system state parameters, if both the training validity identifier and the parameter value validity identifier are valid, then the parameter validity identifier corresponding to the first life system state parameter is determined to be valid; otherwise, the parameter validity identifier corresponding to the first life system state parameter is determined to be invalid. Wherein, the parameter value validity identifier is used to indicate whether the parameter value of the first life system state parameter is valid; the training validity identifier is used to indicate whether the training process of the prediction model is affected by the first life system state parameter; and the parameter validity identifier is used to indicate whether the first life system state parameter is used when making predictions through the prediction model.

6. The parameter prediction method based on biological homeostasis according to claim 2, characterized in that, The step of obtaining the steady-state variable vector corresponding to the state parameter data of the first life system based on the parameter validity identifier, the target parameter value, and the parameter representation vector sequence includes: For each of the first life system state parameters, the parameter weights corresponding to the first life system state parameters are determined based on the parameter validity identifier and the target parameter value. The parameter representation vector sequence is weighted and fused according to the parameter weights to obtain the steady-state variable vector corresponding to the state parameter data of the first life system.

7. The parameter prediction method based on biological homeostasis according to claim 2, characterized in that, The parameters in the preset parameter set also include the first life system state parameters; The parameter prediction results include the parameter name, parameter value, and parameter value source indication flag for each parameter in the preset parameter set.

8. A parameter prediction system based on biological homeostasis, characterized in that, The system includes: The data acquisition module is used to acquire the first life system state parameter data, wherein the first life system state parameter data includes the parameter name, parameter value and parameter value validity identifier corresponding to a plurality of preset first life system state parameters; The data processing module is used to perform scale compression processing on the parameter values ​​of the first life system state parameters according to the preset scale compression method corresponding to the first life system state parameters, so as to map the parameter values ​​to a preset target parameter value range and obtain the target parameter values. The prediction module is used to predict parameter prediction results based on the parameter name, target parameter value, and parameter value validity identifier corresponding to the first life system state parameter, using a trained prediction model. The parameter prediction results include the parameter values ​​corresponding to each parameter in a preset parameter set. The parameters in the preset parameter set include multiple preset second life system state parameters. The prediction model is trained to predict parameters based on the biological homeostatic relationship between parameters.

9. A terminal, characterized in that, The terminal includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the parameter prediction method based on biological homeostasis as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the parameter prediction method based on biological homeostasis as described in any one of claims 1 to 7.

Images