Scheme generation method and related equipment

Abstract

The invention discloses a scheme generation method and related equipment, and relates to the technical field of artificial intelligence, and the method comprises the steps: responding to a scheme generation instruction, collecting multi-modal public health data in real time, carrying out the preprocessing of the multi-modal public health data, obtaining target data, and transmitting the target data to a server; and based on a shared attention mechanism, a domain adaptability optimization algorithm, the target data and a historical knowledge graph, performing event analysis, risk assessment and decision generation by using a preset dynamic sparse attention model to generate a target scheme. According to the invention, based on the shared attention mechanism, the preset dynamic sparse attention model is used to synchronously execute the plurality of subtasks, so that the scheme generation efficiency is improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a solution generation method and related equipment. Background Art

[0002] At present, the iteration of technology is accelerating, public health events are becoming more complex and cross-regional, and the bottleneck of traditional technology is becoming more prominent. In related technologies, as the amount of data continues to increase, the efficiency of generating solutions based on public health events is becoming lower and lower.

[0003] Therefore, how to improve the efficiency of generating plans based on public health events is an issue that needs to be addressed urgently.

[0004] The above contents are only used to assist in understanding the technical solution of the present application and do not constitute an admission that the above contents are related technologies. Summary of the invention

[0005] The main purpose of this application is to provide a plan generation method and related equipment, aiming to solve the technical problem of how to improve the efficiency of generating plans based on public health events.

[0006] To achieve the above purpose, the present application proposes a solution generation method, which includes: In response to the scheme generation instruction, multimodal public health data is collected in real time, and the multimodal public health data is preprocessed to obtain target data; Based on the shared attention mechanism, the domain adaptive optimization algorithm, the target data and the historical knowledge graph, a preset dynamic sparse attention model is used for event analysis, risk assessment and decision generation to generate a target solution.

[0007] In one embodiment, the step of preprocessing the multimodal public health data to obtain target data further includes: Based on the public health data of each modality, the corresponding encoder is used to extract feature data; Based on the contrastive learning method, the feature data of different modalities are mapped into the same semantic space, and the feature data of different modalities are fused using the multi-head cross attention mechanism to obtain the initial data; Based on the spatiotemporal consistency, the initial data is constrained to obtain constrained target data.

[0008] In one embodiment, the step of using a preset dynamic sparse attention model to perform event analysis, risk assessment, and decision generation based on a shared attention mechanism, a domain adaptive optimization algorithm, the target data, and a historical knowledge graph to generate a target solution also includes: Analyze the target data using a preset dynamic sparse attention model to obtain multiple sets of feature data; Based on the shared attention mechanism, the knowledge graph update sub-model in the preset dynamic sparse attention model is used to update the historical knowledge graph to obtain the current knowledge graph; Based on the reinforcement learning optimization framework, the current knowledge graph and the domain adaptability optimization algorithm, the decision generation model in the preset dynamic sparse attention model is used to process the multiple sets of feature data to obtain an optimized target solution.

[0009] In one embodiment, the step of updating the historical knowledge graph based on the shared attention mechanism using the knowledge graph update submodel in the preset dynamic sparse attention model to obtain the current knowledge graph also includes: Generate entity relationship sets based on multiple sets of feature data; Calculate the confidence score of each entity relationship and check the consistency of each entity relationship with existing knowledge; If the confidence score is greater than a preset confidence threshold and each entity relationship is consistent with existing knowledge, the entity relationship is added to the historical knowledge graph to obtain the current knowledge graph.

[0010] In one embodiment, the step of processing the multiple sets of feature data using a decision generation model in a preset dynamic sparse attention model based on a reinforcement learning optimization framework, the current knowledge graph, and a domain adaptability optimization algorithm to obtain an optimized target solution also includes: Based on the reinforcement learning optimization framework, the emergency resource allocation problem of public health emergencies is modeled as a Markov decision process; Based on the Markov decision process, the global decision is decomposed into a strategic level and a tactical level; Based on the current knowledge graph and the domain adaptability optimization algorithm, the decision generation model in the preset dynamic sparse attention model is used to process the multiple sets of feature data to obtain a strategic plan corresponding to the strategic layer and a tactical plan corresponding to the tactical layer; The strategic plan and the tactical plan are optimized, and a target plan is generated based on the optimized strategic plan and tactical plan.

[0011] In one embodiment, before the step of using a preset dynamic sparse attention model to perform event analysis, risk assessment, and decision generation based on a shared attention mechanism, a domain adaptive optimization algorithm, the target data, and a historical knowledge graph to generate a target solution, the step further includes: Acquire sample data, wherein the processing result corresponding to the sample data is a first decision solution; Processing the sample data using the current dynamic sparse attention model to obtain a second decision solution; Determining whether the first decision-making solution is consistent with the second decision-making solution; If there is inconsistency, adjust the parameters of the current dynamic sparse attention model, and based on the current dynamic sparse attention model with adjusted parameters, return to the step of using the current dynamic sparse attention model to process the sample data to obtain a second decision plan, until the first decision plan is consistent with the second decision plan to obtain a preset dynamic sparse attention model.

[0012] In addition, to achieve the above purpose, the present application also proposes a solution generation device, the solution generation device comprising: A data processing module, wherein the data processing module collects multimodal public health data in real time in response to the scheme generation instruction, and pre-processes the multimodal public health data to obtain target data; A generation module is used to perform event analysis, risk assessment and decision generation based on a shared attention mechanism, a domain adaptability optimization algorithm, the target data and a historical knowledge graph using a preset dynamic sparse attention model to generate a target solution.

[0013] In one embodiment, the data processing module includes: An extraction unit, for extracting feature data based on the public health data of each modality using an encoder corresponding thereto; A mapping unit is used to map feature data of different modalities into the same semantic space based on contrastive learning, so as to use a multi-head cross-attention mechanism to fuse feature data between different modalities and obtain initial data; The constraint unit is used to constrain the initial data based on spatiotemporal consistency to obtain constrained target data.

[0014] In one embodiment, the generating module comprises: A parsing unit, configured to parse the target data using a preset dynamic sparse attention model to obtain multiple sets of feature data; An updating unit, which is used to update the historical knowledge graph based on a shared attention mechanism using a knowledge graph updating sub-model in a preset dynamic sparse attention model to obtain a current knowledge graph; The first generation unit is used to process the multiple groups of feature data using a decision generation model in a preset dynamic sparse attention model based on a reinforcement learning optimization framework, the current knowledge graph and a domain adaptability optimization algorithm to obtain an optimized target solution.

[0015] In one embodiment, the generating module further includes: A second generating unit, used for generating an entity relationship set based on the multiple sets of feature data; A calculation unit, used to calculate the confidence score of each entity relationship and check the consistency of each entity relationship with existing knowledge; An adding unit is used to add the entity relationship to the historical knowledge graph to obtain the current knowledge graph if the confidence score is greater than a preset confidence threshold and each entity relationship is consistent with the existing knowledge.

[0016] In one embodiment, the generating module further includes: A modeling unit, which is used to model the emergency resource allocation problem of public health emergencies as a Markov decision process based on a reinforcement learning optimization framework; A decomposition unit, used for decomposing the global decision into a strategic layer and a tactical layer based on the Markov decision process; A first data processing unit is used to process the multiple sets of feature data using a decision generation model in a preset dynamic sparse attention model based on the current knowledge graph and the domain adaptability optimization algorithm to obtain a strategic plan corresponding to the strategic layer and a tactical plan corresponding to the tactical layer; The optimization unit is used to optimize the strategic plan and the tactical plan, and generate a target plan based on the optimized strategic plan and tactical plan.

[0017] In one embodiment, the solution generating device further includes a training module, and the training module includes: An acquisition unit, used for acquiring sample data, wherein the processing result corresponding to the sample data is a first decision solution; A second data processing unit, configured to process the sample data using the current dynamic sparse attention model to obtain a second decision solution; A judging unit, configured to judge whether the first decision-making scheme is consistent with the second decision-making scheme; The training unit is used to adjust the parameters of the current dynamic sparse attention model if there is any inconsistency, and based on the current dynamic sparse attention model after the adjustment of the parameters, return to the step of using the current dynamic sparse attention model to process the sample data to obtain a second decision plan, until the first decision plan is consistent with the second decision plan to obtain a preset dynamic sparse attention model.

[0018] In addition, to achieve the above-mentioned purpose, the present application also proposes a solution generation device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the solution generation method described above.

[0019] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the solution generation method described above are implemented.

[0020] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps of the solution generation method described above are implemented.

[0021] One or more technical solutions proposed in this application have at least the following technical effects: The present application proposes a solution generation method and related equipment, which relate to the field of artificial intelligence technology. Compared with the related technology in which the efficiency of generating solutions based on public health events is very low, in the present application, first, in response to the solution generation instruction, multimodal public health data is collected in real time, and the multimodal public health data is preprocessed to obtain target data. Then, based on the shared attention mechanism, the domain adaptive optimization algorithm, the target data and the historical knowledge graph, a preset dynamic sparse attention model is used to perform event analysis, risk assessment and decision generation to generate a target solution.

[0022] It can be understood that the present application is based on a shared attention mechanism and uses a preset dynamic sparse attention model to directly and synchronously execute multiple subtasks, thereby improving the efficiency of solution generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0024] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0025] Figure 1 A flowchart of the first embodiment of the method for generating a solution of the present application is provided; Figure 2 A schematic diagram of a sparse attention model provided in Example 1 of the method for generating the solution of this application; Figure 3 A schematic diagram of a flow chart for implementing the second embodiment of the generation method of this application; Figure 4 This is a schematic diagram of the module structure of the scheme generating device of the embodiment of the present application; Figure 5Schematic diagram of the device structure of the hardware operating environment involved in the solution generation method in the embodiment of the present application.

[0026] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0027] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0028] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0029] The main solutions of the embodiments of this application are: In this embodiment, for the convenience of description, the following description is made with the solution generating device as the execution subject.

[0030] Due to existing technologies, as the amount of data continues to increase, the efficiency of generating solutions based on public health events is becoming increasingly lower.

[0031] The present application provides a solution, which enables: first, in response to a solution generation instruction, multimodal public health data is collected in real time, the multimodal public health data is preprocessed to obtain target data, and then, based on a shared attention mechanism, a domain adaptability optimization algorithm, the target data and a historical knowledge graph, a preset dynamic sparse attention model is used to perform event analysis, risk assessment and decision generation to generate a target solution. It can be understood that the present application is based on a shared attention mechanism and uses a preset dynamic sparse attention model to directly and synchronously execute multiple subtasks, thereby improving the efficiency of solution generation.

[0032] It should be noted that the execution subject of this embodiment may be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, a solution generation device, etc. The following takes the solution generation device as an example to illustrate this embodiment and the following embodiments.

[0033] Based on this, the present application embodiment provides a solution generation method, referring to Figure 1 , Figure 1 This is a flowchart of the first embodiment of the method for generating a solution of the present application.

[0034] In this embodiment, the solution generation method includes steps S100 to S200: Step S100, in response to a plan generation instruction, collecting multimodal public health data in real time, and preprocessing the multimodal public health data to obtain target data; It should be noted that the execution subject of this embodiment may be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, a solution generation device, etc. The following takes the solution generation device as an example to illustrate this embodiment and the following embodiments.

[0035] It should be noted that multimodal data refers to data that contains multiple types, such as: text data (news reports, social media posts, etc.), image data (medical images, on-site photos, surveillance video screenshots, etc.), numerical data (case statistics, test results, environmental data, etc.), and audio data (voice broadcasts, meeting minutes, etc.).

[0036] It should be noted that the purpose of preprocessing is to convert the collected raw data into a format suitable for subsequent processing, improve data quality, and reduce noise and redundancy.

[0037] Specifically, the step of preprocessing the multimodal public health data to obtain target data further includes steps S110 to S130: Step S110, based on the public health data of each modality, extract feature data using the encoder corresponding thereto; It should be noted that modalities can include different types of data sources, such as text (such as case reports, news reports), images (such as medical images, surveillance video screenshots), and time series data (such as disease incidence, environmental monitoring data, etc.).

[0038] It should be noted that for text modality, natural language processing (NLP) technology is usually used, such as encoders such as BERT and Transformer, to convert text into word embedding vectors to capture the semantic information in the text.

[0039] It should be noted that for the image modality, a convolutional neural network (CNN) encoder, such as ResNet and VGG, is used to extract the spatial features of the image and capture the texture, shape and other information in the image.

[0040] It should be noted that for time series modality, a recurrent neural network (RNN) or Transformer encoder is used to extract the time dependency and trend information in the time series data.

[0041] Step S120, based on the contrastive learning method, the feature data of different modalities are mapped into the same semantic space, and the feature data between different modalities are fused using a multi-head cross attention mechanism to obtain initial data; It should be noted that contrastive learning is an unsupervised learning method that makes the features of different modalities comparable in the same space by learning the similarities between similar samples and the differences between different samples.

[0042] It should be noted that the multi-head cross-attention mechanism can capture the relationship between different modalities, such as the relationship between text and images, or the relationship between time series data and text. Specifically, it learns the interaction between different modalities through multiple attention heads, and then performs weighted summation of these interaction results to obtain the fused features.

[0043] Step S130: constraining the initial data based on spatiotemporal consistency to obtain constrained target data.

[0044] It should be noted that spatiotemporal consistency refers to the continuity and consistency of data in time and space. For example, in public health data, the spread of disease has temporal and spatial regularities, so constraints are needed to ensure that the data conforms to this regularity.

[0045] It should be noted that spatiotemporal constraints can be introduced, such as temporal smoothness constraints (ensuring that data is smooth in time and has no mutations) and spatial consistency constraints (ensuring that data is reasonable in space, for example, data in adjacent areas should have a certain similarity). These constraints can be achieved by optimizing the objective function, for example, by adding spatiotemporal consistency constraints to the loss function.

[0046] Step S200, based on the shared attention mechanism, the domain adaptability optimization algorithm, the target data and the historical knowledge graph, uses a preset dynamic sparse attention model to perform event analysis, risk assessment and decision generation to generate a target solution.

[0047] It should be noted that the core of the shared attention mechanism is to reduce redundant calculations through parameter sharing while improving the efficiency and performance of the model. For example, the SAPT framework proposed by Harbin Institute of Technology enables the learning module and the selection module to work in coordination by sharing attention weights, thereby solving the problems of catastrophic forgetting and knowledge transfer. This mechanism can effectively utilize the information in the historical knowledge graph and provide richer contextual support for event analysis.

[0048] It should be noted that the domain-adaptive optimization algorithm is designed to enable the model to better adapt to the data distribution in a specific field. Through the optimization algorithm, the model can dynamically adjust parameters to better handle specific features and patterns in the target data. This helps to improve the accuracy and reliability of the model in public health event analysis, especially when faced with data from different regions or at different times.

[0049] It should be noted that the target data provides detailed information about current events, while the historical knowledge graph contains structured knowledge about past events. Combining the two can provide a more comprehensive input for the model, helping it to better understand the context and potential impact of events. For example, through the historical event association information in the knowledge graph, the model can more accurately assess the risk level of the current event.

[0050] Understandably, reference can be made to Figure 2 Sparse attention models (such as the NSA model) significantly improve the efficiency of long text modeling through a dynamic hierarchical sparse strategy, combined with coarse-grained token compression and fine-grained token selection. This model can dynamically adjust the attention distribution according to the characteristics of the input data, thereby more efficiently extracting key information in event analysis. For example, when processing time series data of public health events, dynamic sparse attention can quickly locate key time nodes and improve analysis efficiency.

[0051] It can be understood that based on the above mechanism, the model can conduct in-depth analysis of the target data and extract the key features and patterns of the events.

[0052] Furthermore, the dynamic sparse attention model and the information in the historical knowledge graph are used to quantitatively assess the potential risks of the event.

[0053] Furthermore, targeted decision-making plans are generated based on the analysis and evaluation results. The efficiency and adaptability of the dynamic sparse attention model enable it to quickly respond to complex and changing public health events.

[0054] Finally, by integrating the shared attention mechanism, domain-adaptive optimization algorithm, target data, and historical knowledge graph, the dynamic sparse attention model can generate optimized target solutions. These solutions are not only based on the analysis of current data, but also combine historical experience, thus providing more scientific and effective decision support in responding to public health events.

[0055] Through the above steps, efficient analysis, accurate assessment and scientific decision-making of public health events can be achieved, and the public health response capabilities can be improved.

[0056] Specifically, the step of using a preset dynamic sparse attention model to perform event analysis, risk assessment and decision generation based on the shared attention mechanism, the domain adaptive optimization algorithm, the target data and the historical knowledge graph to generate a target solution also includes steps S210 to S230: Step S210, using a preset dynamic sparse attention model to parse the target data to obtain multiple sets of feature data; Step S220, based on the shared attention mechanism, the historical knowledge graph is updated using the knowledge graph update sub-model in the preset dynamic sparse attention model to obtain the current knowledge graph; Specifically, the step of updating the historical knowledge graph based on the shared attention mechanism using the knowledge graph update submodel in the preset dynamic sparse attention model to obtain the current knowledge graph also includes steps S221 to S223: Step S221, generating an entity relationship set based on multiple sets of feature data; It can be understood that, first, entities are identified from feature data, for example, entities such as names of people, places, and organizations are extracted from text data, and entities such as objects and scenes are identified from image data.

[0057] Then, based on the identified entities, the relationships between entities are extracted. For example, from text, relationships such as "XX belongs to YY" and "XX leads to YY" are extracted; from images, spatial relationships such as "XX is next to YY" are extracted.

[0058] Furthermore, the identified entities and their relationships are combined into a set, namely, the entity relationship set.

[0059] Step S222, calculating the confidence score of each entity relationship, and checking the consistency of each entity relationship with existing knowledge; In this embodiment, a machine learning or deep learning model (such as a relationship extraction model) is used to generate a confidence score for each entity relationship, indicating the model's confidence in the prediction of the relationship. The similarity between entity features (such as cosine similarity, Jaccard similarity, etc.) is calculated and used as the basis for the confidence score.

[0060] Furthermore, each entity relationship is compared with the knowledge in the existing knowledge graph to check whether there are conflicts or inconsistencies. For example, if it is known in the knowledge graph that "XX is a subclass of YY", and the newly extracted relationship is "XX is not a subclass of YY", it is considered inconsistent.

[0061] Step S223: If the confidence score is greater than a preset confidence threshold and each entity relationship is consistent with existing knowledge, the entity relationship is added to the historical knowledge graph to obtain the current knowledge graph.

[0062] It should be noted that the confidence threshold is a preset confidence threshold (such as 0.8), and only entity relationships with a confidence score greater than the threshold are considered reliable. If an entity relationship meets the above conditions, it will be added to the historical knowledge graph and the content of the knowledge graph will be updated. The updated knowledge graph is called the current knowledge graph.

[0063] This method can effectively integrate the knowledge in new data into the knowledge graph while ensuring the quality and consistency of the knowledge graph, providing support for subsequent knowledge management and application.

[0064] Step S230, based on the reinforcement learning optimization framework, the current knowledge graph and the domain adaptability optimization algorithm, use the decision generation model in the preset dynamic sparse attention model to process the multiple groups of feature data to obtain an optimized target solution.

[0065] Specifically, the step of processing the multiple sets of feature data using a decision generation model in a preset dynamic sparse attention model based on the reinforcement learning optimization framework, the current knowledge graph and the domain adaptability optimization algorithm to obtain an optimized target solution also includes steps S231 to S234: Step S231, based on the reinforcement learning optimization framework, modeling the emergency resource allocation problem of public health emergencies as a Markov decision process; It should be noted that the Markov decision process (MDP) modeling includes states, actions, rewards, and transition probabilities. Specifically: State: Defined as the current status of the event, including the progress of the event, the allocation of resources, the status of the affected area, etc.

[0066] Action: Defined as a resource allocation decision, such as allocating a certain resource to a certain area or task.

[0067] Reward: defined as the effect of resource allocation, such as reduced number of infections, improved rescue efficiency, reduced economic losses, etc.

[0068] Transition Probability: describes the probability of transitioning to the next state after taking a certain action in the current state.

[0069] It should be noted that reinforcement learning algorithms (such as Q-learning, DQN, PPO, etc.) are used to optimize resource allocation strategies. Through interaction with the environment, the optimal sequence of resource allocation actions is learned to maximize the cumulative reward.

[0070] Step S232, based on the Markov decision process, decomposing the global decision into a strategic layer and a tactical layer; It should be noted that strategic-level decision-making is a long-term, macro-level resource allocation strategy, such as the overall resource allocation direction, priority sorting, etc. The optimal strategy at the strategic level can be determined by analyzing the state space and reward function of the MDP.

[0071] It should be noted that tactical-level decisions are short-term, specific resource allocation actions, such as allocating specific resources to a specific area at a certain point in time, which can be further refined into a specific sequence of actions based on the results of strategic-level decisions.

[0072] Step S233, based on the current knowledge graph and the domain adaptability optimization algorithm, use the decision generation model in the preset dynamic sparse attention model to process the multiple sets of feature data to obtain a strategic plan corresponding to the strategic layer and a tactical plan corresponding to the tactical layer; Step S234, optimizing the strategic plan and the tactical plan, and generating a target plan based on the optimized strategic plan and tactical plan.

[0073] It should be noted that in this embodiment, the tactical plan can be adjusted through optimization algorithms (such as genetic algorithms, simulated annealing, etc.) to improve the efficiency and effectiveness of resource allocation. Further, the optimized strategic plan and tactical plan are integrated to generate the final target plan.

[0074] It should be noted that, in the present application, a monitoring, evaluation and feedback module is also included, which is used to continuously monitor the development of events and the effects of interventions, and input feedback information into the system for self-optimization.

[0075] The present application proposes a solution generation method and related equipment, which relate to the field of artificial intelligence technology. Compared with the related technology in which the efficiency of generating solutions based on public health events is very low, in the present application, first, in response to the solution generation instruction, multimodal public health data is collected in real time, and the multimodal public health data is preprocessed to obtain target data. Then, based on the shared attention mechanism, the domain adaptive optimization algorithm, the target data and the historical knowledge graph, a preset dynamic sparse attention model is used to perform event analysis, risk assessment and decision generation to generate a target solution.

[0076] It can be understood that the present application is based on a shared attention mechanism and uses a preset dynamic sparse attention model to directly and synchronously execute multiple subtasks, thereby improving the efficiency of solution generation.

[0077] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can refer to the above introduction, and will not be repeated later. Figure 3 Before step S200, the solution generation method further includes steps A10 to A40: Step A10, obtaining sample data, wherein the processing result corresponding to the sample data is a first decision solution; It should be noted that sample data is a data set used to train and verify the model. This data usually contains input features and corresponding processing results (i.e. known correct decision solutions). Sample data will serve as the basis for model training and adjustment.

[0078] Step A20, using the current dynamic sparse attention model to process the sample data to obtain a second decision solution; It should be noted that the second decision plan is a decision plan generated by the model according to the current parameters and is used for comparison with the first decision plan.

[0079] Step A30, determining whether the first decision-making scheme is consistent with the second decision-making scheme; Step A40, if there is inconsistency, adjust the parameters of the current dynamic sparse attention model, and based on the current dynamic sparse attention model after adjusting the parameters, return to the step of using the current dynamic sparse attention model to process the sample data to obtain a second decision plan, until the first decision plan is consistent with the second decision plan, and obtain the preset dynamic sparse attention model.

[0080] It is understandable that if the first decision plan is inconsistent with the second decision plan, the parameters of the dynamic sparse attention model need to be adjusted. The adjustment methods include gradient descent, reinforcement learning, genetic algorithm, Bayesian optimization, etc.

[0081] It should be noted that reinforcement learning uses the difference between the model's output and the target solution as a reward signal and adjusts the parameters through the reinforcement learning algorithm.

[0082] This process is a typical model optimization process, which continuously adjusts the model parameters to make its output consistent with the known correct results. In this way, it can ensure that the model has good performance on known data and provide reliable decision support for subsequent new data processing.

[0083] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the method for generating the solution of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0084] This application also provides a solution generation device, please refer to Figure 4 , the solution generating device comprises: A data processing module 10, which collects multimodal public health data in real time in response to a scheme generation instruction, and preprocesses the multimodal public health data to obtain target data; Generation module 20, the generation module is used to perform event analysis, risk assessment and decision generation based on a shared attention mechanism, a domain adaptability optimization algorithm, the target data and a historical knowledge graph using a preset dynamic sparse attention model to generate a target solution.

[0085] In one embodiment, the data processing module includes: An extraction unit, for extracting feature data based on the public health data of each modality using an encoder corresponding thereto; A mapping unit is used to map feature data of different modalities into the same semantic space based on contrastive learning, so as to use a multi-head cross-attention mechanism to fuse feature data between different modalities and obtain initial data; The constraint unit is used to constrain the initial data based on spatiotemporal consistency to obtain constrained target data.

[0086] In one embodiment, the generating module comprises: A parsing unit, configured to parse the target data using a preset dynamic sparse attention model to obtain multiple sets of feature data; An updating unit, which is used to update the historical knowledge graph based on a shared attention mechanism using a knowledge graph updating sub-model in a preset dynamic sparse attention model to obtain a current knowledge graph; The first generation unit is used to process the multiple groups of feature data using a decision generation model in a preset dynamic sparse attention model based on a reinforcement learning optimization framework, the current knowledge graph and a domain adaptability optimization algorithm to obtain an optimized target solution.

[0087] In one embodiment, the generating module further includes: A second generating unit, used for generating an entity relationship set based on the multiple sets of feature data; A calculation unit, used to calculate the confidence score of each entity relationship and check the consistency of each entity relationship with existing knowledge; An adding unit is used to add the entity relationship to the historical knowledge graph to obtain the current knowledge graph if the confidence score is greater than a preset confidence threshold and each entity relationship is consistent with the existing knowledge.

[0088] In one embodiment, the generating module further includes: A modeling unit, which is used to model the emergency resource allocation problem of public health emergencies as a Markov decision process based on a reinforcement learning optimization framework; A decomposition unit, used for decomposing the global decision into a strategic layer and a tactical layer based on the Markov decision process; A first data processing unit is used to process the multiple sets of feature data using a decision generation model in a preset dynamic sparse attention model based on the current knowledge graph and the domain adaptability optimization algorithm to obtain a strategic plan corresponding to the strategic layer and a tactical plan corresponding to the tactical layer; The optimization unit is used to optimize the strategic plan and the tactical plan, and generate a target plan based on the optimized strategic plan and tactical plan.

[0089] In one embodiment, the solution generating device further includes a training module, and the training module includes: An acquisition unit, used for acquiring sample data, wherein the processing result corresponding to the sample data is a first decision solution; A second data processing unit, configured to process the sample data using the current dynamic sparse attention model to obtain a second decision solution; A judging unit, configured to judge whether the first decision-making scheme is consistent with the second decision-making scheme; The training unit is used to adjust the parameters of the current dynamic sparse attention model if there is any inconsistency, and based on the current dynamic sparse attention model after the adjustment of the parameters, return to the step of using the current dynamic sparse attention model to process the sample data to obtain a second decision plan, until the first decision plan is consistent with the second decision plan to obtain a preset dynamic sparse attention model.

[0090] The solution generation device provided by the present application adopts the solution generation method in the above embodiment to solve the technical problem of solution generation. Compared with the prior art, the beneficial effects of the solution generation device provided by the present application are the same as the beneficial effects of the solution generation method provided by the above embodiment, and other technical features in the solution generation device are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0091] The present application provides a solution generation device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the solution generation method in the above-mentioned embodiment 1.

[0092] Reference below Figure 5, which shows a schematic diagram of the structure of a solution generation device suitable for implementing the embodiment of the present application. The solution generation device in the embodiment of the present application may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 5 The solution generation device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0093] like Figure 5 As shown, the solution generation device may include a processing device 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the solution generation device are also stored. The processing device 1001, ROM1002, and RAM1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the solution generation device to communicate with other devices wirelessly or by wire to exchange data. Although the solution generation device with various systems is shown in the figure, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or have alternatively.

[0094] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0095] The solution generation device provided by the present application adopts the solution generation method in the above embodiment to solve the technical problem. Compared with the prior art, the beneficial effects of the solution generation device provided by the present application are the same as the beneficial effects of the solution generation method provided by the above embodiment, and the other technical features in the solution generation device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0096] It should be understood that the various parts disclosed in this application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0097] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0098] The present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the solution generation method in the above-mentioned embodiment.

[0099] The computer-readable storage medium provided in the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM: Random Access Memory), a read-only memory (ROM: Read Only Memory), an erasable programmable read-only memory (EPROM: Erasable Programmable Read Only Memory or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM: CD-Read Only Memory), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency: Radio Frequency), etc., or any suitable combination of the above.

[0100] The computer-readable storage medium may be included in the solution generation device; or may exist independently without being assembled into the solution generation device.

[0101] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the solution generation device, the solution generation device: In response to the scheme generation instruction, multimodal public health data is collected in real time, and the multimodal public health data is preprocessed to obtain target data; Based on the shared attention mechanism, the domain adaptive optimization algorithm, the target data and the historical knowledge graph, a preset dynamic sparse attention model is used for event analysis, risk assessment and decision generation to generate a target solution.

[0102] Computer program code for performing the operations of the present application may be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0103] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0104] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0105] The readable storage medium provided in this application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned solution generation method, and can solve the technical problem of solution generation. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as the beneficial effects of the solution generation method provided in the above-mentioned embodiment, and will not be repeated here.

[0106] The present application also provides a computer program product, including a computer program, which implements the steps of the solution generation method as described above when executed by a processor.

[0107] The computer program product provided by this application can solve the technical problem of solution generation. Compared with the prior art, the beneficial effects of the computer program product provided by this application are the same as the beneficial effects of the solution generation method provided by the above embodiment, which will not be repeated here.

[0108] The above descriptions are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect applications in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A solution generation method, characterized in that: The solution generation method comprises: In response to the scheme generation instruction, multimodal public health data is collected in real time, and the multimodal public health data is preprocessed to obtain target data; Based on the shared attention mechanism, the domain adaptive optimization algorithm, the target data and the historical knowledge graph, a preset dynamic sparse attention model is used for event analysis, risk assessment and decision generation to generate a target solution.

2. The solution generation method according to claim 1, characterized in that: The step of preprocessing the multimodal public health data to obtain target data further includes: Based on the public health data of each modality, the corresponding encoder is used to extract feature data; Based on the contrastive learning method, the feature data of different modalities are mapped into the same semantic space, and the feature data of different modalities are fused using the multi-head cross attention mechanism to obtain the initial data; Based on the spatiotemporal consistency, the initial data is constrained to obtain constrained target data.

3. The solution generation method according to claim 1, characterized in that: The step of using a preset dynamic sparse attention model to perform event analysis, risk assessment and decision generation based on the shared attention mechanism, the domain adaptive optimization algorithm, the target data and the historical knowledge graph to generate a target solution also includes: Analyze the target data using a preset dynamic sparse attention model to obtain multiple sets of feature data; Based on the shared attention mechanism, the knowledge graph update sub-model in the preset dynamic sparse attention model is used to update the historical knowledge graph to obtain the current knowledge graph; Based on the reinforcement learning optimization framework, the current knowledge graph and the domain adaptability optimization algorithm, the decision generation model in the preset dynamic sparse attention model is used to process the multiple sets of feature data to obtain an optimized target solution.

4. The solution generation method according to claim 3, characterized in that: The step of updating the historical knowledge graph based on the shared attention mechanism using the knowledge graph update sub-model in the preset dynamic sparse attention model to obtain the current knowledge graph also includes: Generate entity relationship sets based on multiple sets of feature data; Calculate the confidence score of each entity relationship and check the consistency of each entity relationship with existing knowledge; If the confidence score is greater than a preset confidence threshold and each entity relationship is consistent with existing knowledge, the entity relationship is added to the historical knowledge graph to obtain the current knowledge graph.

5. The solution generation method according to claim 3, characterized in that: The step of processing the multiple sets of feature data using a decision generation model in a preset dynamic sparse attention model based on the reinforcement learning optimization framework, the current knowledge graph and the domain adaptability optimization algorithm to obtain an optimized target solution also includes: Based on the reinforcement learning optimization framework, the emergency resource allocation problem of public health emergencies is modeled as a Markov decision process; Based on the Markov decision process, the global decision is decomposed into a strategic level and a tactical level; Based on the current knowledge graph and the domain adaptability optimization algorithm, the decision generation model in the preset dynamic sparse attention model is used to process the multiple sets of feature data to obtain a strategic plan corresponding to the strategic layer and a tactical plan corresponding to the tactical layer; The strategic plan and the tactical plan are optimized, and a target plan is generated based on the optimized strategic plan and tactical plan.

6. The solution generation method according to claim 1, characterized in that: Before the step of using a preset dynamic sparse attention model to perform event analysis, risk assessment, and decision generation based on a shared attention mechanism, a domain adaptive optimization algorithm, the target data, and a historical knowledge graph to generate a target solution, the step further includes: Acquire sample data, wherein the processing result corresponding to the sample data is a first decision solution; Processing the sample data using the current dynamic sparse attention model to obtain a second decision solution; Determining whether the first decision-making solution is consistent with the second decision-making solution; If there is inconsistency, adjust the parameters of the current dynamic sparse attention model, and based on the current dynamic sparse attention model with adjusted parameters, return to the step of using the current dynamic sparse attention model to process the sample data to obtain a second decision plan, until the first decision plan is consistent with the second decision plan to obtain a preset dynamic sparse attention model.

7. A solution generating device, characterized in that: The solution generating device comprises: A data processing module, wherein the data processing module collects multimodal public health data in real time in response to the scheme generation instruction, pre-processes the multimodal public health data, and obtains target data; A generation module is used to perform event analysis, risk assessment and decision generation based on a shared attention mechanism, a domain adaptability optimization algorithm, the target data and a historical knowledge graph using a preset dynamic sparse attention model to generate a target solution.

8. A solution generation device, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the solution generation method according to any one of claims 1 to 6.

9. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the solution generation method according to any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the solution generation method according to any one of claims 1 to 6 are implemented.

Images