Artificial intelligence-based medical resource intelligent scheduling method and system

Abstract

This invention discloses an artificial intelligence-based intelligent scheduling method and system for medical resources, relating to the technical field of medical resource scheduling. The method includes: real-time acquisition and construction of a multi-source heterogeneous dataset of medical resources; real-time inference and prediction to obtain resource demand prediction results, conducting conflict assessment of medical resources, and obtaining resource conflict risk values; triggering an intelligent scheduling engine to optimize the multi-objective allocation of medical resources and formulate an intelligent scheduling scheme; simulating the execution of the intelligent scheduling scheme to monitor key aspects of medical resources in real time, generating adjustment suggestions, updating the intelligent scheduling scheme and automatically executing the scheduling, and generating a medical resource scheduling report. This invention solves the technical problems in existing technologies that struggle to cope with the complexity of heterogeneous medical data, the volatility of resource demand, and multi-objective conflicts, leading to poor efficiency and real-time performance in medical resource scheduling. It achieves the technical effect of accurate prediction and intelligent scheduling of resource demand, improving the efficiency and real-time performance of medical resource scheduling.

Description

Technical Field

[0001] This invention relates to the field of medical resource scheduling technology, specifically to a method and system for intelligent scheduling of medical resources based on artificial intelligence. Background Technology

[0002] With the rising incidence of chronic diseases, the contradiction between the supply and demand of medical resources is becoming increasingly prominent, especially during public health emergencies, when the medical system faces challenges such as resource strain and uneven distribution. Traditional medical resource scheduling, based on static management with fixed rules, is ill-suited to cope with complex and ever-changing medical needs. Furthermore, medical resource data is often scattered across different systems such as hospital information systems, drug inventory systems, and emergency dispatch platforms, further increasing the difficulty of medical resource scheduling. While artificial intelligence can extract valuable information from multi-source heterogeneous data, enabling accurate prediction and dynamic optimization of resource demand, it lacks the ability to coordinate multi-resource scheduling. Additionally, the mechanisms for resource conflict risk assessment and real-time adjustment are still imperfect, resulting in insufficient adaptability of scheduling schemes in dynamic environments, thus affecting the efficiency and reliability of medical resource utilization.

[0003] Therefore, current technologies face challenges in addressing the complexity of heterogeneous medical data, the volatility of resource demands, and conflicts among multiple objectives, resulting in poor efficiency and real-time performance in medical resource scheduling. Summary of the Invention

[0004] This application provides an AI-based intelligent medical resource scheduling method and system, which solves the technical problems in the prior art that make it difficult to cope with the complexity of heterogeneous medical data, the volatility of resource demand, and the conflict of multiple objectives, resulting in poor efficiency and real-time performance of medical resource scheduling. It achieves the technical effect of realizing accurate prediction and intelligent scheduling of resource demand, and improving the efficiency and real-time performance of medical resource scheduling.

[0005] This application provides an artificial intelligence-based intelligent scheduling method for medical resources. The method includes: real-time data acquisition through a multi-source data acquisition module to construct a multi-source heterogeneous dataset of medical resources; real-time inference and prediction based on the multi-source heterogeneous dataset of medical resources to obtain resource demand prediction results; conflict assessment of medical resources based on the resource demand prediction results to obtain resource conflict risk values; triggering an intelligent scheduling engine based on the resource conflict risk values ​​to optimize the multi-objective allocation of medical resources and formulate an intelligent scheduling scheme; simulating the execution of the intelligent scheduling scheme to monitor key aspects of medical resources in real time, generating adjustment suggestions; updating the intelligent scheduling scheme based on the adjustment suggestions and automatically scheduling and executing the scheme to generate a medical resource scheduling report.

[0006] In a possible implementation, the AI-based intelligent medical resource scheduling method further performs the following processes: establishing medical data access rules, activating information interfaces according to the medical data access rules for real-time data collection, and obtaining a first dataset; monitoring medical devices in real-time through IoT sensors to obtain a second dataset; performing regional analysis based on a public health platform to retrieve a third dataset; cleaning the first dataset, the second dataset, and the third dataset, and standardizing and fusing them according to the cleaning results to construct the multi-source heterogeneous dataset of medical resources.

[0007] In a possible implementation, the AI-based intelligent medical resource scheduling method further performs the following processes: feature analysis based on a multi-source heterogeneous dataset of medical resources to obtain patient characteristics, time-series characteristics, and resource status characteristics; multi-dimensional resource demand extrapolation based on the patient characteristics, time-series characteristics, and resource status characteristics to obtain patient demand extrapolation results, equipment demand extrapolation results, and personnel demand extrapolation results; prediction of the patient demand extrapolation results, equipment demand extrapolation results, and personnel demand extrapolation results using a random forest algorithm to obtain resource demand prediction results, which include predicted patient inflow, predicted equipment demand, and predicted personnel demand; supply-demand matching of the predicted patient inflow, predicted equipment demand, and predicted personnel demand; graph theory analysis based on the matching results to determine resource dependencies; traversing the resource demand prediction results according to the resource dependencies to locate resource competition hotspot areas; conflict assessment based on the resource competition hotspot areas to generate resource conflict risk values.

[0008] In a possible implementation, the AI-based intelligent medical resource scheduling method further performs the following processing: constructing a multi-dimensional feature input layer, synchronizing the patient demand projection results, equipment demand projection results, and personnel demand projection results to the multi-dimensional feature input layer for analysis, and extracting patient visit characteristics, equipment operation characteristics, and personnel load characteristics; based on the patient visit characteristics, equipment operation characteristics, and personnel load characteristics, dividing the patient demand projection results, equipment demand projection results, and personnel demand projection results into patient flow prediction sub-tasks, equipment demand prediction tasks, and personnel demand prediction tasks; and based on the patient flow prediction sub-tasks... A first random forest model is constructed based on the patient visit characteristics; a second random forest model is constructed based on the equipment demand prediction task and the equipment operation characteristics; a third random forest model is constructed based on the personnel demand prediction task and the personnel load characteristics; the first, second, and third random forest models are performed in parallel and collaboratively to generate initial resource demand prediction results; the reliability of the initial resource demand prediction results is evaluated, and the initial resource demand prediction results are updated in multiple dimensions according to the reliability, and the predicted patient inflow, predicted equipment demand, and predicted personnel demand are integrated to generate the final resource demand prediction result.

[0009] In a possible implementation, the AI-based intelligent medical resource scheduling method further performs the following processing: establishing a parallel model collaboration framework; performing parallel communication analysis on the first random forest model, the second random forest model, and the third random forest model based on the parallel model collaboration framework, and setting parallel execution environment information; coordinating and sharing the first random forest model, the second random forest model, and the third random forest model according to the parallel execution environment information to generate multiple model prediction results; performing collaborative evaluation based on the multiple model prediction results; and dynamically fusing the multiple model prediction results when the collaborative efficiency is higher than the expected efficiency threshold to generate the initial resource demand prediction result.

[0010] In a possible implementation, the AI-based intelligent medical resource scheduling method further performs the following processes: mapping the resource dependencies to the resource demand prediction results for traversal and identification, constructing a resource dependency graph; performing a depth-first traversal of the resource dependency graph to extract resource temporal overlap parameters and resource spatial overlap parameters; identifying peak resource demand periods and generating a hotspot distribution map based on the resource temporal overlap parameters and the resource spatial overlap parameters; performing data competition analysis based on the hotspot distribution map to locate resource competition hotspot areas; performing multi-dimensional, layer-by-layer traversal evaluation according to the resource dependencies based on the resource competition hotspot areas to generate temporal conflict evaluation results and spatial conflict evaluation results; and performing risk aggregation calculation based on the temporal conflict evaluation results and spatial conflict evaluation results to generate the resource conflict risk value.

[0011] In a possible implementation, the AI-based intelligent medical resource scheduling method further performs the following processes: constructing multi-level risk trigger threshold intervals; matching the multi-level risk trigger threshold intervals based on the resource conflict risk value to determine the target scheduling response mode; performing scheduling impact analysis on medical resources according to the target scheduling response mode and setting scheduling priorities; generating trigger signals based on the scheduling priorities and activating the intelligent scheduling engine based on the trigger signals; allocating and scheduling medical resources through the intelligent scheduling engine to generate initial resource allocation information; performing multi-objective optimization based on the initial resource allocation information to determine multiple optimization objectives; performing scheduling verification based on the multiple optimization objectives; and formulating the intelligent scheduling scheme based on the verification results.

[0012] In a possible implementation, the AI-based intelligent medical resource scheduling method further performs the following processes: constructing digital twin environment information; loading the intelligent scheduling scheme based on the digital twin environment information to simulate the execution of key aspects of medical resources, generating a simulation scheduling parameter set; performing multi-dimensional operational status perception of key aspects of medical resources based on the simulation scheduling parameter set, obtaining a real-time monitoring dataset; performing multi-level deviation analysis based on the real-time monitoring dataset to obtain the target deviation level, performing adjustment reasoning according to the target deviation level, and determining initial adjustment suggestions; performing effectiveness evaluation based on the initial adjustment suggestions, performing feasibility analysis based on the effectiveness evaluation results, and generating adjustment suggestions; performing full-cycle updates to the intelligent scheduling scheme according to the adjustment suggestions, generating multi-stage scheduling update instructions; automatically executing the multi-stage scheduling update instructions on key aspects of medical resources, generating the medical resource scheduling report.

[0013] In a possible implementation, the AI-based intelligent medical resource scheduling method further performs the following processes: extracting multiple key indicators from the real-time monitoring dataset, performing deviation analysis on the multiple key indicators to generate multiple indicator deviation degrees; performing a weighted comprehensive evaluation based on the multiple indicator deviation degrees, aggregating the multiple indicator deviation degrees according to the evaluation results to determine a target deviation value; defining multi-level deviation thresholds, mapping the target deviation value to the multi-level deviation thresholds for multi-level deviation analysis to obtain a target deviation level; integrating based on the target deviation level to determine multi-source deviation data, identifying data adjustment based on the multi-source deviation data, and performing collaborative reasoning based on the identification results to determine the initial adjustment suggestion.

[0014] This application also provides an AI-based intelligent medical resource scheduling system, comprising: a data acquisition unit for real-time data acquisition via a multi-source data acquisition module to construct a multi-source heterogeneous dataset of medical resources; a conflict assessment unit for real-time extrapolation and prediction based on the multi-source heterogeneous dataset of medical resources to obtain resource demand prediction results, and for conducting conflict assessment of medical resources based on the resource demand prediction results to obtain resource conflict risk values; a scheduling scheme formulation unit for triggering an intelligent scheduling engine to optimize the multi-objective allocation of medical resources based on the resource conflict risk values, and formulating an intelligent scheduling scheme; and a scheduling report generation unit for simulating the execution of the intelligent scheduling scheme to monitor key aspects of medical resources in real time, generating adjustment suggestions, updating the intelligent scheduling scheme based on the adjustment suggestions, automatically executing the scheduling, and generating a medical resource scheduling report.

[0015] This application proposes an AI-based intelligent medical resource scheduling method and system. This system involves real-time data collection and construction of a multi-source heterogeneous dataset of medical resources; real-time prediction and extrapolation of resource demand forecasts; conflict assessment of medical resources to obtain resource conflict risk values; triggering an intelligent scheduling engine to optimize multi-objective allocation of medical resources and formulate an intelligent scheduling plan; simulating the execution of the intelligent scheduling plan to monitor key aspects of medical resource allocation in real time, generating adjustment suggestions, updating the intelligent scheduling plan, and automatically executing the scheduling, generating a medical resource scheduling report. This addresses the technical problems in existing technologies that struggle to handle the complexity of heterogeneous medical data, the volatility of resource demand, and multi-objective conflicts, leading to poor efficiency and real-time performance in medical resource scheduling. It achieves the technical effect of accurate prediction and intelligent scheduling of resource demand, improving the efficiency and real-time performance of medical resource scheduling. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0017] Figure 1 This is a schematic diagram of the intelligent scheduling method for medical resources based on artificial intelligence, provided in an embodiment of this application.

[0018] Figure 2 This is a schematic diagram of the structure of an AI-based intelligent medical resource scheduling system provided in an embodiment of this application.

[0019] Figure labeling: Data acquisition unit 10, conflict assessment unit 20, scheduling scheme formulation unit 30, scheduling report generation unit 40. Detailed Implementation

[0020] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0021] This application provides an intelligent scheduling method for medical resources based on artificial intelligence, such as... Figure 1 As shown, the method includes:

[0022] Step S100: Real-time data collection is performed through the multi-source data acquisition module to construct a multi-source heterogeneous dataset of medical resources.

[0023] Step S100 further includes step S110, establishing medical data access rules, activating the information interface according to the medical data access rules for real-time data collection, and obtaining a first dataset; step S120, using IoT sensors to monitor medical devices in real time and obtaining a second dataset; step S130, performing regional analysis based on the public health platform and retrieving a third dataset; and step S140, cleaning the first dataset, the second dataset, and the third dataset, and standardizing and merging them according to the cleaning results to construct the multi-source heterogeneous dataset of medical resources.

[0024] Preferably, real-time data collection is performed through a multi-source data acquisition module. This module serves as the data engine within the medical resource scheduling system, collecting medical-related data from multiple information systems. Specifically, it establishes medical data access rules, including data format rules based on the commonly used HL7 FHIR standard in the medical field; clarifies interface protocol rules, such as API calling methods and authentication mechanisms; establishes data security and privacy rules, such as specifying access permissions; and defines data quality verification rules, such as integrity, timeliness, and validity verification. Then, real-time data collection is performed by activating information interfaces according to the medical data access rules. This includes configuring connectors that conform to the medical data access rules for each data source to collect medical data, including obtaining basic patient information, registration information, diagnostic results, examination and test results, and other patient flow data from hospital information systems, laboratory information systems, and image archiving systems. This data forms the first dataset, reflecting the real-time operational status of the hospital. Real-time monitoring of medical equipment through IoT sensors, including using sensors, RFID tags, and smart meters installed on medical equipment and beds, to obtain medical resource status information such as the usage status of medical equipment, bed occupancy, and patient flow data, obtains a second dataset to provide high-frequency, error-free real-time status data, greatly improving the accuracy and timeliness of resource status perception.

[0025] Preferably, regional analysis is performed on public health platforms such as regional health information platforms and disease control and prevention centers to obtain regional epidemiological data such as the incidence and prevalence of infectious diseases and the prevalence of key chronic diseases, as well as seasonal factor data such as temperature, humidity, and air quality index, to obtain a third dataset. This dataset provides external environmental factors that influence medical demand and can be used to predict future medical resource usage. Then, the first, second, and third datasets are cleaned, including handling missing values, correcting erroneous values, standardizing data formats, and removing duplicate records, to obtain cleaned results. These cleaned results are then standardized and fused, meaning the cleaned multi-source medical resource data are aligned and uniformly mapped based on coding standards, and finally integrated to obtain a multi-source heterogeneous dataset of medical resources.

[0026] Step S200: Real-time extrapolation and prediction are performed based on the multi-source heterogeneous dataset of medical resources to obtain resource demand prediction results. Conflict assessment of medical resources is then conducted based on the resource demand prediction results to obtain resource conflict risk values.

[0027] Step S200 further includes step S210, performing feature analysis based on a multi-source heterogeneous dataset of medical resources to obtain patient characteristics, time series characteristics, and resource status characteristics; step S220, performing multi-dimensional resource demand extrapolation based on the patient characteristics, time series characteristics, and resource status characteristics to obtain patient demand extrapolation results, equipment demand extrapolation results, and personnel demand extrapolation results; step S230, using a random forest algorithm to predict the patient demand extrapolation results, equipment demand extrapolation results, and personnel demand extrapolation results to obtain resource demand prediction results, which include predicted patient inflow, predicted equipment demand, and predicted personnel demand; step S240, performing supply and demand matching on the predicted patient inflow, predicted equipment demand, and predicted personnel demand, and performing graph theory analysis based on the matching results to determine resource dependencies; step S250, traversing the resource demand prediction results according to the resource dependencies to locate resource competition hotspots, performing conflict assessment based on the resource competition hotspots, and generating resource conflict risk values.

[0028] Preferably, feature analysis is performed on multi-source heterogeneous datasets of medical resources to extract patient features, time-series features, and resource status features. Among them, patient features are used to describe the static and dynamic attributes of the patient group, such as patient age distribution, main disease types, proportion of critical and severe cases, and region of origin. Time-series features are used to describe the regularity and pattern of data changes over time, such as peak hours for outpatient visits within a day, differences between weekdays and weekends within a week, seasonal epidemic trends, and the impact of holidays on outpatient volume. Resource status features are used to describe the current availability and load of medical resources, such as the real-time occupancy rate of beds in each department, the current number and workload of doctors and nurses on duty, the usage status and queuing situation of key equipment, and drug inventory levels.

[0029] Preferably, based on patient characteristics, time series characteristics, and resource status characteristics, a multi-dimensional resource demand projection is performed to simulate possible future scenarios. Specifically, based on the current patient inflow rate, historical data from the same period, and epidemiological data, the number, type, and severity of patients who may arrive in the next few hours or days are projected to determine the patient demand projection results. Then, combined with the projected patient demand, the demand for specific equipment is predicted to determine the equipment demand projection results. Similarly, based on the patient projection results, the number of medical staff and workload required for each department are predicted to determine the personnel demand projection results. Then, prediction models are built for patients, equipment, and personnel using random forests to transform the projection results into accurate numerical predictions. That is, the patient demand projection results, equipment demand projection results, and personnel demand projection results are input into the corresponding prediction models for prediction, thereby obtaining the predicted patient inflow, such as a prediction of 350 patients arriving in a specific time period; the predicted equipment demand, such as a prediction of 80 monitors and 25 ventilators needed during the same period; and the predicted personnel demand, such as a prediction of 45 nurses and 20 doctors needed to meet the demand.

[0030] Preferably, the predicted patient inflow, predicted equipment demand, and predicted personnel demand are matched for supply and demand. This involves comparing the predicted demand with the current status and capacity of resources, analyzing the complex relationships between resources, and then using graph theory to model all medical resource data into a resource relationship graph. In this graph, nodes represent hospital beds, equipment, medical staff, operating rooms, etc., and edges represent the dependencies between them, thereby determining resource dependencies. Next, the resource demand prediction results are traversed according to resource dependencies. That is, through resource dependency analysis, resources that are depended on by multiple processes and whose predicted demand exceeds supply capacity are identified and located as resource competition hotspots. Finally, conflict assessment is performed on resource competition hotspots, including comprehensively calculating the severity of conflict in resource competition hotspots based on the gap between demand and supply, the criticality of the resource, and the scope of impact, generating a resource conflict risk value.

[0031] Furthermore, step S230 also includes step S231, constructing a multi-dimensional feature input layer, synchronizing the patient demand projection results, the equipment demand projection results, and the personnel demand projection results to the multi-dimensional feature input layer for analysis, and extracting patient visit characteristics, equipment operation characteristics, and personnel load characteristics; step S232, based on the patient visit characteristics, equipment operation characteristics, and personnel load characteristics, dividing the patient demand projection results, equipment demand projection results, and personnel demand projection results into patient flow prediction sub-tasks, equipment demand prediction tasks, and personnel demand prediction tasks; step S233, based on the patient flow prediction sub-tasks and the patient visit characteristics, constructing a first random... Forest Model; Step S234, construct a second random forest model based on the equipment demand prediction task and the equipment operation characteristics; Step S235, construct a third random forest model based on the personnel demand prediction task and the personnel load characteristics; Step S236, perform parallel collaboration of the first random forest model, the second random forest model, and the third random forest model to generate an initial resource demand prediction result; Step S237, evaluate the credibility of the initial resource demand prediction result, update the initial resource demand prediction result in multiple dimensions according to the credibility, and integrate the predicted patient inflow, the predicted equipment demand, and the predicted personnel demand to generate the resource demand prediction result.

[0032] Preferably, a multi-dimensional feature input layer is constructed for data preprocessing and feature management. The results of patient demand projection, equipment demand projection, and personnel demand projection are simultaneously input into this multi-dimensional feature input layer for analysis, extracting more refined and effective patient visit characteristics, such as the ratio of outpatients to inpatients, the proportion of emergency and critical care cases, and the distribution of common diseases; equipment operation characteristics, such as average equipment usage time, failure rate, appointment queuing sequence, and usage preferences of different departments; and personnel workload characteristics, such as the average time for doctors of different titles to handle patients, the nurse-to-patient ratio, and workload fluctuations at shift handover times. Then, based on the patient visit characteristics, equipment operation characteristics, and personnel workload characteristics, the results of patient demand projection, equipment demand projection, and personnel demand projection are divided to determine sub-tasks for patient flow prediction, such as predicting the future number and type of patients; equipment demand prediction tasks, such as predicting the future quantity of specific equipment needed; and personnel demand prediction tasks, such as predicting the future number of medical staff and their skill requirements.

[0033] Preferably, a first random forest model, a second random forest model, and a third random forest model are constructed. The first random forest model is trained using a patient flow prediction subtask combined with patient visit characteristics as training data to predict the future number of patients. The second random forest model is then trained using an equipment demand prediction task combined with equipment operation characteristics as training data to predict the future number of equipment needed. The third random forest model is then trained using a personnel demand prediction task combined with personnel load characteristics as training data to predict the future number of medical staff needed. The first, second, and third random forest models are then run in parallel and coordinated, meaning that the three random forest models run simultaneously and merge to output prediction results, forming the initial resource demand prediction result. The credibility of the initial resource demand forecast is assessed. Specifically, the credibility of the initial resource demand forecast is determined by comprehensively calculating the variance of all decision tree forecasts in the random forest model, the quality of the input data, and the magnitude of the deviation from the historical average. The initial resource demand forecast is then updated in multiple dimensions based on the credibility. If the credibility of a forecast is very low, it is corrected to the historical average or trend value based on ensemble learning or a Bayesian model. The predicted patient inflow, predicted equipment demand, and predicted personnel demand are obtained and integrated to generate the resource demand forecast, thereby ensuring the robustness and reliability of the forecast results.

[0034] Furthermore, step S236 also includes step S2361, establishing a parallel model collaboration framework, performing parallel communication analysis on the first random forest model, the second random forest model, and the third random forest model based on the parallel model collaboration framework, and setting parallel execution environment information; step S2362, according to the parallel execution environment information, performing shared interaction coordination on the first random forest model, the second random forest model, and the third random forest model to construct and generate multiple model prediction results; step S2363, performing collaborative evaluation based on the multiple model prediction results, and when the collaborative efficiency is higher than the expected efficiency threshold, dynamically fusing the multiple model prediction results to generate the initial resource demand prediction result.

[0035] Preferably, a patient feature input layer is configured for the first random forest model to specifically handle features such as visit volume, disease type, and visit time; an equipment feature input layer is configured for the second random forest model to specifically handle equipment utilization rate, maintenance status, and efficiency indicators; and a personnel feature input layer is configured for the third random forest model to specifically handle features such as manpower allocation, skill level, and workload, thereby establishing a parallel model collaboration framework. Based on this framework, the first, second, and third random forest models are then subjected to parallel communication analysis, i.e., a lightweight message passing or memory sharing communication mechanism is established, and a parallel execution environment is set up, such as using a distributed computing framework to deploy the three models on different nodes of a computing cluster or using multi-core CPUs / GPUs for parallel computation, ensuring that the three models can run simultaneously and maintain data synchronization. Then, according to the parallel execution environment information, the first, second, and third random forest models are shared and coordinated, i.e., data exchange of the three prediction information is performed before generating the final prediction results, allowing the three prediction data to corroborate each other.

[0036] Preferably, the prediction results of multiple models are evaluated collaboratively. Specifically, the consistency of the prediction results of the three models is assessed and the execution efficiency is calculated, such as whether the parallel computation was completed within the expected time and whether there was any delay in data synchronization, thereby determining the collaborative efficiency. Then, a pre-set expected efficiency threshold is used as a quality standard based on historical data. The collaborative efficiency is compared with the expected efficiency threshold. If the collaborative efficiency is higher than the threshold, it indicates that the prediction quality is high and the result is reliable. In this case, dynamic weights are used for weighted fusion. For example, if the data quality of the device model is high, its prediction result is given a higher weight, and the initial resource demand prediction result is finally generated. If the collaborative efficiency is lower than the expected efficiency threshold, it indicates that the data is out of sync and the model results are seriously contradictory. In this case, an alarm is triggered, and the administrator is notified to intervene or initiate a degradation process, such as using the historical average value for resource demand prediction.

[0037] Furthermore, step S250 also includes step S251, mapping the resource dependency relationship to the resource demand prediction result for traversal identification, and constructing a resource dependency relationship graph; step S252, performing a depth-first traversal of the resource dependency relationship graph to extract resource temporal overlap parameters and resource spatial overlap parameters; step S253, identifying peak resource demand periods and generating a hotspot area distribution map by combining the resource temporal overlap parameters and the resource spatial overlap parameters, and performing data competition analysis based on the hotspot area distribution map to locate resource competition hotspot areas; step S254, performing multi-dimensional layer-by-layer traversal evaluation according to the resource dependency relationship based on the resource competition hotspot areas to generate temporal conflict evaluation results and spatial conflict evaluation results; step S255, performing risk aggregation calculation based on the temporal conflict evaluation results and the spatial conflict evaluation results to generate the resource conflict risk value.

[0038] Preferably, resource dependencies are mapped and overlaid onto resource demand prediction results and traversed and identified to construct a resource dependency graph. Nodes represent various resources, edges represent dependencies between resources (connecting resource nodes based on dependencies), and line thickness represents resource competition intensity. Then, a depth-first traversal is performed on the resource dependency graph, exploring the entire graph, exploring each dependency path, and backtracking to identify complex and indirect dependency chains. This allows for the extraction of resource temporal overlap parameters and resource spatial overlap parameters. The resource temporal overlap parameter indicates the degree of competition for the same resource over time between different medical activities. For example, if two different surgical plans use the same operating room at the same time, the resource temporal overlap parameter is 100%, indicating complete conflict. The resource spatial overlap parameter quantifies the degree of congestion in physical space or logical grouping. For example, multiple critically ill patients need to be placed in a space-constrained ICU simultaneously, causing spatial or logical overlap and competition.

[0039] Preferably, the peak resource demand periods are identified by combining the prediction results, i.e., determining the peak time period of future resource demand. Then, a hotspot distribution map is generated by combining resource temporal overlap parameters and resource spatial overlap parameters. Specifically, the data of resource temporal overlap parameters and resource spatial overlap parameters during the peak resource demand period are projected onto the time axis and category axis in the form of a heat map to show the heat of different time periods and different resource regions. By performing data competition analysis on the hotspot distribution map, resource competition hotspots are accurately identified and located. Then, based on the resource competition hotspots, a multi-dimensional, layer-by-layer traversal evaluation is performed according to resource dependence relationships, i.e., a layer-by-layer traversal evaluation is performed from multiple perspectives such as time, space, and resource type to analyze the direct and indirect impacts of the conflict, generating temporal conflict evaluation results and spatial conflict evaluation results. The temporal conflict evaluation results are used to quantify the severity of the conflict in time, and the spatial conflict evaluation results are used to quantify the severity of the conflict in space. Finally, risk aggregation calculation is performed based on the temporal conflict assessment results and spatial conflict assessment results. This includes inputting the temporal conflict assessment results and spatial conflict assessment results of each hotspot area, along with other factors such as the criticality of the resource and the number of patients that the conflict may affect, into the risk calculation model to perform risk aggregation calculation and generate resource conflict risk values, thereby providing accurate and actionable early warning signals.

[0040] Step S300: Based on the resource conflict risk value, trigger the intelligent scheduling engine to perform multi-objective allocation optimization of medical resources and formulate an intelligent scheduling scheme.

[0041] Step S300 further includes step S310, constructing a multi-level risk trigger threshold range, matching the multi-level risk trigger threshold range based on the resource conflict risk value, and determining the target scheduling response mode; step S320, performing scheduling impact analysis on medical resources according to the target scheduling response mode, and setting scheduling priorities; step S330, generating trigger signals according to the scheduling priorities, and activating the intelligent scheduling engine based on the trigger signals; step S340, allocating and scheduling medical resources through the intelligent scheduling engine, generating initial resource allocation information, performing multi-objective optimization based on the initial resource allocation information, and determining multiple optimization objectives; step S350, performing scheduling verification based on the multiple optimization objectives, and formulating the intelligent scheduling scheme based on the verification results.

[0042] Preferably, multi-level risk trigger threshold ranges of different levels are pre-set. For example, the green range is low risk and does not require active scheduling; the yellow range is medium risk and triggers an early warning mode to formulate an optimized scheduling plan; and the red range is high risk and triggers an emergency mode to formulate an aggressive scheduling plan. Then, the multi-level risk trigger threshold ranges are matched based on the resource conflict risk value. According to the range in which the resource conflict risk value falls, the system automatically switches to the corresponding target scheduling response mode. Next, the scheduling impact analysis of medical resources is performed according to the target scheduling response mode, that is, the impact chain that different scheduling actions may produce is simulated. Based on the scheduling impact analysis results, the conflicts of the resources to be resolved are sorted. The scheduling priority is set according to clinical urgency, scope of impact and resolution efficiency. For example, equipment resources involving the lives of critically ill patients are prioritized over ordinary resources; equipment resources that can solve more patients' problems are prioritized over resources with a smaller impact; and scheduling tasks that are easy to adjust and have short processing time are prioritized over complex tasks.

[0043] Preferably, a trigger signal is generated based on the scheduling priority and sent to the intelligent scheduling engine for activation, initiating intelligent resource scheduling. The intelligent scheduling engine allocates and schedules medical resources, generating initial resource allocation information as a starting point for multi-objective optimization. This involves simultaneously weighing multiple objectives and determining the optimal balance point. These optimization objectives may include maximizing efficiency, ensuring fairness in task scheduling, minimizing costs, and maintaining scheduling stability. Finally, the intelligent engine verifies the scheduling based on these multiple optimization objectives. This includes iteratively verifying the initial resource allocation information and simulating its scheduling effects to determine if scheduling balance can be achieved. Ultimately, a detailed and executable optimal scheduling scheme is output as the intelligent scheduling scheme, specifying information such as scheduling time, scheduling resources, allocation information, and duration. This achieves an automated closed loop of risk warning, intelligent decision-making, and scheduling scheme generation.

[0044] Step S400: Simulate the execution of the intelligent scheduling scheme to monitor key aspects of medical resources in real time, generate adjustment suggestions, update the intelligent scheduling scheme according to the adjustment suggestions, and automatically schedule and execute the scheme to generate a medical resource scheduling report.

[0045] Step S400 further includes step S410, constructing digital twin environment information, loading the intelligent scheduling scheme based on the digital twin environment information to simulate and execute key links of medical resources, and generating a simulation scheduling parameter set; step S420, performing multi-dimensional operational status perception of key links of medical resources based on the simulation scheduling parameter set, and obtaining a real-time monitoring dataset; step S430, performing multi-level deviation analysis based on the real-time monitoring dataset, obtaining the target deviation level, and performing adjustment reasoning according to the target deviation level to determine the initial adjustment suggestion; step S440, performing an effectiveness evaluation based on the initial adjustment suggestion, performing a feasibility analysis based on the effectiveness evaluation result, and generating adjustment suggestions; step S450, performing a full-cycle update of the intelligent scheduling scheme according to the adjustment suggestions, and generating a multi-stage scheduling update instruction; step S460, automatically executing the multi-stage scheduling update instruction on key links of medical resources, and generating the medical resource scheduling report.

[0046] Preferably, a digital twin environment is constructed, incorporating real-world medical environment information such as hospitals, departments, equipment, and patient flow. This integrates physical spaces like ward layouts, logical rules like treatment processes, and real-time dynamic patient information. The intelligent scheduling scheme is then loaded into the digital twin environment, and key aspects of medical resources are simulated. This involves rapidly simulating the execution of the intelligent scheduling scheme over the next few hours or days using a simulation clock, generating a set of operational data, such as average patient waiting time, peak CT scanner utilization, nurse workload curves, and bed turnover rates. This data forms a simulation scheduling parameter set, reflecting the effectiveness of the intelligent scheduling scheme. Based on this simulation scheduling parameter set, the operational status of key aspects of medical resources is then comprehensively monitored, acquiring the operational status of each key aspect as a real-time monitoring dataset.

[0047] Preferably, multi-level deviation analysis is performed on the real-time monitoring dataset, which involves comparing the real-time monitoring dataset with the expected ideal target to obtain the deviation and classifying it into different levels according to the severity of the deviation, such as slight deviation, moderate deviation, and severe deviation, to determine the target deviation level. Then, adjustment reasoning is performed according to the target deviation level, including analyzing the causes of the deviation and automatically reasoning corrective measures. For example, if it is found that the delay in operating room setup is a severe deviation caused by insufficient staff in the preoperative preparation room, an initial adjustment suggestion is determined, such as adding two nurses to the preoperative preparation room. Then, an effectiveness evaluation is performed on the initial adjustment suggestion, which involves quickly simulating whether the initial adjustment suggestion can effectively correct the deviation, and evaluating the feasibility of the initial adjustment suggestion based on the effectiveness evaluation results, and finally generating a confirmed adjustment suggestion. Next, the intelligent scheduling scheme is updated throughout its entire lifecycle according to the adjustment suggestions. This involves integrating the adjustment suggestions into the intelligent scheduling scheme to generate a dynamic updated version that covers the entire scheduling cycle and includes different instructions triggered at different stages, thereby generating multi-stage scheduling update instructions. Finally, multi-stage scheduling update instructions are automatically executed on key links of medical resources. The simulation results, deviation analysis, adjustment logic, and final execution scheme of the entire process are integrated to generate a medical resource scheduling report, thereby improving the scientific nature and safety of medical resource scheduling decisions.

[0048] Furthermore, step S430 also includes step S431, extracting multiple key indicators based on the real-time monitoring dataset, performing deviation analysis on the multiple key indicators, and generating multiple indicator deviation degrees; step S432, performing a weighted comprehensive evaluation based on the multiple indicator deviation degrees, aggregating the multiple indicator deviation degrees according to the evaluation results, and determining the target deviation value; step S433, defining multi-level deviation thresholds, mapping the target deviation value to the multi-level deviation thresholds for multi-level deviation analysis, and obtaining the target deviation level; step S434, integrating based on the target deviation level to determine multi-source deviation data, identifying data adjustment based on the multi-source deviation data, performing collaborative reasoning based on the identification results, and determining the initial adjustment suggestion.

[0049] Preferably, multiple key indicators reflecting operational status are extracted from the real-time monitoring dataset, such as patient waiting time, bed utilization rate, medical staff workload rate, equipment utilization rate, and operating room turnaround delay time. Deviation analysis is performed between these key indicators and the expected target values. The degree of deviation is determined by (key indicator - expected target value) / expected target value, generating multiple indicator deviation scores. Then, based on the importance of the indicators and historical data, different weights are assigned to each indicator. A weighted comprehensive evaluation is performed based on the multiple indicator deviation scores to generate an evaluation result. Finally, all indicator deviation scores are multiplied by their weights and aggregated to obtain the target deviation value, representing the overall severity of deviation from the expected target. Then, multi-level deviation thresholds are preset, i.e., several threshold ranges are preset. For example, the target deviation value for minor deviation is 0-20%, the target deviation value for moderate deviation is 21%-50%, and the target deviation value for severe deviation is above 51%. The target deviation value is mapped to the multi-level deviation thresholds to perform multi-level deviation analysis and determine the current target deviation level. Then, based on the target deviation level, multi-source deviation data is integrated to determine the data adjustment labels, i.e., the specific resources and scheduling links where different types of deviations occur are accurately identified to obtain the labeling results. Finally, collaborative reasoning is performed based on the labeling results, i.e., intelligent decision-making is made on the labeling results to analyze and determine the initial adjustment suggestions. For example, if the target deviation level is moderate and the cause is long CT waiting time due to competition between appointments and emergency services, the initial adjustment suggestion may be to reserve specific CT scan time slots for emergency patients and adjust some non-urgent appointment patients to the afternoon off-peak hours, thereby achieving accurate prediction and intelligent scheduling of resource demand and improving the efficiency and real-time performance of medical resource scheduling.

[0050] In the above text, refer to Figure 1 This paper describes in detail an artificial intelligence-based intelligent scheduling method for medical resources according to embodiments of the present invention. Next, reference will be made to... Figure 2 This invention describes an artificial intelligence-based intelligent medical resource scheduling system according to an embodiment of the present invention.

[0051] The AI-based intelligent medical resource scheduling system according to embodiments of the present invention addresses the technical problems in existing technologies, such as the difficulty in handling the complexity of heterogeneous medical data, the volatility of resource demand, and multi-objective conflicts, leading to poor efficiency and real-time performance in medical resource scheduling. It achieves the technical effect of accurate prediction and intelligent scheduling of resource demand, thereby improving the efficiency and real-time performance of medical resource scheduling. Figure 2 As shown, the AI-based intelligent medical resource scheduling system includes: a data acquisition unit 10, a conflict assessment unit 20, a scheduling scheme formulation unit 30, and a scheduling report generation unit 40.

[0052] The data acquisition unit 10 is used to collect data in real time through a multi-source data acquisition module to construct a multi-source heterogeneous dataset of medical resources; the conflict assessment unit 20 is used to perform real-time inference and prediction based on the multi-source heterogeneous dataset of medical resources to obtain resource demand prediction results, and to conduct conflict assessment of medical resources based on the resource demand prediction results to obtain resource conflict risk values; the scheduling scheme formulation unit 30 is used to trigger the intelligent scheduling engine to optimize the multi-objective allocation of medical resources based on the resource conflict risk values ​​and formulate an intelligent scheduling scheme; the scheduling report generation unit 40 is used to simulate the execution of the intelligent scheduling scheme to monitor key links of medical resources in real time, generate adjustment suggestions, update the intelligent scheduling scheme according to the adjustment suggestions, and automatically schedule and execute the scheme to generate a medical resource scheduling report.

[0053] The specific configuration of the data acquisition unit 10 will be described in detail below. The data acquisition unit 10 further includes: establishing medical data access rules; activating the information interface according to the medical data access rules for real-time data acquisition to obtain a first dataset; monitoring medical devices in real-time using IoT sensors to obtain a second dataset; performing regional analysis based on a public health platform to retrieve a third dataset; cleaning the first, second, and third datasets; standardizing and fusing the data according to the cleaning results to construct the multi-source heterogeneous dataset of medical resources.

[0054] The specific configuration of the conflict assessment unit 20 will be described in detail below. The conflict assessment unit 20 further includes: performing feature analysis based on a multi-source heterogeneous dataset of medical resources to obtain patient characteristics, time-series characteristics, and resource status characteristics; performing multi-dimensional resource demand extrapolation based on the patient characteristics, time-series characteristics, and resource status characteristics to obtain patient demand extrapolation results, equipment demand extrapolation results, and personnel demand extrapolation results; using a random forest algorithm to predict the patient demand extrapolation results, equipment demand extrapolation results, and personnel demand extrapolation results to obtain resource demand prediction results, which include predicted patient inflow, predicted equipment demand, and predicted personnel demand; performing supply-demand matching on the predicted patient inflow, predicted equipment demand, and predicted personnel demand, and performing graph theory analysis based on the matching results to determine resource dependencies; traversing the resource demand prediction results according to the resource dependencies to locate resource competition hotspot areas, and performing conflict assessment based on the resource competition hotspot areas to generate resource conflict risk values.

[0055] The specific configuration of the conflict assessment unit 20 will be described in detail below. The conflict assessment unit 20 further includes: constructing a multi-dimensional feature input layer, synchronizing the patient demand projection results, equipment demand projection results, and personnel demand projection results to the multi-dimensional feature input layer for analysis, and extracting patient visit characteristics, equipment operation characteristics, and personnel load characteristics; based on the patient visit characteristics, equipment operation characteristics, and personnel load characteristics, dividing the patient demand projection results, equipment demand projection results, and personnel demand projection results into patient flow prediction sub-tasks, equipment demand prediction tasks, and personnel demand prediction tasks; and constructing a system based on the patient flow prediction sub-tasks combined with the patient visit characteristics. A first random forest model is used; a second random forest model is constructed based on the equipment demand prediction task and the equipment operation characteristics; a third random forest model is constructed based on the personnel demand prediction task and the personnel load characteristics; the first, second, and third random forest models are performed in parallel and collaboratively to generate initial resource demand prediction results; the reliability of the initial resource demand prediction results is evaluated, and the initial resource demand prediction results are updated in multiple dimensions according to the reliability, and the predicted patient inflow, predicted equipment demand, and predicted personnel demand are integrated to generate the resource demand prediction results.

[0056] The specific configuration of the conflict assessment unit 20 will be described in detail below. The conflict assessment unit 20 further includes: establishing a parallel model collaboration framework; performing parallel communication analysis on the first random forest model, the second random forest model, and the third random forest model based on the parallel model collaboration framework; setting parallel execution environment information; coordinating and sharing the first random forest model, the second random forest model, and the third random forest model according to the parallel execution environment information to generate multiple model prediction results; and performing collaborative evaluation based on the multiple model prediction results. When the collaborative efficiency is higher than the expected efficiency threshold, the multiple model prediction results are dynamically fused to generate the initial resource requirement prediction result.

[0057] The specific configuration of the conflict assessment unit 20 will be described in detail below. The conflict assessment unit 20 further includes: mapping the resource dependency relationships to the resource demand prediction results for traversal identification, constructing a resource dependency relationship graph; performing a depth-first traversal of the resource dependency relationship graph to extract resource temporal overlap parameters and resource spatial overlap parameters; identifying peak resource demand periods and generating a hotspot area distribution map based on the resource temporal overlap parameters and the resource spatial overlap parameters; performing data competition analysis based on the hotspot area distribution map to locate resource competition hotspot areas; performing multi-dimensional, layer-by-layer traversal assessment according to the resource dependency relationships based on the resource competition hotspot areas to generate temporal conflict assessment results and spatial conflict assessment results; and performing risk aggregation calculation based on the temporal conflict assessment results and spatial conflict assessment results to generate the resource conflict risk value.

[0058] The specific configuration of the scheduling scheme formulation unit 30 will be described in detail below. The scheduling scheme formulation unit 30 further includes: constructing a multi-level risk triggering threshold range; matching the multi-level risk triggering threshold range based on the resource conflict risk value to determine the target scheduling response mode; performing a scheduling impact analysis on medical resources according to the target scheduling response mode; and setting scheduling priorities.

[0059] A trigger signal is generated based on the scheduling priority, and the intelligent scheduling engine is activated based on the trigger signal. Medical resources are allocated and scheduled through the intelligent scheduling engine to generate initial resource allocation information. Multi-objective optimization is performed based on the initial resource allocation information to determine multiple optimization objectives. Scheduling verification is performed based on the multiple optimization objectives, and the intelligent scheduling scheme is formulated based on the verification results.

[0060] The specific configuration of the scheduling report generation unit 40 will be described in detail below. The scheduling report generation unit 40 further includes: constructing digital twin environment information; loading the intelligent scheduling scheme based on the digital twin environment information to simulate and execute key aspects of medical resources, generating a simulation scheduling parameter set; performing multi-dimensional operational status perception on key aspects of medical resources based on the simulation scheduling parameter set, obtaining a real-time monitoring dataset; performing multi-level deviation analysis based on the real-time monitoring dataset to obtain the target deviation level; performing adjustment reasoning according to the target deviation level to determine initial adjustment suggestions; conducting effectiveness evaluation based on the initial adjustment suggestions; performing feasibility analysis based on the effectiveness evaluation results to generate adjustment suggestions; performing full-cycle updates to the intelligent scheduling scheme according to the adjustment suggestions, generating multi-stage scheduling update instructions; and automatically executing the multi-stage scheduling update instructions on key aspects of medical resources to generate the medical resource scheduling report.

[0061] The specific configuration of the scheduling report generation unit 40 will be described in detail below. The scheduling report generation unit 40 further includes: extracting multiple key indicators based on the real-time monitoring dataset; performing deviation analysis on the multiple key indicators to generate multiple indicator deviation degrees; performing a weighted comprehensive evaluation based on the multiple indicator deviation degrees; aggregating the multiple indicator deviation degrees according to the evaluation results to determine a target deviation value; defining multi-level deviation thresholds; mapping the target deviation value to the multi-level deviation thresholds for multi-level deviation analysis to obtain a target deviation level; integrating based on the target deviation level to determine multi-source deviation data; identifying data adjustment based on the multi-source deviation data; and performing collaborative reasoning based on the identification results to determine the initial adjustment suggestion.

[0062] The AI-based intelligent medical resource scheduling system provided in this invention can execute the AI-based intelligent medical resource scheduling method provided in any embodiment of this invention, and has the corresponding functional modules and beneficial effects of the method execution.

[0063] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. An artificial intelligence-based medical resource intelligent scheduling method, characterized in that, The method includes: Real-time data collection is performed using a multi-source data acquisition module to construct a multi-source heterogeneous dataset of medical resources. Real-time extrapolation and prediction are performed based on the multi-source heterogeneous dataset of medical resources to obtain resource demand prediction results. Based on the resource demand prediction results, a conflict assessment of medical resources is conducted to obtain resource conflict risk values. Based on the resource conflict risk value, the intelligent scheduling engine is triggered to perform multi-objective allocation optimization of medical resources and formulate an intelligent scheduling scheme. The intelligent scheduling scheme is simulated to monitor key aspects of medical resources in real time, generate adjustment suggestions, update the intelligent scheduling scheme according to the adjustment suggestions, and automatically schedule and execute the scheme to generate a medical resource scheduling report. Based on the aforementioned multi-source heterogeneous dataset of medical resources, real-time extrapolation and prediction are performed to obtain resource demand prediction results. Then, based on these resource demand prediction results, a conflict assessment of medical resources is conducted to obtain a resource conflict risk value. The method includes: Feature analysis was performed on multi-source heterogeneous datasets of medical resources to obtain patient characteristics, time series characteristics, and resource status characteristics. Based on the patient characteristics, the time series characteristics, and the resource status characteristics, multi-dimensional resource demand extrapolation is performed to obtain patient demand extrapolation results, equipment demand extrapolation results, and personnel demand extrapolation results. The random forest algorithm is used to predict the patient demand projection results, the equipment demand projection results, and the personnel demand projection results to obtain resource demand prediction results, which include the predicted patient inflow, the predicted equipment demand, and the predicted personnel demand. The predicted patient inflow, the predicted equipment demand, and the predicted personnel demand are matched for supply and demand. Based on the matching results, graph theory analysis is performed to determine resource dependencies. Based on the resource dependency relationship, the resource demand forecast results are traversed to locate resource competition hotspots. Conflict assessment is performed based on the resource competition hotspots to generate resource conflict risk values. The random forest algorithm is used to predict the patient demand projection results, the equipment demand projection results, and the personnel demand projection results to obtain resource demand prediction results. The method includes: A multi-dimensional feature input layer is constructed, and the patient demand projection results, equipment demand projection results, and personnel demand projection results are synchronized to the multi-dimensional feature input layer for analysis to extract patient visit characteristics, equipment operation characteristics, and personnel load characteristics. Based on the patient visit characteristics, the equipment operation characteristics, and the personnel load characteristics, the patient demand projection results, the equipment demand projection results, and the personnel demand projection results are divided into patient flow prediction sub-tasks, equipment demand prediction tasks, and personnel demand prediction tasks. A first random forest model is constructed based on the patient flow prediction subtask and the patient visit characteristics. A second random forest model is constructed based on the equipment demand prediction task and the equipment operation characteristics. A third random forest model is constructed based on the personnel demand prediction task and the personnel load characteristics. The first random forest model, the second random forest model, and the third random forest model are used in parallel and coordinated to generate initial resource demand prediction results. The initial resource demand forecast results are evaluated for reliability. Based on the reliability, the initial resource demand forecast results are updated in multiple dimensions. The predicted patient inflow, the predicted equipment demand, and the predicted personnel demand are integrated to generate the resource demand forecast results. The method involves traversing the resource demand forecast results according to the resource dependency relationships to locate resource competition hotspots, conducting conflict assessments based on these hotspots, and generating resource conflict risk values. The resource dependencies are mapped to the resource demand prediction results and traversed and identified to construct a resource dependency graph. A depth-first traversal is performed on the resource dependency graph to extract resource temporal overlap parameters and resource spatial overlap parameters; The peak periods of resource demand are identified, and a hotspot distribution map is generated by combining the resource time overlap parameter and the resource spatial overlap parameter. Based on the hotspot distribution map, data competition analysis is performed to locate resource competition hotspot areas. Based on the resource competition hotspots, a multi-dimensional, layer-by-layer evaluation is performed according to the resource dependencies to generate time conflict evaluation results and spatial conflict evaluation results. Based on the time conflict assessment results and the spatial conflict assessment results, risk aggregation calculations are performed to generate the resource conflict risk value. 2.The AI-based medical resource intelligent scheduling method of claim 1, wherein, Real-time data acquisition using a multi-source data acquisition module is employed to construct a multi-source heterogeneous dataset of medical resources. The methods include: Establish medical data access rules, activate the information interface according to the medical data access rules to collect data in real time, and obtain the first dataset. A second dataset is obtained by real-time monitoring of medical devices using IoT sensors; Regional analysis was conducted based on a public health platform, and a third dataset was retrieved. The first dataset, the second dataset, and the third dataset are cleaned, and then standardized and fused based on the cleaning results to construct the multi-source heterogeneous dataset of medical resources.

3. The intelligent medical resource scheduling method based on artificial intelligence as described in claim 1, characterized in that, The first, second, and third random forest models are used in parallel and collaboratively to generate initial resource demand prediction results. The method includes: A parallel model collaboration framework is established, and the first random forest model, the second random forest model, and the third random forest model are analyzed in parallel communication based on the parallel model collaboration framework. The parallel execution environment information is set. Based on the parallel execution environment information, the first random forest model, the second random forest model, and the third random forest model are shared and coordinated to generate multiple model prediction results; Based on the prediction results of the multiple models, a collaborative evaluation is performed. When the collaborative efficiency is higher than the expected efficiency threshold, the prediction results of the multiple models are dynamically fused to generate the initial resource demand prediction result.

4. The intelligent medical resource scheduling method based on artificial intelligence as described in claim 1, characterized in that, Based on the resource conflict risk value, the intelligent scheduling engine is triggered to perform multi-objective allocation optimization of medical resources, and an intelligent scheduling scheme is formulated. The method includes: Construct a multi-level risk trigger threshold range, and match the multi-level risk trigger threshold range based on the resource conflict risk value to determine the target scheduling response mode; Based on the target scheduling response mode, an impact analysis of medical resource scheduling is performed, and scheduling priorities are set. A trigger signal is generated based on the scheduling priority, and the intelligent scheduling engine is activated based on the trigger signal. The intelligent scheduling engine allocates and schedules medical resources, generates initial resource allocation information, and performs multi-objective optimization based on the initial resource allocation information to determine multiple optimization objectives. The scheduling is verified based on the multiple optimization objectives, and the intelligent scheduling scheme is formulated based on the verification results. 5.The AI-based medical resource intelligent scheduling method of claim 1, wherein, The method includes simulating the execution of the intelligent scheduling scheme to monitor key aspects of medical resources in real time, generating adjustment suggestions, updating the intelligent scheduling scheme according to the adjustment suggestions, automatically scheduling and executing the scheme, and generating a medical resource scheduling report. Construct digital twin environment information, load the intelligent scheduling scheme based on the digital twin environment information to simulate the execution of key links of medical resources, and generate a simulation scheduling parameter set; Based on the simulation scheduling parameter set, the operational status of key links in medical resources is perceived from multiple perspectives to obtain a real-time monitoring dataset. Based on the real-time monitoring dataset, multi-level deviation analysis is performed to obtain the target deviation level. Adjustment reasoning is then performed according to the target deviation level to determine the initial adjustment recommendations. An effectiveness assessment is conducted based on the initial adjustment recommendations, and a feasibility analysis is performed based on the effectiveness assessment results to generate adjustment recommendations. The intelligent scheduling scheme is updated throughout its entire lifecycle according to the adjustment recommendations, generating multi-stage scheduling update instructions. The multi-stage scheduling update instructions are automatically executed for key aspects of medical resources, generating the medical resource scheduling report. 6.The AI-based medical resource intelligent scheduling method of claim 5, wherein, Based on the real-time monitoring dataset, multi-level deviation analysis is performed to obtain the target deviation level. Adjustment inference is then conducted according to the target deviation level to determine initial adjustment recommendations. The method includes: Based on the real-time monitoring dataset, multiple key indicators are extracted, and deviation analysis is performed on the multiple key indicators to generate multiple indicator deviation degrees. A weighted comprehensive evaluation is performed based on the deviation of the multiple indicators, and the deviation of the multiple indicators is aggregated according to the evaluation results to determine the target deviation value; Define multi-level deviation thresholds, map the target deviation value to the multi-level deviation thresholds to perform multi-level deviation analysis, and obtain the target deviation level; Based on the target deviation level, multi-source deviation data is identified through integration. Data adjustment labels are then applied to the multi-source deviation data. Based on the labeling results, collaborative reasoning is performed to determine the initial adjustment recommendations.

7. The intelligent medical resource scheduling system based on artificial intelligence, characterized in that, The system is used to implement the AI-based intelligent medical resource scheduling method according to any one of claims 1 to 6, and the system comprises: The data acquisition unit is used to collect data in real time through the multi-source data acquisition module to build a multi-source heterogeneous dataset of medical resources. The conflict assessment unit is used to perform real-time extrapolation and prediction based on the multi-source heterogeneous dataset of medical resources, obtain resource demand prediction results, and conduct conflict assessment of medical resources based on the resource demand prediction results to obtain resource conflict risk values. The scheduling scheme formulation unit is used to trigger the intelligent scheduling engine to perform multi-objective allocation optimization of medical resources based on the resource conflict risk value, and to formulate an intelligent scheduling scheme. The scheduling report generation unit is used to simulate the execution of the intelligent scheduling scheme to monitor key aspects of medical resources in real time, generate adjustment suggestions, update the intelligent scheduling scheme according to the adjustment suggestions, and automatically schedule and execute the scheme to generate a medical resource scheduling report.

Images