A method and system for managing a data analysis platform

Abstract

The present invention discloses a method and system for managing a data analysis platform, belonging to the technical field of computer data processing. The method includes obtaining multi-source heterogeneous data from a medical information system; establishing a cross-system data association model, performing dynamic feature extraction and weight assignment on the multi-source heterogeneous data, and outputting a resource supply-demand relationship graph; analyzing the resource supply-demand relationship graph to identify business conflicts in medical resource scheduling; generating a conflict solution for the business conflicts based on a multi-objective optimization algorithm and generating an execution instruction; sending the execution instruction to a target medical system, verifying the execution result of the execution instruction and updating system parameters. The present invention adopts a cross-system dynamic modeling and multi-objective optimization strategy, combines a closed-loop verification mechanism and a multi-source heterogeneous data fusion technology, can realize intelligent prediction, precise scheduling and self-learning optimization of medical resource conflicts, ensure traceable execution of the whole process of high-value consumables, and comprehensively improve resource utilization efficiency and emergency response capabilities.

Description

Technical Field

[0001] The present invention relates to the technical field of computer data processing, and in particular to a method and system for managing a data analysis platform. Background Art

[0002] Currently, the scheduling of medical resources faces the problem of difficult integration of multi-source heterogeneous data, and the phenomenon of data islands is common due to the independent operation of each system within the hospital. Patient diagnosis and treatment, consumable inventory, and equipment operation data are scattered on different platforms and the formats are non-standardized, making it difficult to achieve dynamic perception and real-time response to business requirements. Resource conflicts mainly rely on manual experience to identify, and there are defects such as lag in supply-demand imbalance early warning and inaccurate determination of the priority of urgent tasks. Especially in large-scale emergencies, there is a risk of misallocation of key resources.

[0003] Traditional solutions focus on local optimization of a single system, using periodic data batch processing and static rule engines for resource matching. For example, limited data interoperability is achieved through basic database association technology, or resource warning lines are set using fixed thresholds. Some systems introduce linear programming models to optimize path selection, but fail to effectively integrate spatio-temporal dynamic parameters and multi-dimensional constraint conditions, resulting in a large deviation between decision-making and actual execution. Such methods lack the ability of adaptive tuning when facing complex and changeable medical scenarios and cannot support cross-hospital collaborative requirements.

[0004] Existing technologies have problems such as single modeling dimension and lack of real-time feedback when processing cross-system data. The static weight allocation mechanism is difficult to capture the dynamic changes of the business and easily causes rigid algorithm decision-making. At the same time, there is no closed-loop verification mechanism in the process of medical resource scheduling, and the execution effect is difficult to quantitatively evaluate, and historical experience is not fully transformed into the basis for model optimization. In addition, the safety verification methods for the scheduling of high-value consumables are fragmented, with shortcomings such as difficult traceability of operation risks and low emergency response efficiency, and it is difficult to meet the requirements of modern medical refined management. Summary of the Invention

[0005] To solve the above problems, the present invention provides a method and system for managing a data analysis platform, which adopts a cross-system dynamic modeling and multi-objective optimization strategy, combines a closed-loop verification mechanism and multi-source heterogeneous data fusion technology, and can realize intelligent prediction, precise scheduling and self-learning optimization of medical resource conflicts, ensure traceable execution of the entire process of high-value consumables, and comprehensively improve resource utilization efficiency and emergency response ability.

[0006] The above object can be achieved by the following solutions:

[0007] A method for managing a data analysis platform, comprising: obtaining multi-source heterogeneous data from a medical information system, the data including patient diagnosis and treatment data, medical consumable data, and equipment operation data; establishing a cross-system data association model, performing dynamic feature extraction and weight assignment on the multi-source heterogeneous data, and outputting a resource supply and demand relationship map; analyzing the resource supply and demand relationship map to identify business conflicts in medical resource scheduling; generating a conflict solution for the business conflicts based on a multi-objective optimization algorithm, and generating an execution instruction; sending the execution instruction to a target medical system, the instruction including a resource scheduling path and a time limit constraint condition; verifying the execution result of the execution instruction and updating system parameters.

[0008] Optionally, the obtaining multi-source heterogeneous data from a medical information system includes: connecting an electronic medical record system and a material management system through a standardized protocol conversion module, and outputting a structured interface data stream; based on the structured interface data stream, identifying treatment type keywords and consumable identification code segments in medical order texts, and converting unstructured data into medical resource labels in a unified coding format; according to the attribute characteristics of the medical resource labels, adjusting the data collection mode at a preset time period, the data collection mode including: a real-time data processing mode, grabbing key fields at a second-level frequency; a batch update mode, integrating all-day data to generate an incremental update package.

[0009] Optionally, the establishing a cross-system data association model, performing dynamic feature extraction and weight assignment on the multi-source heterogeneous data, and outputting a resource supply and demand relationship map includes: constructing a three-dimensional feature coordinate system, the dimensions of the three-dimensional feature coordinate system including business type, data format, and time limit level, and generating an initial feature vector set; based on the initial feature vector set, establishing a weight dynamic adjustment mechanism, when detecting that the application frequency of the same type of consumables exceeds a preset frequency threshold, calculating a weight increase parameter through a sliding window algorithm, and adjusting the feature vectors in the initial feature vector set; inputting the adjusted feature vectors into a preset graph database, constructing a multi-level association relationship between nodes, and outputting a resource supply and demand relationship map.

[0010] Optionally, the identifying business conflicts in medical resource scheduling includes: scanning multi-system data records of medical resources, extracting a three-value difference set of inventory quantity, allocated quantity, and in-transit quantity; based on the difference set, analyzing the spatio-temporal matching degree between operating room reservation data and consumable inventory data, and positioning conflict coordinates; according to the node parameters of the resource supply and demand relationship map, calculating a conflict level and generating a conflict warning signal.

[0011] Optionally, the method for generating a conflict resolution solution for the service conflict based on the multi-objective optimization algorithm includes: receiving the conflict warning signal, and configuring the time limit priority strategy for clinical treatment decision-making and the cost priority strategy for administrative material decision-making; generating a multi-path scheduling plan according to the resource supply-demand relationship map, including in-hospital area transfer, emergency inventory call, and cross-hospital area allocation; predicting and calculating the time deviation rate, cost deviation value, and feasibility of each plan according to the preset historical scheduling plan execution database, and generating a three-dimensional decision matrix; and selecting a scheduling plan by using the three-dimensional decision matrix.

[0012] Optionally, the step of sending the execution instruction to the target medical system includes: disassembling the selected scheduling plan into standardized instruction units, where each instruction unit includes a target system identifier, an operation content code, and a time limit tag; arranging electronic tag verification points on the consumable transfer path, collecting and feeding back the consumable coordinates and timestamps; and activating a dual electronic signature confirmation mechanism when a high-value consumable scheduling instruction is detected.

[0013] Optionally, the step of verifying the execution result of the execution instruction and updating the system parameters includes: based on the consumable coordinates and timestamps, collecting the consumable position data and time limit data during the execution process of the instruction in real time; calculating the time deviation degree and space deviation degree between the actual execution path and the optimized path based on the consumable position data and time limit data, generating a comprehensive deviation degree, and generating an alarm level based on the comprehensive deviation degree; triggering an immediate correction instruction for the weight increase parameter based on the alarm level, and adjusting the weight increase parameter by using the comprehensive deviation degree.

[0014] Optionally, the step of verifying the execution result of the execution instruction and updating the system parameters further includes: integrating the execution results to construct an experience knowledge base, where the experience knowledge base includes a successful case template and an exception handling report; periodically adjusting the system parameters according to the historical data of the experience knowledge base; establishing a department execution energy efficiency score, and updating the weight of the corresponding node of the department in the resource supply-demand relationship map according to the score result.

[0015] Optionally, the method further includes: performing an exception judgment according to the space deviation degree, and when an exception is detected, loading the successful case template and switching to a degraded mode; calling the path data with the minimum average value of the space deviation degree within a preset time window according to the successful case template to generate an emergency scheduling plan; reducing the data collection frequency, simplifying the label conversion, and allocating emergency resources according to the score result in the degraded mode.

[0016] Based on the same inventive concept, the present invention also provides a data analysis platform management system, which includes: a data collection module for obtaining multi-source heterogeneous data from a medical information system, the data including patient diagnosis and treatment data, medical consumable data, and equipment operation data; a supply-demand relationship establishment module for establishing a cross-system data association model, performing dynamic feature extraction and weight allocation on the multi-source heterogeneous data, and outputting a resource supply-demand relationship map; a business conflict identification module for analyzing the resource supply-demand relationship map and identifying business conflicts in medical resource scheduling; a decision generation module for generating a conflict solution for the business conflict based on a multi-objective optimization algorithm and generating an execution instruction; an instruction execution module for sending the execution instruction to a target medical system, the instruction including a resource scheduling path and a time limit constraint condition; and a closed-loop verification module for verifying the execution result of the execution instruction and updating system parameters.

[0017] Compared with the prior art, the present invention has the following advantages:

[0018] 1. By standardizing the conversion and dynamically modeling the features of heterogeneous data such as electronic medical records, equipment operation, and consumable inventory, the present invention constructs an accurate resource supply-demand relationship map, breaks through the limitations of data islands in traditional medical systems, and significantly improves the cross-departmental collaboration efficiency; it can automatically identify key resource bottlenecks and quantify the conflict level, providing data support from a global perspective for complex scheduling scenarios.

[0019] 2. Based on the spatio-temporal matching degree analysis and multi-objective optimization algorithm, the present invention achieves a balance between time limit priority and cost control, generating an optimal path that takes into account both emergency treatment needs and operation efficiency; compared with traditional manual scheduling, it can quickly locate the conflict nodes of operating room scheduling and inventory dislocation, and adaptively adjust the decision-making strategy through a historical experience database, avoiding the delay risk caused by misallocation of medical resources.

[0020] 3. The present invention uses electronic tag verification points to track the instruction execution path in real time, combines the spatio-temporal deviation degree to dynamically correct the model weight parameters, and establishes a full-chain monitoring of the transportation process; continuously accumulates successful cases and abnormal handling experiences, realizes the intelligent linkage between department energy efficiency scores and resource allocation weights, and promotes the continuous optimization and upgrading of medical resource allocation strategies.

[0021] 4. By adopting double electronic signature and blockchain evidence storage technology, it ensures that the operation of critical medical supplies scheduling is traceable and the permissions are verifiable; the standardized instruction unit and the downgrade mode switching mechanism effectively prevent the risk of abuse of medical resources while ensuring the continuity of core business, providing a reliable technical guarantee for refined management.

[0022] Other features and advantages of the present invention will be set forth in the following description, and in part will be obvious from the description, or may be learned by practice of the present invention. The objectives and other advantages of the present invention can be realized and attained by the structure pointed out in the description, claims and drawings. Description of the Drawings

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0024] Figure 1 It is a schematic structural diagram of a data analysis platform management method according to an embodiment of the present invention.

[0025] Figure 2 It is an initial feature vector distribution diagram according to an embodiment of the present invention.

[0026] Figure 3 It is a schematic diagram of a weight increase adjustment curve according to an embodiment of the present invention.

[0027] Figure 4 It is a schematic diagram of the spatial structure of a three-dimensional decision matrix according to an embodiment of the present invention.

[0028] Figure 5 It is a bar chart of resource scheduling deviation analysis according to an embodiment of the present invention.

[0029] Figure 6 It is a schematic structural diagram of a data analysis platform management system according to an embodiment of the present invention. Detailed Embodiments

[0030] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0031] Refer to Figure 1 , an embodiment of the present invention proposes a data analysis platform management method, which adopts a cross-system dynamic modeling and multi-objective optimization strategy, combines a closed-loop verification mechanism and multi-source heterogeneous data fusion technology, can realize intelligent prediction, precise scheduling and self-learning optimization of medical resource conflicts, ensure traceable execution of the whole process of high-value consumables, and comprehensively improve resource utilization efficiency and emergency response capabilities.

[0032] The method of this embodiment specifically includes:

[0033] Obtain multi-source heterogeneous data from a medical information system, where the data includes patient diagnosis and treatment data, medical consumable data, and equipment operation data;

[0034] Establish a cross-system data association model, perform dynamic feature extraction and weight assignment on the multi-source heterogeneous data, and output a resource supply-demand relationship map;

[0035] Analyze the resource supply-demand relationship map to identify business conflicts in medical resource scheduling;

[0036] Generate a conflict solution for the business conflict based on a multi-objective optimization algorithm, and generate an execution instruction;

[0037] Send the execution instruction to the target medical system, where the instruction includes a resource scheduling path and time limit constraint conditions;

[0038] Verify the execution result of the execution instruction and update the system parameters.

[0039] Specifically, the present invention constructs a resource supply-demand relationship map through multi-source data fusion and dynamic modeling, uses a three-dimensional decision matrix to achieve conflict classification and optimal path selection, and forms a self-learning optimization mechanism in combination with closed-loop verification of electronic tags. It breaks through the data island limitation of traditional systems, realizes global visualization and accurate prediction of cross-level medical resources; reduces the dependence on manual experience through spatio-temporal matching degree analysis, and improves the decision-making efficiency in emergency scenarios; the dual verification mechanism ensures the safety and compliance of medical resource scheduling; the model self-optimization driven by historical experience reduces repeated execution deviation. For example, after a certain tertiary hospital applies it, it can automatically call cross-hospital consumable resources during sudden emergency tasks and complete the delivery within 30 minutes, which is better than the average time-consuming under the traditional manual scheduling mode.

[0040] Optionally, the obtaining of multi-source heterogeneous data from a medical information system includes:

[0041] Connect an electronic medical record system and a material management system through a standardized protocol conversion module, and output a structured interface data stream;

[0042] Specifically, deploy an HL7 v2.6 protocol adapter in the Hospital Information System (HIS) and define field mapping rules; the Supply Chain Pharmacy System (SPD) extracts the coding through a UDI parser, and its coding format is UDI = (01)[14-bit identification code](17)[6-bit expiration date][other fields]; for example, "(01)00884745041942(17)250831" is parsed as the product identification 00884745041942, and the expiration date is August 31, 2025. Generate a structured data stream, and the format is shown in Table 1.

[0043] Table 1 Data Stream Format Table

[0044]

[0045] Exemplarily, assume that a certain hospital needs to monitor the infusion consumable reserves in different departments in real time, such as normal saline and antibiotics. The hospital's HIS records the infusion orders issued by doctors, such as "intravenously drip 2g of ceftriaxone sodium daily", and the SPD tracks the inventory batches and expiration dates of infusion bags in the warehouse. Through a general protocol converter, convert the data of different systems into a unified format. For example, convert the free text order "2g of ceftriaxone sodium, q12h" in the HIS and the UDI coding in the warehouse, such as "expiration date until August 2025" into a structured table, as shown in Table 2:

[0046] Table 2 Structured Table

[0047]

[0048] Based on the structured interface data stream, identify the treatment type keywords and consumable identification code segments in the order text, and convert the unstructured data into medical resource labels in a unified coding format;

[0049] Specifically, based on the BiLSTM-CRF algorithm, segment words to capture medical entities, and the order text outputs data in the format of {verb, noun, frequency} through the algorithm model; for example, input "intravenously drip ceftriaxone sodium, 2g, q12h" and the output label is {operation = intravenously drip, drug = ceftriaxone sodium, dose = 2g, frequency = every 12 hours}; extract key parameters according to the UDI coding structure, such as:

[0050] Among them, the output format of the unstructured data is shown in Table 3.

[0051] Table 3 Example Table of Unstructured Data Output Results

[0052] Exemplarily, the doctor's handwritten emergency medical order may be unstructured text, such as "Emergency! Prioritize the immunoglobulin infusion for the patient in bed 5". Automatically extracting keywords, the treatment type is immunoglobulin infusion, which is classified as "high-priority biologic". After the consumable identifier is matched with the UDI code in the electronic tag, it is converted into a standardized format, such as "Biologic-Immunoglobulin-Batch B15". The emergency flag is automatically marked as "Level 1 Response" based on "Emergency!". The free-text medical order is automatically translated into a machine-readable label for subsequent scheduling.

[0053] According to the attribute characteristics of the medical resource label, adjust the data collection mode at a preset time period. The data collection mode includes:

[0054] Real-time data processing mode, which grabs key fields at a second-level frequency;

[0055] Batch update mode, which integrates the whole-day data to generate an incremental update package.

[0056] Specifically, select the data collection mode according to the time period. For example, from 8:00 to 18:00, the real-time data processing mode is adopted, and from 0:00 to 6:00, the batch update mode is adopted; among them, when collecting the emergency level label of the medical resource label, when the emergency level label is level 2 or level 1, the real-time data processing mode is adopted, and the collected data is {real-time order, inventory change, operating room status}, and the collection frequency is 0.5 seconds / time; in other cases, the collection frequency is 5 seconds / time.

[0057] Optionally, the establishment of a cross-system data association model, the dynamic feature extraction and weight assignment of the multi-source heterogeneous data, and the output resource supply-demand relationship map include:

[0058] Construct a three-dimensional feature coordinate system, the dimensions of which include business type, data format, and timeliness level, to generate an initial feature vector set;

[0059] Specifically, unify the scattered multi-source heterogeneous data, such as diagnosis and treatment records, consumable identifiers, and inventory status, into a quantifiable mathematical model. The business type dimension is classified according to the use of medical resources, such as surgical consumables, emergency drugs, and test reagents. The data format dimension is to unify the data expression form, such as structured tables, semi-structured JSON, and unstructured text. The timeliness level dimension is classified according to the data update frequency, such as real-time, high / medium / low timeliness. As Figure 2 shown, each data entity is mapped to a three-dimensional coordinate point. For example, the feature vector of a certain surgical consumable can be (business type = surgical consumable, data format = structured data, timeliness level = high), which is converted to (X, Y, Z). Normalize the value ranges of different dimensions to generate an initial feature vector set. For example, numerical codes are assigned to the business type, 1 for surgical consumables and 2 for emergency drugs.

[0060] Based on the initial feature vector set, a dynamic weight adjustment mechanism is established. When the application frequency of the same type of consumables exceeds the preset frequency threshold, the weight increase parameter is calculated through the sliding window algorithm, and the feature vectors in the initial feature vector set are adjusted.

[0061] Specifically, the dynamic weight adjustment mechanism is as Figure 3 shown. The frequency threshold for the application of the same type of consumables is preset. For example, when the application of the same type of consumables exceeds 10 times per hour, the adjustment is triggered. The weight increase parameter calculated through the sliding window algorithm can be to count the application records within a certain time period, such as the recent 30 minutes, and calculate the ratio of the actual frequency to the frequency threshold, that is:

[0062] In the formula, is the weight increase parameter, is the actual frequency, is the frequency threshold, is the correction coefficient; for example, when the correction coefficient is 1, the frequency threshold is 10 times, and the actual frequency is 15 times, the weight increase parameter is 0.5, that is, the weight increases by 50%. Update the weight value of the feature vector according to the weight increase parameter. For example, the consumable with the original weight of 1 is adjusted to 1.5, and the adjustment formula is:

[0063] In the formula, is the new weight of the feature vector, is the original weight of the feature vector.

[0064] Input the adjusted feature vectors into the preset graph database, construct the multi-level association relationship between nodes, and output the resource supply and demand relationship map.

[0065] Specifically, visualize the complex relationship between data to support the scheduling decision. Input the adjusted feature vectors into a graph database such as Neo4j. Each node represents an entity, such as a certain type of consumable, a certain department, and a certain surgical device. The node attributes carry three-dimensional feature data, such as the consumable aging level and the business type weight. Based on the dynamic weight, construct the edge relationship. For example, at level 1, it is the "supply-demand" relationship between the consumable and the department, and at level 2, it is the "dependency call" relationship between the department and the device. The weight of the edge is determined by the dynamically adjusted parameter. For example, the edge with a higher weight has a darker color or a thicker connection. Finally, generate a visualized resource supply and demand relationship map, such as a high-weight consumable associated with the emergency surgical needs of multiple departments.

[0066] Exemplarily, assume that the demand for oxygen cylinders in a certain top - tier hospital has increased sharply due to the flu season, and resource scheduling needs to be optimized immediately. The feature vector of the oxygen cylinder is mapped to (business type = emergency equipment, data format = structured UDI code, timeliness level = real - time), and after conversion, the coordinates (5, 2, 8) are obtained. The range of the business type dimension is 1 - 5. For example, 1 = ordinary equipment, 5 = emergency equipment, and the oxygen cylinder is assigned a value of 5 due to its high priority. It is detected that the application frequency of oxygen cylinders in the respiratory department reaches 15 times within 30 minutes, and the frequency threshold is 10 times. Therefore, the weight increase parameter is calculated to be 0.5, so the weight is adjusted to 1.5 times the original value. Thus, the timeliness level of the oxygen cylinder is increased from 8 to 12, and the priority is significantly improved. Then, graph database modeling is carried out. The nodes are the respiratory department with a weight of 12, oxygen cylinder A with a weight of 12, the ICU with a weight of 10, and the oxygen supply vehicle with a weight of 8. The edges are defined as from the respiratory department to oxygen cylinder A, with an edge weight = 12, red thick line, high urgency; from the ICU to oxygen cylinder A, with an edge weight = 8, yellow medium - line, medium demand. The graph shows that the respiratory department has the highest dependence on oxygen cylinders, and its supply should be prioritized.

[0067] Optionally, the identification of business conflicts in medical resource scheduling includes:

[0068] Scanning the multi - system data records of medical resources to extract the three - value difference set of inventory, allocated quantity, and in - transit quantity;

[0069] Specifically, integrating the hospital information system (HIS), supply chain management system (SPD), and logistics tracking system (LIS) to obtain real - time data. The real - time data obtained includes inventory, allocated quantity, in - transit quantity, and surgical demand to be allocated. The inventory is the currently available inventory in the SPD database, the allocated quantity is the number of consumables marked as "occupied by surgery" in HIS but not yet received, the in - transit quantity is the quantity of consumables in transit in LIS and the estimated arrival time, and the surgical demand to be allocated is extracted from the surgical appointment system for the consumable demand of planned surgeries within the next 72 hours. The data in the difference set includes the shortage quantity, and the calculation logic of the shortage quantity is "shortage quantity = allocated quantity + surgical demand to be allocated-(inventory + in - transit quantity)". When the shortage quantity is greater than 0.2 times the inventory, an exception is triggered, and a triple {inventory, allocated quantity, in - transit quantity} is generated. The difference set is stored in a dynamic table, and each record contains a timestamp, consumable type, and shortage quantity.

[0070] Based on the difference set, analyzing the spatio - temporal matching degree between operating room reservation data and consumable inventory data to locate the conflict coordinates;

[0071] Specifically, analyze the time interval of the surgical plan in the reservation system, and check whether the expected arrival time of the in-transit quantity meets the condition that the expected arrival time of the in-transit quantity is less than the surgical end time minus the buffer time, where the buffer time is defaulted to 2 hours. If there is a situation where the expected arrival time of the in-transit quantity does not meet the condition or the inventory is insufficient to complete the current surgery, generate conflict coordinates to obtain the coordinates (department, time interval of the surgical plan, consumable type).

[0072] Calculate the conflict level and generate a conflict warning signal according to the node parameters of the resource supply and demand relationship graph.

[0073] Specifically, obtaining the node parameters of the resource supply and demand relationship graph includes the coordinate values corresponding to the consumable type. For the conflict level , there is In the formula, is the weight coefficient, defaulted to 0.4, 0.4, 0.2, and can be configured by the hospital. is the shortfall quantity. is the unallocated demand. is the coordinate value corresponding to the consumable type. is the reciprocal of the time from the expected arrival time of the in-transit quantity to the start of the surgery. is the total time window. For , assuming the consumable type is an oxygen cylinder, the coordinates of the node corresponding to the oxygen cylinder in the resource supply and demand relationship graph are the three-dimensional characteristic data of the oxygen cylinder, that is, the characteristic vector of the oxygen cylinder (business type = oxygen cylinder, data format = structured UDI code, timeliness level = real-time), for example, it can be (5, 2, 8) mentioned above. Since it is detected that the application frequency of the oxygen cylinder in the respiratory department reaches 15 times within 30 minutes and the frequency threshold is 10 times, the calculated weight increase parameter is 0.5. Therefore, the weight is adjusted to 1.5 times the original value, so the timeliness level of the oxygen cylinder is increased from 8 to 12. Then at this time is 12. Then, classify the level according to the value of the conflict level. For example, when is greater than 8, it is a red warning. When is greater than 4 and less than or equal to 8, it is a yellow warning. When is less than or equal to 4, it is a blue warning.

[0074] Optionally, the conflict solution for generating the business conflict based on the multi-objective optimization algorithm includes:

[0075] Receive the conflict warning signal, and configure the timeliness priority strategy for clinical treatment decision-making and the cost priority strategy for administrative material decision-making.

[0076] Specifically, when the system receives a conflict warning signal, it matches the decision-making strategy according to the conflict level. For clinical treatment conflicts, such as a shortage of emergency surgery resources, the time-efficiency priority strategy is triggered. This strategy sets the time-efficiency weight ratio of medical supplies arriving at the operating room to 70% and the transportation cost weight to 30%, and stipulates that the response time of the resource scheduling path shall not exceed 15 minutes. For administrative supply conflicts, such as the over-limit inventory of general outpatient consumables, the cost-priority strategy is implemented. The transportation cost weight is increased to 65% and the time-efficiency weight is reduced to 35%, and the dispatching response time is allowed to be extended to 4 hours. When configuring the strategy, the pre-set parameter table is called to dynamically adjust the coefficient combination of each factor in the objective function, and at the same time, a time-efficiency red line constraint condition is inserted into the decision-making model.

[0077] The time-efficiency priority strategy is a medical resource scheduling rule for critically ill patients, requiring key supplies to be delivered to the designated location within the shortest time. The cost-priority strategy is an optimization rule for the distribution of non-emergency supplies, preferentially selecting the path and method with the lowest total transportation cost.

[0078] Exemplarily, an acute myocardial infarction case suddenly occurs in the inpatient department of a certain tertiary hospital, and the electrocardiogram system automatically triggers a catheterization laboratory activation request. At this time, the system identifies that the conflict type is clinical treatment, and activates the time-efficiency priority strategy. The time constraint condition is set to "complete all interventional consumable preparations and personnel arrival within 20 minutes". When making a transportation path decision, the 3rd-floor consumable warehouse closest to the catheterization laboratory is preferentially selected instead of the farther central warehouse. The system automatically shields the cost overrun warning, forcibly retrieves high-priority inventory, and generates the shortest path navigation to the operating room. By automatically distinguishing clinical emergencies from administrative affairs, it ensures that the timeliness of the emergency scenario is not dragged down by the regular process, realizes the seamless connection of medical resources during the rescue window period, and guarantees the life safety of patients.

[0079] Generate a multi-path scheduling plan according to the resource supply-demand relationship map, including in-hospital transfer, emergency inventory call, and cross-hospital transfer;

[0080] Specifically, based on the topological structure of the resource supply-demand relationship map, search for multi-level associated paths of the current material node. The search depth of in-hospital transfer is set to 3 levels of associated nodes, and the path planning algorithm automatically matches the locations of alternative resources within the same hospital and calculates the transportation distance. The trigger condition for emergency inventory call is that the regular inventory is lower than 20% of the safety threshold. When this strategy is activated, it will connect to the regional medical supply reserve center and generate a special scheduling plan including cold chain transportation requirements. The path evaluation of cross-hospital transfer includes real-time access to the traffic condition prediction module, and dynamically calculates the time and toll required for cross-regional logistics. In-hospital transfer is the path planning for the material flow between different departments or warehouses within the medical institution. Emergency inventory call is the distribution plan for obtaining resources from the regional centralized reserve warehouse or strategic reserve. Cross-hospital transfer is the collaborative transportation planning of materials between different medical institutions.

[0081] Exemplarily, in an orthopedic surgery, it is found that a specific type of artificial joint prosthesis is missing, and the system starts a three-level path exploration. First, search within the same hospital area and find that there is available inventory in the bone trauma department warehouse on the 14th floor, generating a transport plan using the walking ladder; second, check the data in the emergency reserve warehouse and find that this type of material is available in the emergency center of the adjacent administrative region, planning a special approval channel for an ambulance to pick up the goods; finally, retrieve the database of the regional medical consortium, locate the inventory in a cooperative hospital 80 kilometers away, generate a helicopter emergency transportation plan and estimate the impact of the route weather. Establish a multi-level resource scheduling network, quickly expand the search radius when the single-point inventory is insufficient, form a three-dimensional resource replenishment plan, and effectively avoid the interruption of the medical process caused by local shortages.

[0082] Execute the database according to the preset historical scheduling plan, predict and calculate the time deviation rate, cost deviation value, and feasibility of each plan, generating a three-dimensional decision matrix;

[0083] Select a scheduling plan using the three-dimensional decision matrix.

[0084] Specifically, obtain the historical scheduling plan, historical time data, and historical cost data based on the historical scheduling records to construct a training set. Use the historical scheduling plan as the input, and the historical time data and historical cost data as the output to establish and train a neural network model. Finally, obtain a time-cost prediction model. Input the selected scheduling plan into the time-cost prediction model to output the predicted time and predicted cost.

[0085] Among them, the time deviation rate The calculation formula is:

[0086] In the formula, is the predicted time, is the actual time. The cost deviation value is calculated by the following formula:

[0087] In the formula, is the predicted cost, is the actual cost. Combine , and the plan feasibility to form a three-dimensional decision space, where the feasibility is calculated by double-parameter weighting of the path passing rate and protocol compliance. For the feasibility , there is In the formula, is the weight of the path passing rate, defaulting to 0.6, which is configurable, is the path passing rate, that is, the historical execution completion rate. For example, 0.95 means a 95% success rate, is the weight of the protocol compliance, defaulting to 0.4, satisfying , It is the protocol compliance degree, that is, the probability of meeting the medical transportation specifications, such as the passing rate of cold chain integrity inspection.

[0088] Finally, it is presented as an interactive three-dimensional matrix view, and the optimal solution is located in the low deviation area of the first quadrant. The three-dimensional decision matrix is a three-dimensional evaluation model that includes the time dimension, cost dimension, and feasibility dimension. The scale of each dimension is transformed into a percentage index through normalization processing. The time deviation rate is the ratio of the relative error between the actual execution time and the predicted time. The cost deviation value is the absolute difference between the actual cost and the estimated cost.

[0089] Exemplarily, the system generates three solutions for the scheduling requirements of a certain oncology drug. Solution A is predicted to take 45 minutes to complete the transportation, with an estimated cost of 1200 yuan; Solution B is predicted to take 60 minutes, with a cost of 900 yuan; Solution C is predicted to take 30 minutes, with a cost of 1800 yuan. Historical data training shows that the average is ±8%, and the fluctuation range is ±15%. After execution, the actual data is that Solution A takes 50 minutes and costs 1250 yuan, Solution B takes 75 minutes and costs 850 yuan, and Solution C takes 28 minutes and costs 2000 yuan. It is calculated that , ; , ; , . Among them, the path passing rate of Solution A is 95%, the compliance rate is 90%, and the calculated feasibility is 0.93. The path passing rate of Solution B is 80%, the compliance rate is 85%, and the calculated feasibility is 0.82. The path passing rate of Solution C is 60%, the compliance rate is 75%, and the calculated feasibility is 0.66. As Figure 4 shown, the X-axis is the time deviation rate, with negative values indicating faster execution than expected and positive values indicating the opposite. The Y-axis is the cost deviation value, with negative values indicating cost overruns and positive values indicating savings. The Z-axis is the feasibility, and the higher the value, the higher the comprehensive score of the solution. Therefore, Solution A is the optimal choice. By visually displaying the multi-objective optimization results in three dimensions, decision-makers can quickly balance among timeliness, cost, and reliability, and preferentially select solutions at high positions, significantly reducing the complexity of manual analysis and improving the scientificity of resource scheduling decisions. By quantitatively evaluating the execution deviations of historical solutions, a traceable decision optimization system is constructed to assist managers in selecting the scheduling solution with the optimal comprehensive benefits under multiple constraints and reducing the decision-making trial-and-error costs.

[0090] Optionally, issuing the execution instruction to the target medical system includes:

[0091] Decomposing the selected scheduling solution into standardized instruction units, and each instruction unit includes the target system identifier, operation content code, and time limit mark;

[0092] Specifically, the selected resource scheduling plan needs to be converted into machine instructions recognizable by the medical system. The splitting process decomposes the scheduling path into independent operation units according to the medical Internet of Things standard protocol. Each instruction unit contains a target system identification field, which is automatically generated by the hospital coding system and has the format of administrative area code + hospital area code + floor number + department code, such as Jing A01-C3053A. The operation content coding uses the mapping of the international general medical operation code library. For example, the coding extension of infusion operations is HCPCS-J7040. The time limit mark presets the execution time limit according to the total time of the scheduling path and realizes millisecond-level synchronization through binary time stamp coding. The generation logic of the standardized instruction unit needs to follow the predefined template in XML format and ensure data integrity through redundant check bits.

[0093] Among them, the standardized instruction unit is a set of computer-executable commands split according to predefined rules, including the operation object, action type, and time requirement. The target system identification is the unique code of the medical node receiving the instruction in the Internet of Things. The operation content coding is the international general medical service coding rule. The time limit mark is the time threshold identification generated based on the total duration of the scheduling path.

[0094] Exemplarily, a tertiary hospital needs to execute a cross-floor ventilator scheduling, and the resource scheduling plan is to transfer from the ICU on the 7th floor to the emergency department on the 3rd floor. The system automatically generates three instruction units: Unit 1, the target identification is the ICU medical equipment controller, and the content coding is to remove the fixing device; Unit 2, the target identification is the logistics robot navigation module, and the content coding is to set the transfer path L7→L3; Unit 3, the target identification is the emergency department material receiving terminal, and the content coding is equipment access detection. Each unit marks the time point that must be completed and is packaged into a JSON instruction stream that conforms to the HL7 FHIR standard. Through instruction splitting, the complex scheduling is accurately disassembled into independently executable atomic operations, ensuring that the responsibility of each link is traceable and the time limit is quantifiable. By standardizing instruction splitting, semantic ambiguity in traditional manual scheduling is avoided, and the risk of operation errors is reduced with machine-readable coding, improving the compatibility and reliability of cross-system instruction execution.

[0095] Install electronic tag verification points on the consumable transfer path to collect and feedback the consumable coordinates and timestamps;

[0096] Specifically, interface devices based on RFID or NFC technology are installed at the key nodes of the medical consumables transportation path. The distance between verification points is set according to the category of consumables. For example, a detection device is arranged every 20 meters for high-value consumables, and a 50-meter interval is set for general consumables. The electronic tag ID on the consumable package is read at each verification point, the real-time geographical coordinates are obtained through the Beidou / GPS positioning module, and the Unix timestamp is recorded. The collected data is transmitted back to the central monitoring system via the LoRa wireless protocol. When the transmission delay between two consecutive verification points exceeds the set threshold, a path deviation warning is automatically triggered. For refrigerated consumables, the verification points synchronously collect temperature sensor data and compare it with the preset cold chain specifications. The storage chip of the verification point has the function of power-off data protection to ensure the integrity of the audit trail data.

[0097] Among them, the electronic tag verification point is an intelligent node device with reading, positioning and communication functions, which is deployed on the physical path of the consumables transportation channel. The consumable coordinates are the longitude and latitude data of the geographical location coordinate system. The timestamp is the execution time identifier recorded by the computer system accurate to milliseconds.

[0098] Exemplarily, a certain tertiary hospital transports a heart stent to the catheterization laboratory, and the path runs through the medical street, the surgical dedicated passage and the clean area. 7 verification points are arranged along the path. The coordinates of the admission gate are (116.403, 39.914), the coordinates of the transition area of the surgical building are (116.4035, 39.915), and the coordinates of the buffer area of the clean area are (116.4037, 39.916). When the transfer box carrying the stent passes by, each verification point reads the tag code and records the time, such as 09:30:15.234, 09:31:03.112, 09:32:57.889. The system compares the preset standard path time curve, detects that the stay time in the clean area exceeds the standard by 120 seconds, and immediately notifies the quality control department to intervene. Through dynamic path monitoring, the whole process of high-value consumables transportation can be traced, and any abnormal deviation can be warned in time. A monitoring network integrating physical and digital is constructed to realize the visualization of the process of medical supplies transportation, and the anti-loss and anti-tampering capabilities are improved by cross-verifying spatio-temporal data.

[0099] When a high-value consumable scheduling instruction is detected, a dual electronic signature confirmation mechanism is activated.

[0100] Specifically, the high-value category is identified through the prefix of the consumable UDI code, triggering the security verification process. The first signature is an asymmetric encryption signature generated by the authorized person in the medical department using the national cryptographic SM2 algorithm to verify the compliance of consumable application. The second signature is the digital certificate signature issued by the head of the equipment receiving department through the PKI system to confirm the receiving qualification. After the dual signatures are generated, blockchain evidence preservation will be executed. Each signature contains the operator's work number, timestamp, and device fingerprint information. The signature data is bound to the consumable logistics track to form an immutable audit chain. If the signature verification fails three times, the system will automatically freeze the relevant scheduling process and activate the manual review channel. For implantable high-value consumables, an additional electronic signature link for the surgeon is added.

[0101] Among them, high-value consumables are medical items with a unit price exceeding the set medical payment standard, such as artificial joints, heart stents, etc. The dual electronic signature confirmation mechanism is to use two independent digital identity authentication technologies to perform security verification on key operations. Blockchain evidence preservation is the encrypted storage of important operation records by distributed ledger technology.

[0102] Exemplarily, a certain patient needs to urgently use an artificial heart valve. The system identifies that the UDI code of this consumable is (01)88521023456784(21)A1015 and determines that it belongs to high-value consumables. When generating the scheduling instruction, it is required that the director of the cardiac surgery department and the head nurse of the interventional catheterization room sign through the mobile digital certificate respectively. The first signature confirms that the surgical indications conform to the medical insurance catalog, and the second signature verifies the material storage qualification of the receiving department. The two signatures are transmitted to the blockchain node through quantum encryption and generate a hash value of 0x7d5a...ef2c. When the consumable arrives, the system compares the symmetric name hash chain. If the match is successful, the safe will be unlocked. Through dual verification, the risk of illegal invocation is eliminated, ensuring the full life cycle supervision of high-value consumables. Establish a multi-layer defense material security guarantee system. Through the combination of cryptography technology and distributed ledger, effectively prevent the theft and abuse of medical resources and protect the rights and interests of patients.

[0103] Optionally, verifying the execution result of the execution instruction and updating the system parameters includes:

[0104] Based on the consumable coordinates and timestamp, the consumable position data and timeliness data during the execution of the instruction are collected in real time;

[0105] Specifically, the system continuously obtains the consumable logistics data through the Internet of Things sensing nodes deployed on the medical transportation path. The position data comes from the built-in Beidou / GPS dual-mode positioning chip in the transportation equipment, and the collection frequency is set to 1 time per second. The timeliness data is calculated by comparing the planned timestamp with the time difference between the actual arrival at the verification point. The time difference The calculation formula is: Among them, is the time node preset in the scheduling plan. The Unix timestamp recorded for the electronic tag verification point. The multi-source heterogeneous data is integrated every 3 seconds and uploaded to the central monitoring system database through the MQTT protocol. The data acquisition mechanism includes a fault-tolerant design. When the signal of a single node is lost, the redundant data interpolation of adjacent verification points is automatically enabled for completion. The consumable coordinates are the real-time geographical location coordinates of medical supplies in three-dimensional space, including longitude, latitude, and floor height. The timeliness data is the difference between the actual transportation time and the planned time, reflecting the deviation of the execution progress. The electronic tag verification point is a fixed-point detection device equipped with a radio frequency identification and time synchronization module, used to verify logistics nodes.

[0106] Based on the consumable location data and the timeliness data, calculate the time deviation degree and the space deviation degree between the actual execution path and the optimized path, generate the comprehensive deviation degree, and generate the alarm level based on the comprehensive deviation degree;

[0107] Specifically, the path deviation degree calculation adopts a spatio-temporal two-dimensional comprehensive evaluation model. The space deviation degree The calculation formula is: The time deviation degree The calculation formula is: The comprehensive deviation degree The fusion formula is: Among them, is the actual coordinate, is the planned coordinate, is the historical maximum deviation distance, is the total allowable duration, is the weight coefficient, default , , and is configurable. The alarm level determination rule is is green for normal, is yellow for observation, is red for emergency. When a red alarm is detected, the system automatically freezes the subsequent scheduling instructions and waits for manual confirmation. The actual execution path is the sequence of actual trajectory coordinates of the consumable transportation. The optimized path is the theoretical trajectory in the optimal scheduling scheme generated by the algorithm. The deviation degree is a quantitative index of the difference between the actual transportation path and the ideal path. The alarm level is a response priority label divided according to the deviation degree.

[0108] Based on the alarm level, trigger the immediate correction instruction of the weight increase parameter, and use the comprehensive deviation degree to adjust the weight increase parameter.

[0109] Specifically, the weight increase parameter is a dynamic coefficient for adjusting the importance of feature vectors in the data association model. The immediate correction instruction is a model adaptive adjustment instruction triggered based on real-time deviation data. When the warning level is non-green, the immediate correction instruction is triggered. The immediate correction instruction adjusts the weight increase parameter by calculating the correction coefficient. For the current correction coefficient , there is In the formula, is the th correction coefficient, that is, the current correction coefficient, is the th correction coefficient, that is, the previous correction coefficient, is the dynamic coefficient, is the constant coefficient, is the comprehensive deviation threshold; among them, is adjusted according to the warning level. When the warning level is yellow, When the warning level is red,

[0110] This method realizes an optimization mechanism of closed-loop feedback by real-time tracking the spatio-temporal data of the medical resource scheduling path, dynamically evaluating the execution deviation and adaptively adjusting the model parameters. It improves the adaptability of the dynamic model to abnormal situations, optimizes the accuracy of subsequent resource scheduling, strengthens the risk control ability through warning classification, and combines electronic tags and correction algorithms to ensure the reliability and self-learning ability of the scheduling process. For example, when the cold chain drug transportation is delayed, the system automatically increases the model weight of such consumables and preferentially matches the spare path resources.

[0111] Exemplarily, a certain hospital schedules a batch of vaccines to the emergency department. The planned path passes through the corridor of Area A, with coordinates X = 102, Y = 58, and the planned time is 8 minutes. During actual transportation, due to temporary control, it detours through Area B, with coordinates X = 105, Y = 55, and it takes 12 minutes. The calculated spatial deviation is 4.24, the time deviation is 50%, and the comprehensive deviation is 22.54%, triggering a red warning. Since the warning level is a red warning, an immediate correction instruction is triggered. At this time, the dynamic coefficient ; assuming that the previous correction coefficient is 1, the comprehensive deviation threshold is 10, is 0.2. Therefore, the calculated current correction coefficient is 1.3. Therefore, the weight increase parameter of this type of vaccine is increased by 30%. It can quickly identify deviations in case of sudden path blockage, dynamically enhance the model weight of cold chain drugs, ensure that emergency needs are preferentially matched with reliable paths, and reduce the execution risk of subsequent tasks.

[0112] Optionally, verifying the execution result of the execution instruction and updating the system parameters further includes:

[0113] Integrate the execution results to build an experience knowledge base, which includes successful case templates and exception handling reports;

[0114] Specifically, after each medical resource scheduling task is completed, core indicator data such as execution timeliness, path deviation degree, and resource utilization rate are automatically collected. The execution timeliness is the time deviation of department consumables from the average value. The path deviation degree is the average value of the time deviation degrees of various department consumables. The resource utilization rate is calculated by the ratio of the actual used quantity of department consumables to the allocated quantity of department consumables. The experience knowledge base adopts a hierarchical storage structure, and data cleaning rules are set to filter out invalid records. The successful case template is defined as historical records with a path deviation degree continuously lower than 5% and a resource waste rate less than 3%, and is classified and stored by department type through the K-means clustering algorithm. The exception handling report records the handling process of emergencies such as resource loss and cold chain breakage, and associates the original conflict warning level with the corrective measures. Each report is marked with a root cause code, such as "EC-102" representing the timeliness delay caused by transportation equipment failure. The construction mechanism performs incremental updates every day at midnight, retaining the valid data of the most recent 365 days.

[0115] Among them, the experience knowledge base is a structured database storing historical scheduling cases, including positive experiences and problem handling records. The successful case template is a scheduling path template that meets the preset excellent standards. The exception handling report is a fault file recording resource scheduling accidents and countermeasures.

[0116] Periodically adjust the system parameters according to the historical data of the experience knowledge base; specifically, the adjustment period can be once a month, and the system parameters to be adjusted include the frequency threshold , the safety threshold of the inventory quantity, and the comprehensive deviation degree threshold . When the proportion of successful cases of a certain type of consumable in the experience knowledge base exceeds 60%, the safety threshold of the inventory quantity automatically decreases by 5%. After the parameters are updated, three months of historical data need to be simulated in the sandbox environment to verify the improvement effect before deployment. For example, due to the stable surgical volume of a certain artificial joint consumable, the safety threshold of the inventory quantity is reduced from 200 sets to 190 sets, releasing storage space while ensuring continuous supply.

[0117] Establish an energy efficiency score for department execution, and update the weight of the corresponding node of the department in the resource supply and demand relationship map according to the score result.

[0118] Specifically, the department execution energy efficiency score is a comprehensive evaluation index for quantifying the department's resource scheduling ability. For the department execution energy efficiency score , there is In the formula, is the weight coefficient, , is the actual used quantity of department consumables, is the allocated quantity of department consumables, is the average spatial deviation degree of department consumables, is the average time deviation of department consumables; among them, the allocated quantity of department consumables is the data in the three-value difference set, is calculated through the spatial deviation degree in the foregoing, and is the average value of the spatial deviation degrees of various consumables in the department; is calculated through the time deviation degree in the foregoing, and is the average value of the time deviation degrees of various consumables in the department. Map the execution energy efficiency score of the department to the corresponding node in the resource supply-demand relationship graph, and update the weight of the corresponding node. For example, if the execution energy efficiency score is above 90 points, it is grade A, then the corresponding weight of the department is multiplied by 1.2; if the execution energy efficiency score is between 75 and 89 points, it is grade B, then the corresponding weight of the department is multiplied by 1.0; if the execution energy efficiency score is between 60 and 74 points, it is grade C, then the corresponding weight of the department is multiplied by 0.8. The execution energy efficiency score is updated monthly and synchronized to the resource supply-demand relationship graph. For example, because the emergency department has scored 92 points for three consecutive months, the associated weight of its node has increased by 20%, and it will be given priority for the allocation of scarce resources. The energy efficiency score is synchronously pushed to the hospital performance appraisal system as the basis for department management evaluation.

[0119] The present invention realizes the continuous improvement of medical resource scheduling by constructing a self-learning optimization system with closed-loop feedback. The experience knowledge base accumulates historical operation data to form a decision-making think tank, and the periodic parameter tuning ensures that the model keeps up with the times. The execution energy efficiency scoring mechanism encourages departments to improve their operation efficiency. It can quickly match the best practice plan in case of sudden resource demand, and the experience of handling abnormal events forms a standardized pre-plan to effectively avoid repeated mistakes. The dynamic adjustment of department weights promotes the reasonable flow of medical resources and constructs an adaptive medical ecosystem.

[0120] Optionally, the method further includes:

[0121] Perform abnormal judgment according to the spatial deviation degree. When an abnormality is detected, load the successful case template and switch to the degraded mode;

[0122] Specifically, the determination of the abnormal state is realized by monitoring the average value of the spatial deviation degree. The analysis of the resource scheduling deviation degree is as Figure 5 shown. When the average value of the spatial deviation degree of any transport unit exceeds 15%, trigger the abnormal detection mechanism and activate the degraded mode. At this time, retrieve the successful case template with matching attributes from the experience knowledge base. For example, select the historical scheduling plan with a time effect deviation rate less than 8% and a path repeatability higher than 90% as the template benchmark. The degraded mode is a simplified operation mode in the emergency state, which closes the non-core algorithm module and compresses the data transmission volume.

[0123] Among them, the anomaly detection mechanism makes judgments through real-time data monitoring. The degradation mode is a disaster tolerance operation state that reduces the system operation accuracy but ensures the continuity of critical services. The successful case template is a standardized storage template for historical scheduling schemes that meet the preset execution criteria, including path trajectories, time-consuming records, and resource utilization parameters. For example, if the spatial deviation of a consumable transportation contains three stages (25%, 12%, 18%), the average deviation is 18.33%, which is greater than 15%. Therefore, the abnormal state is determined and the degradation mode is switched.

[0124] According to the successful case template, call the path data with the smallest average value of the spatial deviation within the preset time window to generate an emergency scheduling scheme;

[0125] Specifically, filter out the path with the smallest average value of the spatial deviation within the preset time window. The time window can be set to 24 hours, and the filtered path data is used as an alternative solution to construct an emergency scheduling scheme. For example, three candidate paths are retrieved. The average value of the spatial deviation of path A is 9.2%, the average value of the spatial deviation of path B is 11.8%, and the average value of the spatial deviation of path C is 8.7%. Select path C with the smallest average value of the spatial deviation to generate an emergency plan, shortening the manual decision-making time.

[0126] In the degradation mode, reduce the data collection frequency, simplify the label conversion, and allocate emergency resources according to the scoring results.

[0127] Specifically, adjust the data transmission protocol to improve the response speed. The data collection frequency is reduced from the second level to the minute level, and non-critical data points are filtered through a filtering algorithm. The redundant verification steps are cancelled in the label conversion process, and the core field mapping is retained. The emergency resource allocation is sorted from largest to smallest according to the latest department execution energy efficiency score, and is allocated proportionally according to the sorting results; for example, the department with the highest execution energy efficiency score of level A ranks first and gets 70% of the emergency resource allocation, the department with the second highest execution energy efficiency score of level B gets 25%, and the rest are allocated to the common resource pool.

[0128] Among them, label conversion is a data cleaning process that transcribes medical identifiers with different coding rules into a standard format. The filtering algorithm is a noise reduction processing method that generates a data fitting curve according to the least squares principle.

[0129] This method realizes fast disaster tolerance on the premise of ensuring the basic functions of the system by constructing a two-layer emergency response mechanism. The organic combination of anomaly detection and degradation mode can avoid system-level crashes, and the case-driven emergency plan generation significantly improves the decision-making efficiency. The dynamically adjusted data collection strategy reduces the system load while ensuring the integrity of core business data, and the scoring-based resource allocation mechanism ensures that emergency resources are tilted towards high-performance departments.

[0130] Based on the same inventive concept, such asFigure 6 As shown in the figure, the present invention also provides a data analysis platform management system, which includes:

[0131] A data acquisition module, configured to obtain multi-source heterogeneous data from a medical information system, where the data includes patient diagnosis and treatment data, medical consumable data, and equipment operation data;

[0132] A supply-demand relationship establishment module, configured to establish a cross-system data association model, perform dynamic feature extraction and weight assignment on the multi-source heterogeneous data, and output a resource supply-demand relationship map;

[0133] A business conflict identification module, configured to analyze the resource supply-demand relationship map and identify business conflicts in medical resource scheduling;

[0134] A decision generation module, configured to generate a conflict solution for the business conflict based on a multi-objective optimization algorithm and generate an execution instruction;

[0135] An instruction execution module, configured to send the execution instruction to a target medical system, where the instruction includes a resource scheduling path and a time limit constraint condition;

[0136] A closed-loop verification module, configured to verify the execution result of the execution instruction and update system parameters;

[0137] It should be noted that the electrical connections between the above-mentioned units do not necessarily mean direct connections of the lines. Indirect connection methods, as long as they can achieve the purpose of the present invention, are applicable to the embodiments of the present invention. The above are only exemplary embodiments of the present invention and cannot be used to limit the scope of the present invention.

[0138] That is, any equivalent changes and modifications made according to the teachings of the present invention still fall within the scope covered by the present invention. Those skilled in the art will easily think of other implementation schemes of the present invention after considering the specification and the disclosure of the practical truth. This application aims to cover any variations, uses, or adaptive changes of the present invention, which follow the general principles of the present invention and include common general knowledge or conventional technical means in the technical field not recorded in the present invention.

Claims

1. A data analysis platform management method, characterized in that: The method comprises: Acquire multi-source heterogeneous data from medical information systems, including patient diagnosis and treatment data, medical consumables data, and equipment operation data; Establish a cross-system data association model, perform dynamic feature extraction and weight assignment on the multi-source heterogeneous data, and output a resource supply and demand relationship map, wherein the resource supply and demand relationship map uses consumables, departments, and equipment as nodes and the association relationship between nodes as edges; Analyze the resource supply and demand relationship map to identify business conflicts in medical resource scheduling; Generate a conflict resolution method for the business conflict based on a multi-objective optimization algorithm and generate an execution instruction; Sending the execution instruction to the target medical system, wherein the instruction includes a resource scheduling path and a time constraint condition; Verifying the execution result of the execution instruction and updating system parameters; The establishment of a cross-system data association model, dynamic feature extraction and weight allocation of the multi-source heterogeneous data, and output of a resource supply and demand relationship map include: Constructing a three-dimensional feature coordinate system, wherein the dimensions of the three-dimensional feature coordinate system include business type, data format and timeliness level, and generating an initial feature vector set; Based on the initial feature vector set, a dynamic weight adjustment mechanism is established. When it is detected that the frequency of claiming for the same type of consumables exceeds a preset frequency threshold, a weight increase parameter is calculated by a sliding window algorithm to adjust the feature vectors in the initial feature vector set; The adjusted feature vector is input into the preset graph database, a multi-level association relationship between nodes is constructed, and a resource supply and demand relationship map is output.

2. A data analysis platform management method according to claim 1, characterized in that: The acquisition of multi-source heterogeneous data from the medical information system includes: Connect the electronic medical record system and the material management system through a standardized protocol conversion module, and output a structured interface data stream; Based on the structured interface data stream, identifying treatment type keywords and consumable identification code segments in the medical order text, and converting unstructured data into medical resource tags in a unified coding format; According to the attribute characteristics of the medical resource tag, the data collection mode is adjusted according to the preset time period, and the data collection mode includes: Real-time data processing mode, capturing key fields at a frequency of seconds; Batch update mode integrates all-day data to generate incremental update packages.

3. A data analysis platform management method according to claim 1, characterized in that: The identifying of business conflicts in medical resource scheduling includes: Scan multi-system data records of medical resources and extract the difference set of three values: inventory quantity, allocated quantity, and in-transit quantity; Based on the difference set, analyzing the spatiotemporal matching between the operating room reservation data and the consumables inventory data, and locating the conflicting coordinates; According to the node parameters of the resource supply and demand relationship map, the conflict level is calculated and a conflict warning signal is generated.

4. A data analysis platform management method according to claim 3, characterized in that: The conflict resolution method for generating the business conflict based on the multi-objective optimization algorithm includes: Receiving the conflict warning signal, configuring a time priority strategy for clinical treatment decisions and a cost priority strategy for administrative material decisions; Generate a multi-path scheduling plan based on the resource supply and demand relationship map, including intra-campus transfer, emergency inventory call and inter-campus transfer; According to the preset historical scheduling plan execution database, the timeliness deviation rate, cost deviation value and feasibility of each plan are predicted and calculated to generate a three-dimensional decision matrix; A scheduling scheme is selected using the three-dimensional decision matrix.

5. A data analysis platform management method according to claim 1, characterized in that: The sending of the execution instruction to the target medical system includes: Decompose the selected scheduling scheme into standardized instruction units, each of which contains the target system identification, operation content coding and timeliness mark; Set up electronic tag verification points on the consumables transfer path to collect and feedback the consumables coordinates and timestamps; When a high-value consumables scheduling instruction is detected, a double electronic signature confirmation mechanism is activated.

6. A data analysis platform management method according to claim 5, characterized in that: The verifying the execution result of the execution instruction and updating the system parameters includes: Based on the consumable coordinates and timestamps, real-time collection of consumable position data and timeliness data during the execution of the instruction; Based on the consumables location data and timeliness data, calculate the time deviation and space deviation between the actual execution path and the optimized path, generate a comprehensive deviation, and generate an alarm level based on the comprehensive deviation; Based on the alarm level, an immediate correction instruction of the weight increase parameter is triggered, and the weight increase parameter is adjusted using the comprehensive deviation.

7. A data analysis platform management method according to claim 6, characterized in that: The verifying the execution result of the execution instruction and updating the system parameters also includes: Integrate the execution results to build an experience knowledge base, which includes successful case templates and exception handling reports; periodically adjusting system parameters according to historical data of the experience knowledge base; Establish department execution energy efficiency scoring, and update the weight of the corresponding node of the department in the resource supply and demand relationship map according to the scoring results.

8. A data analysis platform management method according to claim 7, characterized in that: The method further comprises: Anomaly judgment is performed according to the spatial deviation, and when an anomaly is detected, the success case template is loaded and switched to a degradation mode; According to the successful case template, the path data with the smallest average spatial deviation within the preset time window is called to generate an emergency dispatch plan; In degraded mode, data collection frequency is reduced, label conversion is simplified, and urgent resources are allocated according to the scoring results.

9. A data analysis platform management system, applied to a data analysis platform management method as claimed in any one of claims 1 to 8, characterized in that: The system comprises: A data acquisition module is used to acquire multi-source heterogeneous data from a medical information system, including patient diagnosis and treatment data, medical consumables data, and equipment operation data; A supply and demand relationship establishment module is used to construct a three-dimensional feature coordinate system, the dimensions of which include business type, data format and timeliness level, to generate an initial feature vector set; and based on the initial feature vector set, a weight dynamic adjustment mechanism is established. When it is detected that the frequency of application for the same type of consumables exceeds a preset frequency threshold, a weight increase parameter is calculated by a sliding window algorithm to adjust the feature vector in the initial feature vector set; and the adjusted feature vector is input into a preset graph database to construct a multi-level association relationship between nodes, and output a resource supply and demand relationship map; wherein the resource supply and demand relationship map uses consumables, departments, and equipment as nodes, and the association relationship between nodes as edges; A business conflict identification module, used to analyze the resource supply and demand relationship map and identify business conflicts in medical resource scheduling; A decision generation module, used to generate a conflict resolution method for the business conflict based on a multi-objective optimization algorithm and generate an execution instruction; An instruction execution module, used to send the execution instruction to the target medical system, wherein the instruction includes a resource scheduling path and time constraint conditions; The closed-loop verification module is used to verify the execution result of the execution instruction and update the system parameters.

Images