CN-122000093-A - Cloud platform-based intelligent analysis system and method for medical supply chain

CN122000093ACN 122000093 ACN122000093 ACN 122000093ACN-122000093-A

Abstract

The invention discloses a cloud platform-based intelligent analysis system and a cloud platform-based intelligent analysis method for a medical supply chain, which relate to the technical field of computer Internet, and aim to solve the problems of private leakage, single public opinion modeling, delayed demand perception and the like of a medical supply chain crossing mechanism, realize high-safety and high-precision dynamic management of the medical supply chain, and improve the predictability and resource allocation efficiency of demand prediction by constructing a multi-source heterogeneous data federal perception network, carrying out differential privacy treatment, establishing a space-time coupled public opinion diffusion dynamic graph model to capture a propagation rule, calculating a multi-modal fused regional demand disturbance index based on a prediction result and an inventory state, and generating a collaborative replenishment strategy by using a distributed constraint optimization algorithm.

Inventors

ZHU LIWEI
LIU LIE
XIAO PEIQIANG
ZHU MIN
WANG FANQI

Assignees

南京易联阳光信息技术股份有限公司

Dates

Publication Date: 20260508
Application Date: 20260407

Claims (10)

1. A cloud platform-based intelligent analysis method for a medical supply chain is characterized by comprising the following steps of: Based on a cloud platform architecture, local edge computing nodes of hospitals, chain pharmacies and regional distribution centers are deployed, each node generates a disturbed sales volume characteristic value after applying noise according to the daily average sales volume of medicines of the node, and the compressed sales volume characteristic vector is uploaded to a cloud aggregation server to be aggregated to obtain a regional medicine consumption characteristic vector; Based on social media text, news reports and geographic positioning data, extracting disease keywords, drug names, geographic positions and time stamps, constructing a dynamic graph structure by taking regional demarcation nodes and cross-regional population flow intensity as side weights, and learning propagation paths and intensity attenuation rules of public opinion in space-time dimensions through a gate-controlled graph neural network; The public opinion transmission intensity is aligned with the drug consumption trend, and the regional demand disturbance index of each region is calculated by adopting a weighted fusion mode; When the demand disturbance index of any region continuously exceeds a preset threshold value for a plurality of time units and the safety stock of the subordinate node is lower than a warning line, an emergency response mechanism is started, the minimum transfer total cost is used as an objective function, and the optimal transfer scheme is obtained and then pushed to the management terminal.
2. The intelligent analysis method of the medical supply chain based on the cloud platform as set forth in claim 1, wherein the method for generating the sales feature value comprises the following steps: Obtaining identity information of medicines and forming a medicine set, calculating daily average sales of the p-th medicine in a unit period, adding Laplacian noise to the daily average sales at a local edge calculation node to obtain a sales characteristic value of the p-th medicine, carrying out data compression on the sales characteristic value to generate a sales characteristic vector, uploading the sales characteristic vector to a cloud platform, and carrying out arithmetic average on the sales characteristic value of the p-th medicine from N nodes to obtain the medicine consumption characteristic vector of the p-th medicine in a region.
3. The intelligent analysis method of the medical supply chain based on the cloud platform as set forth in claim 1, wherein the construction of the dynamic graph structure comprises the following steps: Acquiring disease keywords, drug names, geographical region codes and timestamps to form four-tuple; Calculating edge weights between any two nodes based on the number of times of personnel flow between two places in unit time, and establishing a directed edge set; The state updating rule of the gating graph neural network is that for the hidden state of any node at the time t0, the hidden state of any node at the time t0+1 is generated by fusing the current node state and the weighting state of the neighbor node through a gating circulation unit, wherein the contribution of the neighbor node is weighted by the corresponding edge weight.
4. The intelligent analysis method of the medical supply chain based on the cloud platform as set forth in claim 1, wherein the calculating of the regional demand perturbation index comprises: The method comprises the steps of carrying out standardization processing on regional medicine consumption characteristic vectors to obtain trend scores, inputting node hiding states output by a gate control graph neural network into a single-layer fully-connected neural network, outputting regional flow intensity scores normalized to a preset interval, and carrying out weighted calculation on the trend scores and the regional flow intensity scores to obtain regional demand disturbance indexes.
5. The intelligent analysis method of the medical supply chain based on the cloud platform as set forth in claim 1, wherein the local replenishment strategy comprises: The regional demand disturbance index is amplified proportionally according to a disturbance amplification factor to obtain a disturbance reference value, a predicted value of the regional total demand is obtained by multiplying the disturbance reference value by the historical contemporaneous average demand, the disturbance amplification factor takes a value in a preset range according to the degree of shortage of the medicine, a mean value and a standard deviation are calculated based on a plurality of predicted daily demand sequences in the future, a dynamic safety stock quantity is calculated by combining a service level factor and a predicted period length to serve as a local replenishment triggering threshold, and a replenishment flow is started when the local stock is lower than the threshold.
6. The intelligent analysis method of the medical supply chain based on the cloud platform as set forth in claim 1, wherein the response method of the emergency response mechanism comprises the following steps: the demand perturbation index for any region exceeds the response threshold for three consecutive time units, and the current inventory of hospital or pharmacy nodes within the region exceeding the percentage threshold is less than half of the safe inventory; The constraint conditions of the objective function comprise that the total allocation amount of each supplier does not exceed the adjustable inventory of the supplier, the total receiving amount of each demand is not lower than the predicted shortfall amount, all allocation amounts are non-negative, and the obtained optimal allocation scheme comprises an allocation party, medicine codes, quantity and predicted delivery time and is pushed to a management terminal through an application program interface.
7. The cloud platform-based intelligent analysis system for the medical supply chain is used for executing the cloud platform-based intelligent analysis method for the medical supply chain, and is characterized by comprising a federal learning coordination module, a space-time diagram construction module, a joint reasoning module, an operation optimization module and an instruction execution module; the federal learning coordination module is used for managing model training periods, parameter encryption uploading and global model distribution of each participant client; The space-time diagram construction module is used for integrating geographic information, population flow, logistics track and public opinion data, and constructing and maintaining a dynamically updated multi-relation diagram structure; the combined reasoning module is used for fusing the federal learning output requirement and a disturbance signal propagated by a space-time diagram to generate regional requirement prediction; the operation planning optimization module is used for solving an optimal medicine allocation and distribution scheme based on a demand prediction result and resource constraint; the instruction execution module is used for carrying out rule verification on the optimization scheme and transmitting the rule verification to the distribution scheme execution system.
8. The intelligent analysis system of a medical supply chain based on a cloud platform as set forth in claim 7, wherein said federal learning coordination module comprises a local feature extraction unit and a security aggregation unit; the local feature extraction unit is used for performing data desensitization and vector coding on each participant side, and the security aggregation unit is used for performing noise adding aggregation on model parameters by adopting a differential privacy mechanism on a central server side.
9. The cloud platform-based intelligent analysis system of a medical supply chain of claim 7, wherein the space-time diagram construction module comprises a node attribute management unit and a weight management unit, and the joint reasoning module comprises a disturbance detection unit and a diagram propagation calculation unit; The node attribute management unit is used for maintaining basic static attributes and dynamic consumption indexes of each region; The weight management unit is used for calculating edge weights among regional nodes according to population migration records; the disturbance detection unit is used for monitoring public opinion enthusiasm and abnormal diagnosis indexes of each node; the graph propagation calculation unit is used for executing information transfer and state update based on the graph neural network.
10. The intelligent analysis system of the medical supply chain based on the cloud platform as set forth in claim 7, wherein said operation optimization module comprises a constraint modeling unit and a solution scheduling unit; The constraint modeling unit is used for converting inventory, capacity and policy requirements into mathematical constraints, and the solution scheduling unit is used for calling the optimization solver to generate a scheduling scheme meeting aging requirements.

Description

Cloud platform-based intelligent analysis system and method for medical supply chain Technical Field The invention relates to the technical field of computer Internet, in particular to a cloud platform-based intelligent analysis system and method for a medical supply chain. Background Along with the acceleration of the digitization process of the medicine supply chain, the intelligent analysis method based on the cloud platform has great potential in improving the medicine circulation efficiency and emergency response capability. The current mainstream public opinion driven demand prediction method still has significant limitations. For example, patent application CN120875749a discloses a method and a system for analyzing drug inventory requirement based on sales data, wherein public opinion influence is estimated by co-occurrence frequency of pharmacy sales data and drug names in text, but pharmacy sales records relate to enterprise business confidentiality and personal privacy, and real-time sharing across institutions is difficult to realize in practice, so that drug consumption estimation is distorted. Meanwhile, semantic association is measured only by depending on the co-occurrence relation of diseases and medicines, and the diffusion dynamics characteristics of public opinion in the space-time dimension cannot be described, so that the perception of disturbance on regional requirements is delayed and one-sided. Therefore, there is a need for a cloud platform-based intelligent analysis method for a medical supply chain, which can construct a federal perception mechanism with multi-body cooperation on the premise of strictly protecting sensitive data of each medicine provider, fuse multi-mode heterogeneous data to accurately describe region demand mutation under the driving of public opinion, and accordingly realize privacy-safe dynamic inventory optimization and cross-mechanism emergency dispatch. Disclosure of Invention The invention aims to provide a cloud platform-based intelligent analysis system and method for a medical supply chain, which are used for solving the problems in the prior art. In order to achieve the aim, the invention provides the technical scheme that the intelligent analysis method of the medical supply chain based on the cloud platform comprises the following specific steps of; Step 1, constructing a multi-source heterogeneous data federation perception network, namely deploying local edge computing nodes oriented to hospitals, chain pharmacies and regional distribution centers under a cloud platform architecture, wherein each node only uploads a medicine consumption characteristic vector subjected to differential privacy disturbance treatment to a cloud aggregation server, and does not transmit an original sales record; Step 2, constructing a space-time coupled public opinion diffusion dynamic graph model, namely extracting disease keywords, medicine entities, geographic positions and time stamp quadruples based on social media texts, news reports, geographic positioning sign-in data and medicine image recognition results, constructing a dynamic graph structure with regional demarcation nodes and cross-regional population flow intensity as side weights, and learning a propagation path and intensity attenuation law of public opinion in space-time dimension through a gating graph neural network; Step 3, calculating a multi-mode fused regional demand disturbance index, namely aligning the public opinion transmission intensity output in the step 2 with the regional medicine consumption trend obtained by aggregation in the step 1, and calculating the regional demand disturbance index of each region by adopting a weighted fusion formula; Step 4, executing collaborative inventory optimization under privacy protection, namely running a safe multiparty calculation protocol locally by each medicine provider based on a demand disturbance index, and jointly solving a linear programming model with constraint to generate a local replenishment strategy which meets the service level requirement and minimizes the total holding cost, wherein only encrypted gradient information is exchanged in the process, and the respective inventory level and cost parameters are not revealed; And 5, executing cross-mechanism emergency cooperative scheduling simulation, namely starting an emergency response mechanism when a demand disturbance index of any region continuously exceeds a preset threshold value for a plurality of time units and the safety stock of nodes which belong to a preset proportion above is lower than a warning line, solving a minimum cost flow problem based on an operation optimizing engine embedded in a cloud platform, and pushing the optimal scheduling scheme to a management terminal after constraint conditions including supply capacity, a demand gap and non-negative scheduling quantity are obtained for the objective function to minimize the total scheduling cost. The construction process