Hospital pharmacy competitiveness evaluation method and system based on multi-source data fusion

Abstract

The invention relates to the technical field of medical data analysis and business intelligence, and discloses a hospital pharmacy competitiveness evaluation method and system based on multi-source data fusion, and the method comprises the following steps: collecting and processing multi-source data, and constructing a market entity feature vector; constructing an initial market causal atlas based on the feature vectors; initializing a digital twinning environment, and setting a dynamic behavior rule of a simulation entity; running simulation to obtain a reference competitiveness evolution trajectory; real data is periodically acquired to calculate simulation deviation; and if the deviation is greater than a preset threshold value, correcting the causal atlas and updating the simulation entity rule. The system comprises a data processing and feature construction module; a market causal relationship graph construction module; a market digital twin environment initialization module; a competitiveness evolution simulation module; a simulation reality deviation calculation module; and a causal atlas correction module. According to the method, the dynamic causal model is constructed by fusing multi-source data, so that accurate and prospective competitiveness evaluation is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical data analysis and business intelligence, in particular to a hospital pharmacy competitiveness evaluation method and system based on multi-source data fusion. BACKGROUND

[0002] In the medical and health field, market competition is becoming increasingly fierce. For hospitals and pharmacy market entities, accurate competitiveness analysis is the basis for formulating scientific decisions and strategic planning.

[0003] Currently, the analysis method commonly used in the industry is based on a statistical model to process single-source data of medical institutions and pharmaceutical companies to obtain preliminary business insights such as market share at a specific point in time.

[0004] However, such methods have significant limitations. First, relying on single-source data cannot fully reflect the multi-dimensional market pattern. Second, statistical analysis can only reveal the correlation between variables, but cannot explain the causal transmission mechanism behind it. Therefore, the analysis conclusions derived therefrom are often static and one-sided, making it difficult to predict the future dynamic evolution of the market, leading to blind spots in the formulated competition strategy, and thus affecting the final effectiveness of the decision.

[0005] Therefore, the present application proposes a hospital pharmacy competitiveness evaluation method and system based on multi-source data fusion to solve the problems of the prior art. SUMMARY

[0006] The purpose of the present application is to provide a hospital pharmacy competitiveness evaluation method and system based on multi-source data fusion, which solves the problem of the prior art that relying on single-source data for analysis leads to an inability to fully and dynamically reflect the market competition pattern, thereby affecting the effectiveness of market strategy decision-making.

[0007] To achieve the above purpose, the present application is implemented by the following technical solutions: The present application provides, in a first aspect, a hospital pharmacy competitiveness evaluation method based on multi-source data fusion, comprising the following steps: collecting multi-source heterogeneous data, and performing cleaning, format conversion and standardization processing on the collected multi-source heterogeneous data to form fusion data, and determining hospitals or pharmacies as market entities based on the fusion data, and constructing a market entity feature vector for each market entity; Based on the market entity feature vector, an initial market causal relationship graph is constructed, which represents the causal relationship between market variables; Based on the initial market causal relationship graph, a market digital twin environment is initialized, which contains a plurality of simulation entities, each of which corresponds to a market entity, and the dynamic behavior rules of the simulation entities are determined by the initial market causal relationship graph; running an evolutionary simulation in the market digital twin environment, to obtain a competitive evolution trajectory of each of the simulation entities corresponding to a market entity under a benchmark scenario; periodically obtaining market observation data in the real world, and calculating a simulation reality deviation between a simulation state of the market digital twin environment and the market observation data; determining whether the simulation reality deviation is greater than a preset threshold, and if so, modifying the initial market causal relationship graph using the simulation reality deviation to obtain an updated market causal relationship graph, and updating the dynamic behavior rules of the simulation entities using the updated market causal relationship graph.

[0008] In a possible implementation, the initial market causal relationship graph is a directed acyclic graph, and the directed acyclic graph includes a node set and a directed edge set, where each node corresponds to a market variable, and each directed edge represents a direct causal influence between two market variables.

[0009] In a possible implementation, the step of obtaining a competitive evolution trajectory of each of the simulation entities corresponding to a market entity under a benchmark scenario includes: running the evolutionary simulation in a discrete time step manner; and at each discrete time step, dynamically calculating a competitive score of each of the market entities, where the calculation of the competitive score takes a state vector of the simulation entity at a current time step as input, and a calculation formula of the competitive score is as follows: wherein C i (i, t) is a competitive score of the i-th market entity at simulation time t; s i,j (i, t) is a j-th state variable in a state vector of the i-th simulation entity at simulation time t; w j is a preset weight of the j-th state variable; and k is a total number of state variables in the state vector. Subsequently, the competitive scores of all the market entities at all the discrete time steps are combined to form the competitive evolution trajectory.

[0010] In a possible implementation, the method further includes: receiving a market intervention strategy specified by a user; modifying a parameter in the market digital twin environment or a dynamic behavior rule of the simulation entities based on the market intervention strategy; running an evolutionary simulation again in the modified market digital twin environment to obtain a competitive evolution trajectory under a strategy scenario; and quantifying a net effect of the market intervention strategy.

[0011] In a possible implementation, the quantification of the net influence of the market intervention strategy is calculated based on the competitive evolution trajectory in the benchmark scenario and the competitive evolution trajectory in the strategy scenario, and the calculation formula of the net influence is: AC i =C' i (T s )-C i (T s ); In the formula, AC i is the net influence of the market intervention strategy on the i-th market entity; C i ′ (T s ) is the competitive score of the i-th market entity at the end time T s in the strategy scenario; C i (T s ) is the competitive score of the i-th market entity at the end time T s in the benchmark scenario.

[0012] In a possible implementation, the step of calculating the simulation reality deviation of the simulation state of the market digital twin environment from the market observation data comprises: obtaining the simulation state of the market digital twin environment at a real-world time point corresponding to the market observation data; based on the simulation state and the market observation data, calculating the simulation reality deviation, and the calculation formula of the simulation reality deviation is: In the formula, D(t real ) is the simulation reality deviation at the real-world time point t real ; t real is the real-world time point at which the market observation data is obtained; C j is the simulation state value of the j-th market variable in the simulation state; C norm is the market observation data value of the j-th market variable in the market observation data; k is the total number of market variables used for calculating the deviation; ω max is the deviation sensitivity weight of the j-th market variable.

[0013] In a possible implementation, the step of modifying the initial market causal relationship graph using the simulation reality deviation comprises: identifying a high-deviation market variable according to the simulation reality deviation; tracing an upstream causal chain of the high-deviation market variable in the market causal relationship graph; and modifying the topology or parameters of the market causal relationship graph according to the upstream causal chain and the simulation reality deviation.

[0014] The second aspect of the present application provides a hospital pharmacy competitiveness evaluation system based on multi-source data fusion, comprising: A data processing and feature construction module is configured to collect multi-source heterogeneous data, clean, format convert and standardize the collected multi-source heterogeneous data to form fusion data, determine hospitals or pharmacies as market entities based on the fusion data, and construct a market entity feature vector for each market entity. A market causal relationship graph construction module is configured to construct an initial market causal relationship graph based on the market entity feature vector, wherein the initial market causal relationship graph represents the causal relationship between market variables. A market digital twin environment initialization module is configured to initialize a market digital twin environment based on the initial market causal relationship graph, wherein the market digital twin environment includes a plurality of simulation entities, each simulation entity corresponds to a market entity, and the dynamic behavior rules of the simulation entities are determined by the initial market causal relationship graph. A competitiveness evolution simulation module is configured to run an evolution simulation in the market digital twin environment to obtain the competitiveness evolution trajectory of each market entity in a benchmark scenario. A simulation-reality deviation calculation module is configured to periodically obtain market observation data of the real world, and calculate the simulation-reality deviation between the simulation state of the market digital twin environment and the market observation data. A causal graph correction module is configured to determine whether the simulation-reality deviation is greater than a preset threshold, and if so, correct the initial market causal relationship graph using the simulation-reality deviation to obtain an updated market causal relationship graph, and update the dynamic behavior rules of the simulation entities using the updated market causal relationship graph.

[0015] The third aspect of the present application provides an electronic device, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor executes the computer program to implement the method of any one of the embodiments of the first aspect of the present application.

[0016] The fourth aspect of the present application provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the method of any one of the embodiments of the first aspect of the present application.

[0017] In summary, the present application includes at least one of the following beneficial technical effects: 1.The application constructs a comprehensive feature vector for each market entity by collecting and fusing multi-source heterogeneous data, overcoming the one-sidedness problem caused by relying on single-source data. This design makes the competitiveness evaluation of hospitals and pharmacies based on a multi-source data fusion that is closer to the real market, thereby significantly improving the objectivity and accuracy of the evaluation results and providing a solid data foundation for market strategy formulation.

[0018] 2.The application converts static evaluation into dynamic trajectory deduction by constructing a market digital twin environment and running evolutionary simulation. Users can simulate market intervention strategies in the digital environment and predict their potential impact. This multi-source data fusion-based analysis method enables the evaluation of the competitiveness of hospitals and pharmacies to be forward-looking, allowing the quantification of the net impact of different strategies and providing the possibility of scientific argumentation for decision-makers before resource investment.

[0019] 3.The application establishes a closed-loop feedback and correction mechanism for simulation reality deviation, enabling the model to self-calibrate according to real market data. This evaluation method for the competitiveness of hospitals and pharmacies is dynamically evolving, and the causal graph constructed based on multi-source data fusion can continuously approximate the real market logic, thereby ensuring the high fidelity and reliability of the evaluation system during long-term operation and avoiding the accumulation of biases caused by model solidification. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 A multi-source data fusion-based hospital and pharmacy competitiveness evaluation system architecture diagram for an embodiment of the application; Figure 2 A multi-source data fusion-based hospital and pharmacy competitiveness evaluation method flowchart for an embodiment of the application; Figure 3 A strategy net impact evaluation schematic diagram for an embodiment of the application; Figure 4 An electronic device structure schematic diagram of the application.

[0021] Among them, 10, data processing and feature construction module; 20, market causal relationship graph construction module; 30, market digital twin environment initialization module; 40, competitiveness evolutionary simulation module; 50, simulation reality deviation calculation module; 60, causal graph correction module; 70, electronic device; 71, processor; 72, memory; 73, storage medium. DETAILED DESCRIPTION

[0022] The following will be described in detail with reference to the accompanying drawings Figure 1 -APPENDIX Figure 4 The application will be further described in detail.

[0023] Refer to the accompanying drawings Figure 1 ,Figure 1 is a hospital pharmacy competitiveness evaluation system architecture based on multi-source data fusion according to an embodiment of the application.

[0024] Embodiment one: The embodiment provides a hospital pharmacy competitiveness evaluation system based on multi-source data fusion, which establishes a complete technical architecture integrating data processing, causal modeling, simulation deduction, deviation calculation and closed-loop correction.

[0025] The system can comprise a data processing and feature construction module 10, a market causal relationship graph construction module 20, a market digital twin environment initialization module 30, a competitiveness evolution simulation module 40, a simulation reality deviation calculation module 50 and a causal graph correction module 60.

[0026] The data processing and feature construction module 10 is used to collect multi-source heterogeneous data, and performs cleaning, format conversion and standardization processing on the collected data to form fusion data, and determines a hospital or pharmacy as a market entity based on the fusion data, and constructs a market entity feature vector for each market entity. The output of this module provides standardized data input for the subsequent modules.

[0027] The market causal relationship graph construction module 20 is connected with the output end of the data processing and feature construction module 10, and is used to construct an initial market causal relationship graph representing the causal relationship between market variables based on the received market entity feature vector.

[0028] The market digital twin environment initialization module 30 is connected with the output end of the market causal relationship graph construction module 20, and is used to initialize the market digital twin environment based on the initial market causal relationship graph. The environment contains a plurality of simulation entities, each simulation entity corresponds to a market entity, and the dynamic behavior rules of the simulation entities are determined by the initial market causal relationship graph.

[0029] The competitiveness evolution simulation module 40 is connected with the market digital twin environment initialization module 30, and is used to run evolution simulation in the market digital twin environment to obtain the competitiveness evolution trajectory of each market entity corresponding to the simulation entity in the benchmark scenario.

[0030] The simulation reality deviation calculation module 50 is used to periodically obtain market observation data of the real world, and receives the simulation state from the competitiveness evolution simulation module 40, so as to calculate the simulation reality deviation between the simulation state of the market digital twin environment and the market observation data.

[0031] The causal graph correction module 60 is connected with the output end of the simulation reality deviation calculation module 50, and is used for judging whether the simulation reality deviation is greater than a preset threshold. If the judgment result is yes, the simulation reality deviation is used to correct the market causal relationship graph to obtain an updated market causal relationship graph. The updated graph is then used to update the dynamic behavior rules of the simulation entities in the market digital twin environment initialization module 30, thereby forming a feedback correction closed loop.

[0032] Referring to the drawings Figure 2 , Figure 2 is a hospital pharmacy competitiveness evaluation method based on multi-source data fusion according to an embodiment of the application.

[0033] Embodiment two: The hospital pharmacy competitiveness evaluation method based on multi-source data fusion provided in this embodiment details the specific steps followed by the system in executing its functions.

[0034] The method can include the following steps: S1, collecting multi-source heterogeneous data, and performing cleaning, format conversion and standardization processing on the collected multi-source heterogeneous data to form fusion data, and determining a hospital or a pharmacy as a market entity based on the fusion data, and constructing a market entity feature vector for each market entity; S2, constructing an initial market causal relationship graph based on the market entity feature vector, the initial market causal relationship graph representing the causal relationship between market variables; S3, initializing a market digital twin environment based on the initial market causal relationship graph, the market digital twin environment containing a plurality of simulation entities, each simulation entity corresponding to a market entity, and the dynamic behavior rules of the simulation entities being determined by the initial market causal relationship graph; S4, running an evolutionary simulation in the market digital twin environment to obtain the competitiveness evolution trajectory of each market entity corresponding to the simulation entity under a benchmark scenario; S5, periodically obtaining market observation data in the real world, and calculating the simulation reality deviation between the simulation state of the market digital twin environment and the market observation data; S6, judging whether the simulation reality deviation is greater than a preset threshold, and if so, using the simulation reality deviation to correct the initial market causal relationship graph to obtain an updated market causal relationship graph, and using the updated market causal relationship graph to update the dynamic behavior rules of the simulation entities.

[0035] The first step of the method is S1, that is, to perform data processing and market entity feature construction.

[0036] In the present embodiment, the specific execution process of step S1 is described in detail. First, the multi-source heterogeneous data is collected by the data processing and feature construction module 10. The multi-source heterogeneous data is different in both source and structure, and can specifically include: structured diagnosis and treatment data from the hospital internal information system (HIS), such as medical order records, drug prescription amounts; structured transaction data from the point of sale (POS) system of a chain drugstore, such as drug sales and inventory turnover rate; semi-structured data from the geographic information system (GIS), such as the latitude and longitude coordinates of the hospital or drugstore, and the distribution of surrounding transportation facilities; and unstructured text data from Internet public platforms, such as online comments by patients on hospital services, discussions on specific drugstores on social media, etc.

[0037] After obtaining the above data, the collected multi-source heterogeneous data is cleaned, format-converted and standardized to form fusion data for subsequent steps. Data cleaning aims to handle inconsistencies and noise in the original data. For example, for missing values in numerical data, the median of the feature column is used for filling; for data outliers, the Z-score method can be used for identification, i.e. calculating the Z-score value of each data point, and removing or adjusting the data points with an absolute value of Z-score greater than a preset threshold (e.g. 3).

[0038] Data format conversion aims to unify the data from different sources into a standard format. For example, date strings in different formats (such as 2025-07-09, 07 / 09 / 2025) are uniformly converted to the standard ISO 8601 date format; text type classification data (such as hospital level three A, two B) is converted to the corresponding numerical code (such as 31, 22).

[0039] Data standardization processing aims to eliminate the influence caused by different dimensions between different market variables. For example, the min-max normalization method can be used to linearly map numerical data to the [0, 1] interval, and the calculation formula is: In the formula, X norm is the normalized numerical value; X is the original numerical value; X max and X min are the maximum and minimum values of the variable in the data set, respectively.

[0040] After the above processing and fusion data are formed, the market entity is determined based on the data. For example, taking the hospital practice license number or the unique business registration number of the drugstore as the unique identifier, each independent hospital or drugstore is defined as a market entity.

[0041] Finally, a multi-dimensional market entity feature vector V is constructed for each market entity i . The feature vector V i = [v i,1 , v i,2 ,..., v i,n ], where each dimension v i,j corresponds to a processed market variable, for example, v i,1 is the standardized geographical location convenience score, v i,2 is the average daily prescription amount in the last three months, and v i,3 is the average user evaluation score extracted from web text.

[0042] By performing step S1, the dispersed and heterogeneous raw data is transformed into regular and quantifiable market entity feature vectors, providing a comprehensive and objective data basis for subsequent construction of causal relationship graphs and competitive assessment, thereby overcoming the inherent one-sidedness problem when relying on a single data source for analysis, and improving the accuracy and reliability of the assessment results.

[0043] After performing step S1, the method flow enters step S2, i.e., constructing an initial market causal relationship graph based on the market entity feature vector.

[0044] In this embodiment, the specific execution process of step S2 is described in detail. The market causal relationship graph construction module 20 receives the set of market entity feature vectors V i constructed for all market entities in step S1 as input data. The construction of this graph aims to discover and represent the inherent and directional causal relationship between market variables from high-dimensional feature data, rather than simply statistical correlation.

[0045] The initial market causal relationship graph is structurally defined as a directed acyclic graph (DAG), denoted as G = (N, E). The directed acyclic graph G includes a node set N and a directed edge set E. Each node n j ∈ N in the node set N corresponds to a market variable, which is a dimension in the market entity feature vector V i , such as geographical location convenience or average daily prescription amount. Each directed edge e jk ∈ E (denoted as n j → n k ) in the directed edge set E represents the direct causal influence between two market variables, i.e., the market variable corresponding to node n j is the direct cause of the market variable corresponding to node n k .

[0046] The method for constructing the initial market causal graph G can use a constraint-based causal discovery algorithm, such as the PC algorithm (Peter-Clark algorithm), combined with a domain expert knowledge base. The process mainly includes the following sub-steps: first, all market variables are nodes, and a fully connected undirected graph is constructed.

[0047] Secondly, through conditional independence test, the edges in the undirected graph are iteratively removed. For any pair of nodes (variables) X and Y connected by an edge in the graph, the conditional independence of X and Y given a node set S is systematically tested, denoted as (X XY·S The conditional independence test can select specific methods according to variable types. For continuous variables, the partial correlation coefficient can be calculated. If the partial correlation coefficient p of X and Y given S is greater than the preset significance level a (for example, a = 0.05), then X and Y are considered to be conditionally independent given S, and the edge between them is removed. The test starts from an empty set S and gradually increases the number of nodes in S until no more edges can be removed, thus obtaining the skeleton of the graph.

[0048] Finally, the undirected edges in the skeleton graph are oriented. This step is based on specific orientation rules, such as identifying V-type structures (V-structure). If there are three nodes X, Y, and Z, where X and Y have no edge between them, but they both have an edge with Z, and when looking for a separation set S, Z is not in the separation set S XY , then the two edges are oriented as X→Z←Y. Based on the oriented V-type structure, other rules (such as avoiding the generation of new V-type structures or directed loops) are applied to the remaining edges to orient them as much as possible, and finally a directed acyclic graph is formed. For example, for brand reputation (X), drug price (Y), and monthly sales (Z), the algorithm may find that (X

[0049] By performing step S2, the invention explicitly and structurally expresses the complex relationships originally implied in the data as a market causal graph with clear causal logic. This not only provides the basic logical rules for subsequent dynamic simulation, but more importantly, it reveals the internal mechanism of market operation.

[0050] After performing step S2, the method continues to perform step S3, which initializes the market digital twin environment based on the initial market causal graph.

[0051] In the present embodiment, the specific execution process of step S3 is described in detail. First, the market digital twin environment is defined, which refers to a computer-constructed computing environment that maps entities, relationships, and dynamic processes with the real-world market. This environment provides a basic running framework for subsequent evolution simulation, including time stepping mechanism and interaction rules. The market digital twin environment initialization module 30 initializes the environment based on the initial market causal relationship graph G=(N, E) constructed in step S2.

[0052] The initialization process first includes: for each market entity determined in step S1, a one-to-one corresponding simulation entity is generated in the environment. Here, the simulation entity is defined as an independent, structured software entity that encapsulates internal state and behavior rules as the functional counterpart of its corresponding real-world market entity (hospital or pharmacy) in the digital twin environment. The initial internal state of each simulation entity at simulation time zero (t=0) is determined by the market entity feature vector V i is assigned.

[0053] The core is that the dynamic behavior rules of the simulation entity are completely determined by the initial market causal relationship graph G. Specifically, the dynamic behavior rules of the simulation entity are defined as a set of state update functions. For any non-root node n k (i.e., a market variable with a parent node) in the initial market causal relationship graph G, the state variable s i,k of its corresponding simulation entity at the next simulation time t+1 is determined by a function f k . The input of this function is the value of all parent node variables of the simulation entity at the current simulation time t, and its mathematical expression is: s i,k (t+1)=f k ({s i,j (t)|n j →n k ∈E}); where s i,k (t+1) is the value of the kth state variable of the ith simulation entity at time t+1; {s i,j (t)|n j →n k ∈E} represents the set of state variable values corresponding to all parent nodes n k pointing to node n j in the initial market causal relationship graph G at time t.

[0054] In the present embodiment, the state update function f k can be set as a weighted linear model to quantify causal effects: where w jk is the weight of the directed edge from node n j to node n k , which represents the causal influence strength of the parent variable s i,j on the child variable s i,k , whose initial value can be set by regression analysis or expert knowledge; b k is the baseline value or intercept term of the variable s i,k ; ∈ k is the random disturbance term, which is used to simulate the uncertainty factors in the market. For the root nodes in the graph (i.e., market variables without parent nodes), their state update rules can be set to keep their initial values unchanged or vary according to an exogenous time series.

[0055] By performing step S3, the present application converts the static, structured market causal relationship graph constructed in step S2 into a computable, deductive dynamic system. Each simulation entity is not only a digital mirror of the real market entity, but its behavior logic also strictly follows the causal laws discovered from the data.

[0056] After performing step S3, the method continues to perform step S4, i.e., running evolutionary simulation in the market digital twin environment to obtain the competitive evolution trajectory of each market entity corresponding to the simulation entity under the baseline scenario.

[0057] In this embodiment, the specific execution process of step S4 is described in detail. The competitive evolution simulation module 40 runs evolutionary simulation in the market digital twin environment initialized in step S3 in a discrete time step manner. The discrete time step can be a preset time unit, such as one day, one week, or one month. The simulation starts from the initial time t = 0 and proceeds to the preset simulation end time T according to the time step. At each simulation time t, all simulation entities in the environment update their internal state vectors simultaneously according to the dynamic behavior rules assigned to them in step S3.

[0058] At each discrete time step t, the system dynamically calculates the competitive score of each market entity. This calculation takes the state vector of the simulation entity corresponding to the market entity at the current time step t as input. The calculation formula of the competitive score is: where C i (t) is the competitive score of the i-th market entity at simulation time t, which is a comprehensive quantitative indicator representing the market comprehensive competitiveness level of the entity at that time.

[0059] s i,j(t) is the jth state variable in the state vector of the ith simulation entity at simulation time t, for example, s i,1 (t) can be the simulated daily passenger flow, s i,2 (t) can be the simulated drug sales. The values of these state variables are calculated according to the state update function defined in step S3.

[0060] w j is the preset weight of the jth state variable, k is the total number of state variables in the state vector.

[0061] Finally, the competitiveness scores C i (t) of each market entity calculated at all discrete time steps (from t = 0 to T) are combined in chronological order to form the competitiveness evolution trajectory of the market entity. The trajectory is a time series, which is specifically represented as: {C i (0), C i (1), C i (2),..., C i (T)} which intuitively shows the dynamic changes of the competitiveness of the market entity over time under the benchmark scenario (i.e. natural evolution without external intervention).

[0062] By performing step S4, the present application converts the multi-dimensional static characteristics and dynamic causal rules of the market entity into an observable and quantifiable competitiveness time series. Thus, the evaluation of competitiveness changes from a static numerical evaluation based on a single time point to a quantitative presentation of the dynamic evolution process within a complete time period. More importantly, the competitiveness evolution trajectory under the benchmark scenario generated thereby provides precise simulation-side data for the simulation-reality deviation calculation in subsequent step S5, is an indispensable link to realize the closed-loop correction function of the entire system, and also provides a basic reference system for subsequent strategy deduction.

[0063] After performing step S4, the method continues to perform step S5, which periodically acquires market observation data of the real world and calculates the simulation-reality deviation between the simulation state of the market digital twin environment and the market observation data.

[0064] In this embodiment, the specific execution process of step S5 is described in detail. The simulation-reality deviation calculation module 50 is used to automatically perform this step with a fixed time period (for example, monthly or quarterly). First, the simulation-reality deviation calculation module 50 acquires market observation data of the real world. The market observation data refers to the values of the state variables of the market entity at a specific real-world time point t real, which can be objectively obtained from the outside, and is used to calibrate the quantitative index of the real state of the market. In this embodiment, the data can specifically include, but not limited to, one or a combination of the following: Financial data: specific data obtained from the financial statements publicly released or internally provided by market entities (such as hospitals, pharmacies), such as total drug sales revenue, gross profit margin, etc. in a specific quarter or month.

[0065] Operation data: data obtained through internal information systems or third-party monitoring services, such as the number of outpatient visits, total prescriptions, average length of stay, or customer transaction count in a pharmacy within a specific time period.

[0066] Market share data: sales or sales volume data of each market entity in a specific region published by professional market research institutions (such as medical and health information service providers).

[0067] Publicly quantifiable reputation data: from mainstream network platforms at time point t real Aggregated quantitative score, such as the average user evaluation score of a specific hospital or pharmacy on a certain day.

[0068] At the same time, the module obtains the simulation state of the market digital twin environment at the real world time point t real corresponding to the market observation data. This means that from the simulation data running in step S4, the simulation state value matching the simulation time t real is extracted.

[0069] After obtaining the simulation state and market observation data, the module calculates the simulation reality deviation based on the two sets of data. The calculation formula of the simulation reality deviation is In the formula, D(t real ) is the simulation reality deviation at real world time point t real . This value is a scalar that quantifies the overall accuracy of the simulation system at this time.

[0070] t real is the real world time point when the market observation data is obtained, for example, July 31, 2025.

[0071] is the simulation state value of the jth market variable in the simulation state. This value is the output value of the digital twin environment for this market variable (such as total market sales) at the simulation time step corresponding to t real .

[0072] is the market observation data value of the jth market variable in the market observation data. The value is the actual occurrence value of the market variable (e.g., market total sales) in the real world.

[0073] k is the total number of market variables used to calculate the deviation. The variable set selected for comparison must be available in both the simulation environment and the real world and be consistently defined.

[0074] ω j is the deviation sensitivity weight of the jth market variable, and

[0075] By performing step S5, the present application establishes a quantitative feedback channel connecting the simulation world and the real world. Instead of relying on subjective judgment to evaluate the pros and cons, it measures the consistency between the digital twin environment and the real market through an accurate and objective deviation value D(t real ), which provides a direct and quantifiable decision basis for whether to start the correction mechanism and how to correct in subsequent step S6, and is the core prerequisite for realizing system adaptation and self-optimization, thereby ensuring the effectiveness and accuracy of the evaluation method in long-term operation.

[0076] After performing step S5, the method continues to perform step S6, i.e., adaptive correction of the causal graph.

[0077] In this embodiment, the specific execution process of step S6 is described in detail. The causal graph adaptive correction module 60 first compares the simulation reality deviation D(t real ) calculated in step S5 with the preset deviation threshold D threshold . The setting of this threshold aims to ensure that the correction mechanism is only triggered when there is a significant deviation between the simulation model and the real world, thereby avoiding excessive correction caused by random disturbances in the market.

[0078] If D(t real )>D threshold , the system starts the correction process of the initial market causal relationship graph. The process specifically includes the following sub-steps: First, the high-deviation market variables are identified according to the simulation reality deviation. In the deviation calculation formula in step S5, the overall deviation D(t real ) is composed of the deviation components of each market variable. Therefore, this step identifies one or more market variables that contribute most to the overall deviation by checking the size of the weighted deviation square term of each variable j, which are the high-deviation market variables.

[0079] Secondly, trace the upstream causal chain of the high-biased market variable n k in the initial market causal graph G = (N, E). For example, if variable n k is identified as a high-biased variable, the system will start from n k and perform a reverse graph traversal in the initial market causal graph G to find all the parent nodes of n k (i.e. the direct causes of n k ), and then find the parent nodes of these parent nodes, and so on recursively until reaching the root nodes in the graph or reaching a preset trace depth. The set of nodes and edges thus obtained constitutes the upstream causal chain that leads to the bias of n k .

[0080] Finally, modify the topology or parameters of the initial market causal graph according to the upstream causal chain and the simulation reality bias.

[0081] Parameter modification: This is the preferred modification method. The system takes the observed value of the high-biased variable n k as the target, takes the variables in the upstream causal chain as the input, and recalibrates the parameters of the state update function f that determines the value (such as the weighted linear model described in step S3). For example, gradient descent can be used to minimize the bias as the optimization objective, iteratively update the relevant weights w jk and baseline values b k , and the update rule can be expressed as: where η is the learning rate.

[0082] Topology modification: If the bias cannot be reduced to an acceptable range through parameter modification alone, the system will consider that the structure of the initial graph may be incorrect. At this time, the system will perform a local causal discovery algorithm (such as the PC algorithm described in step S2) again for the subset of variables involved in the upstream causal chain using the newly added market observation data. This process may discover new causal relationships between original variables (i.e. add a directed edge) or find that an original causal relationship is no longer significant (i.e. remove a directed edge).

[0083] Through the above modification, an updated market causal graph can be obtained. Finally, the system uses this updated graph to update the dynamic behavior rules of the simulation entities defined in step S3, i.e. replaces the original state update function with the modified parameters or topology.

[0084] By performing step S6, the present application builds a closed-loop adaptive correction mechanism. This mechanism enables the market digital twin to dynamically learn from changes in the real world and continuously optimize its internal logic, ensuring consistency between the model and reality.

[0085] Referring to the accompanying drawings Figure 3 , the figure is a schematic diagram of the strategy net impact evaluation of an embodiment of the present application.

[0086] Embodiment three: This embodiment further describes how the present application can be applied to evaluate the effectiveness of a specific market intervention strategy based on embodiments one and two.

[0087] In this embodiment, when the system has completed steps from S1 to S6, i.e., it already has a digital twin environment that has been calibrated with real-world data and can accurately reflect market dynamics, the user can start an optional strategy evaluation process.

[0088] First, the user specifies a market intervention strategy. The market intervention strategy refers to a specific action plan proposed by the user (e.g., a hospital manager, a pharmacy operator, or a policy maker) to change the market structure or improve their competitiveness. In this embodiment, the strategy can be specifically: For a hospital entity: Set an increase of 20% in research funding or introduce 2 top experts in a specific field.

[0089] For a pharmacy entity: Set a 10% reduction in the retail price of major over-the-counter (OTC) drugs or launch a three-month online promotion campaign, expected to increase brand awareness by 15%.

[0090] Next, the system modifies the parameters in the market digital twin environment or the dynamic behavior rules of the simulated entities based on the market intervention strategy. This step is to convert abstract strategy concepts into computable operations in the digital twin environment. For example: If the strategy is to reduce the retail price of major OTC drugs by 10%, the system will locate the simulated entity corresponding to the pharmacy that initiated the strategy and adjust the initial value or relevant parameters in the update rule of the state variable s i,price drug price in its internal state vector. If the strategy is to increase research funding by 20%, this may affect the new therapy output rate variable. The system will modify the state update function f 新疗法产出率 inside the simulated entity, for example, by adjusting the weight parameter related to research funding to reflect the positive impact of increased input on output.

[0091] In the modified market digital twin environment, the system runs the evolutionary simulation again. This simulation process is similar to step S4, running from t=0 to a preset simulation end time T. Since the initial conditions or operating rules of the environment have been changed by the policy, this simulation will generate a completely new set of competitive evolution trajectories {C′} under the policy scenario. i (0),C′ i (1),...,C′ i (T s )_}.

[0092] Finally, the system quantifies the net impact of the market intervention strategy. The quantification of net impact is based on a comparative calculation of the competitiveness evolution trajectory under the baseline scenario obtained in Example 2 and the competitiveness evolution trajectory under the strategy scenario obtained in this example. The formula for calculating the net impact is: ΔC i =C′ i (T s )-C i (T s ); In the formula, ΔC i C′ represents the net impact of the market intervention strategy on the i-th market entity; i (T s Let be the market entity i at the end of the simulation under the strategy scenario. s Competitiveness score; C i (T s ) represents the i-th market entity at the end of the simulation under the baseline scenario. s The competitiveness score.

[0093] Through the process described in this embodiment, the present invention provides a risk-free and low-cost strategy simulation and evaluation capability. Users can pre-examine the potential positive and negative effects of a possible decision without incurring any real-world costs, and may even uncover unexpected systemic impacts (for example, a pharmacy's price reduction strategy may trigger a chain reaction from competitors, leading to a decline in the overall market's profit margin). This evaluation method, based on causal simulation and quantitative comparison, transforms strategic decision-making from relying on intuition and experience to relying on data-driven scientific deduction, greatly enhancing the predictability and scientific rigor of decisions and providing strong support for hospitals and pharmacies to formulate efficient competitive strategies in fierce market competition.

[0094] When the system completes the complete process of steps S1 to S6, that is, a calibrated digital twin environment capable of reliably reflecting market dynamics is constructed, the system will output a comprehensive competitiveness evaluation report for all market entities. The core result of the report is generated by dynamic and comprehensive quantitative calculation of the state vector of each simulation entity in the simulation environment. Specifically, at each time step in the simulation period, the system aggregates the simulation values of multiple key state variables (such as service quality, drug price, brand reputation, geographical location convenience, etc.) representing each simulation entity by weighting, obtains a comprehensive quantitative index, that is, a competitiveness score, and collects these scores to form a dynamic evolution trajectory.

[0095] Embodiment four: The application also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method as above when executing the computer program.

[0096] Embodiment five: The application also provides a storage medium, wherein the storage medium stores a computer program, and the computer program is executable on a processor to implement the method as above.

[0097] The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0098] Although the embodiments of the application have been shown and described, it is understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the application, and the scope of the application is defined by the appended claims and their equivalents.

Claims

1. A hospital pharmacy competitiveness evaluation method based on multi-source data fusion, characterized in that, The method comprises the following steps: Collecting multi-source heterogeneous data, and performing cleaning, format conversion and standardization processing on the collected multi-source heterogeneous data to form fusion data, and determining hospitals or pharmacies as market entities based on the fusion data, and constructing a market entity feature vector for each market entity; Based on the market entity feature vector, an initial market causal relationship graph is constructed, which represents the causal relationship between market variables; Based on the initial market causal relationship graph, a market digital twin environment is initialized, which includes a plurality of simulation entities, each of which corresponds to a market entity, and the dynamic behavior rules of the simulation entities are determined by the initial market causal relationship graph; Running an evolutionary simulation in the market digital twin environment to obtain the competitive evolution trajectory of each market entity in the benchmark scenario; Periodically obtaining market observation data in the real world, calculating the simulation reality deviation between the simulation state of the market digital twin environment and the market observation data; Judging whether the simulation reality deviation is greater than a preset threshold, if yes, modifying the initial market causal relationship graph using the simulation reality deviation to obtain an updated market causal relationship graph, and updating the dynamic behavior rules of the simulation entities using the updated market causal relationship graph.

2. The hospital pharmacy competitiveness evaluation method based on multi-source data fusion according to claim 1, characterized in that, The initial market causal relationship graph is a directed acyclic graph, which includes a node set and a directed edge set, wherein each node corresponds to a market variable, and each directed edge represents the direct causal influence between two market variables. 3.The hospital pharmacy competitiveness evaluation method based on multi-source data fusion of claim 1, characterized in that, The step of running an evolutionary simulation in the market digital twin environment to obtain the competitive evolution trajectory of each market entity in the benchmark scenario comprises: Running the evolutionary simulation in a discrete time step manner; At each discrete time step, dynamically calculating the competitive score of each market entity, wherein the calculation of the competitive score takes the state vector of the simulation entity at the current time step as input, and the calculation formula of the competitive score is: wherein C i (t) is the competitiveness score of the market entity i at simulation time t; s i,j (t) is the jth state variable in the state vector of the simulation entity i at simulation time t; w j is the preset weight of the jth state variable; k is the total number of state variables in the state vector; Combining the competitive scores of all market entities at all discrete time steps to form the competitive evolution trajectory. 4.The hospital pharmacy competitiveness evaluation method based on multi-source data fusion of claim 1, characterized in that, The method further comprises: Receiving a user-specified market intervention strategy; Modifying the parameters in the market digital twin environment or the dynamic behavior rules of the simulation entities based on the market intervention strategy, running an evolutionary simulation again in the modified market digital twin environment to obtain the competitive evolution trajectory under the strategy scenario, and quantifying the net effect of the market intervention strategy.

5. The hospital pharmacy competitiveness evaluation method based on multi-source data fusion according to claim 4, characterized in that, The quantification of the net effect of the market intervention strategy is calculated based on the competitive evolution trajectory under the benchmark scenario and the competitive evolution trajectory under the strategy scenario, and the calculation formula of the net effect is: AC i =C i ′ (T s )-C i (T s ); where ΔC i is the net impact of the market intervention strategy on the i-th market entity; C i ′ (T s ) is the competitiveness score of the i-th market entity at the end of the simulation T s under the strategy scenario; C i (T s ) is the competitiveness score of the i-th market entity at the end of the simulation T s under the baseline scenario. 6.The hospital pharmacy competitiveness evaluation method based on multi-source data fusion of claim 1, characterized in that, The step of periodically obtaining market observation data in the real world, calculating the simulation reality deviation between the simulation state of the market digital twin environment and the market observation data comprises: Obtaining the simulation state of the market digital twin environment at the real world time point corresponding to the market observation data; Based on the simulation state and the market observation data, a simulation reality deviation is calculated, and a calculation formula of the simulation reality deviation is: where D(t real ) is the simulated reality bias for a real-world time point t real ; t real is the real-world time point at which market observation data was acquired; is the simulated state value of the jth market variable in the simulated state; is the market observation data value of the jth market variable in the market observation data; k is the total number of market variables used to calculate the bias; ω j is the bias sensitivity weight for the jth market variable.

7. The hospital pharmacy competitiveness evaluation method based on multi-source data fusion according to claim 1, characterized in that, The step of correcting the initial market causal relationship graph by using the simulation reality deviation comprises: Identifying a high-deviation market variable according to the simulation reality deviation; Tracing an upstream causal chain of the high-deviation market variable in the initial market causal relationship graph; According to the upstream causal chain and the simulation reality deviation, the topology structure or the parameter of the initial market causal relationship graph is corrected.

8. A hospital pharmacy competitiveness evaluation system based on multi-source data fusion, applied to the hospital pharmacy competitiveness evaluation method based on multi-source data fusion according to any one of claims 1-7, characterized in that, Comprise: A data processing and feature construction module is configured to collect multi-source heterogeneous data, clean, format convert and standardize the collected multi-source heterogeneous data to form fusion data, determine a hospital or a drugstore as a market entity based on the fusion data, and construct a market entity feature vector for each market entity; A market causal relationship graph construction module is configured to construct an initial market causal relationship graph based on the market entity feature vector, wherein the initial market causal relationship graph represents the causal relationship between market variables; A market digital twin environment initialization module is configured to initialize a market digital twin environment based on the initial market causal relationship graph, wherein the market digital twin environment comprises a plurality of simulation entities, each simulation entity corresponds to a market entity, and the dynamic behavior rule of the simulation entity is determined by the initial market causal relationship graph; A competitiveness evolution simulation module is configured to run an evolution simulation in the market digital twin environment to obtain the competitiveness evolution trajectory of each market entity in a benchmark scenario; A simulation reality deviation calculation module is configured to periodically obtain market observation data of a real world, calculate the simulation state of the market digital twin environment and the simulation reality deviation of the market observation data; A causal graph correction module is configured to determine whether the simulation reality deviation is greater than a preset threshold, if so, correct the initial market causal relationship graph by using the simulation reality deviation to obtain an updated market causal relationship graph, and update the dynamic behavior rule of the simulation entity using the updated market causal relationship graph.

9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the method of any one of claims 1-7.

10. A storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to implement the method of any one of claims 1-7.

Images