Pharmacy department medicine information service platform based on cloud computing

Abstract

The invention relates to the technical field of medicine information management, and particularly discloses a pharmacy department medicine information service platform based on cloud computing, which is characterized in that a cloud computing storage module realizes medicine information cloud distributed storage; the information acquisition and input module supports automatic and manual data collection; the information processing and analyzing module is used for predicting drug requirements, analyzing adverse reactions and detecting incompatibility by applying various algorithms; the information sharing and interaction module promotes multi-party information intercommunication; the security guarantee module encrypts transmission, sets authority and backs up data; and the user management module allocates authority and authenticates login. Wherein drug demand prediction is combined with multiple models for dynamic integrated prediction and correction; analyzing untoward effects, comprehensively detecting anomalies by multiple methods, constructing a causal chain, and generating a risk index; compatibility taboo is detected, multi-source knowledge is integrated, efficient processing is achieved through multi-mode detection, risk quantification and grading and other processes by means of cloud computing distributed computing, and intelligent, safe and reliable services are provided for medicine management of the pharmacy department.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medicine information management, and particularly discloses a medicine information service platform for a pharmacy department based on cloud computing. BACKGROUND

[0002] In the medical industry, the pharmacy department undertakes important tasks such as medicine management, supply, and rational drug use guidance. With the development of medical informatization, the amount of medicine information is growing explosively, and the traditional medicine information management method gradually exposes many drawbacks.

[0003] On the one hand, the traditional medicine information storage relies on local servers, and the storage capacity is limited, which is difficult to meet the long-term storage needs of a large amount of medicine information, and there is a risk of data loss. On the other hand, in terms of information processing and analysis, there is a lack of effective technical means to deeply mine and analyze massive medicine information, it is difficult to accurately predict the medicine demand trend, and it is impossible to timely discover the adverse reactions and incompatibility of medicines, resulting in low medicine management efficiency and hidden dangers in drug safety.

[0004] At the same time, the information sharing between different medical institutions, the pharmacy department and other departments of the hospital, and the pharmacy department and the pharmaceutical suppliers is not smooth, the information interaction is hindered, the collaborative management and efficient use of medicine information cannot be realized, and the improvement of the quality of medical services is seriously restricted.

[0005] Therefore, a medicine information service platform for a pharmacy department based on cloud computing has become a problem that people urgently need to solve. SUMMARY

[0006] The technical problem to be solved by the present application is to provide a medicine information service platform for a pharmacy department based on cloud computing to realize efficient storage, accurate analysis, safe sharing and intelligent management of medicine information, and improve the efficiency of medicine management and drug safety in the pharmacy department.

[0007] To solve the above technical problems, the technical solution provided by the present application is as follows: a medicine information service platform for a pharmacy department based on cloud computing, comprising the following modules:

[0008] The cloud computing storage module is used to store the medicine information of the pharmacy department in the cloud server, and the medicine information is distributedly stored in multiple physical nodes. By selecting a cloud computing service provider to configure cloud storage resources, and regularly maintaining and optimizing the stored data, the safety and reliability of data storage can be effectively improved, data loss caused by single node failure can be avoided, and the growing storage needs of medicine information can be met.

[0009] Information Collection and Entry Module: This module is used to collect and enter the basic information of drugs, production information, pharmacological information, inventory information, and procurement information. It can automatically collect drug information by connecting with hospital information systems and drug supplier systems, while allowing manual entry. Connecting with hospital information systems can obtain patient medication information and medical order information, and connecting with drug supplier systems can synchronize drug procurement information and inventory information. Manual entry is set with data verification function to ensure the accuracy and integrity of the collected and entered information.

[0010] Information Processing and Analysis Module: This module processes and analyzes the collected drug information. Data mining and machine learning algorithms are used to analyze drug usage, predict drug demand trends, analyze drug adverse reactions, and detect drug incompatibility.

[0011] Predicting Drug Demand Trends: ARIMA model is used to capture the autoregressive and moving average characteristics of drug demand, and historical observations and prediction errors are used to predict future demand. LSTM network is used to handle long-term dependencies in data, and the cell state is updated through forget gate, input gate and output gate to obtain the predicted value. XGBoost model is used to build decision tree ensemble for comprehensive analysis and prediction of multi-dimensional features. The prediction results of ARIMA model, LSTM network and XGBoost model are dynamically integrated with weights, and the weights are constantly updated according to the loss function and learning rate to make the integrated prediction value more consistent with the actual demand. The deviation rate is calculated to evaluate the difference between the integrated prediction value and the actual value. When the deviation rate exceeds the threshold, the correction coefficient is dynamically adjusted according to the drug category to correct the subsequent integrated prediction value, thereby achieving accurate prediction of drug demand trends and providing scientific basis for drug procurement and inventory management.

[0012] Analyzing adverse drug reaction information: Enumerate all possible drug combinations and adverse reactions; Calculate the probability of simultaneous occurrence of drug combinations and adverse reactions, and screen out combinations with a frequency higher than the threshold; Evaluate the conditional probability of adverse reactions when drug combinations occur; Determine the significance of the rules and identify the positive correlation between drug combinations and adverse reactions; Finally, generate a rule base to provide prior assumptions for subsequent causal analysis. Construct the adverse reaction distribution vector according to the time dimension, and detect abnormalities through Z-score: Calculate the mean and standard deviation of the adverse reaction distribution vector within the time window; The Z-score of the current time adverse reaction distribution vector reflects its deviation from the historical baseline; When |Z-score| exceeds the preset threshold, automatically mark the abnormal event. Calculate the probability of patients using a specific drug through logistic regression, balance the baseline characteristics of the medication group and the non-medication group, and reduce selection bias; Estimate the potential outcome difference between medication and non-medication through weighting method; Construct the causal chain of "drug-metabolic pathway-adverse reaction". Extract key information from electronic medical records and adverse reaction reports: Identify drugs and adverse reactions through the BERT model, output entity pairs; Quantify the semantic intensity of the description sentence, output a severity value between 0 and 1. Generate a comprehensive risk index based on the severity and frequency of adverse reactions; When the adverse reaction monitoring index exceeds the preset threshold, push an alert to the pharmacy department with a list of risky drugs to promptly identify and handle adverse drug reactions and ensure patient safety.

[0013] Detecting drug compatibility contraindications:

[0014] S1, Multi-source knowledge integration and feature construction

[0015] Extract drug components, effects, and interaction relationships from pharmaceutical databases, clinical guidelines, and literature to construct a drug component-effect knowledge graph G = (V, E), where nodes V include drugs, effects, and relationships, and edges E represent the association between nodes; Establish a drug interaction matrix and calculate the interaction strength of drug pairs through weighting, incorporating molecular structure similarity and metabolic pathway overlap; Extract drug chemical structure features and medication regimen features to provide a data foundation for subsequent detection;

[0016] S2, Multi-modal compatibility contraindication detection

[0017] Determine whether there are metabolic competition and target overlap contraindications between drug pairs through path search and semantic matching in the knowledge graph, outputting a binary result; Input drug structure and medication regimen features into the XGBoost model to calculate the compatibility risk probability RiskScore through the Sigmoid function, capturing non-linear associations; Based on the preset rule base, perform logical judgment on drug combinations and output rule penalty values RulePenalty;

[0018] S3, Risk quantification and hierarchical early warning

[0019] The knowledge graph inference result, machine learning risk probability and rule penalty value are weighted and integrated into a comprehensive risk score; according to the ROC curve of historical data, a threshold is set to divide the risk into four levels: none, low, medium and high;

[0020] S4, cloud computing distributed computing

[0021] By using the distributed characteristics of cloud computing, the drug knowledge graph is divided into different computing nodes according to the dimensions of pharmacological classification and chemical structure for parallel processing through the MapReduce framework: in the Map stage, a detection key-value pair is generated for each drug combination, and in the Reduce stage, the results are combined; according to the real-time query load, the computing resources are dynamically adjusted to ensure the detection efficiency under large-scale data;

[0022] S5, real-time feedback and knowledge update

[0023] When new compatibility cases are fed back by clinicians, the posterior probability of the contraindication hypothesis is updated through the Bayes formula, and the model parameters are optimized in real time using incremental learning algorithm; if the confidence of the newly discovered drug relationship exceeds the threshold, the knowledge graph is automatically expanded, and the dynamic evolution of the contraindication knowledge base is realized;

[0024] S6, multi-dimensional verification and explanation generation

[0025] Explanations are constructed through multi-source evidence chains, and cross-validation is performed using heterogeneous models; the uncertainty of the calculation results is calculated, and when the uncertainty is too high, an expert review process is triggered to ensure the accuracy of high-risk contraindication judgments and effectively prevent the occurrence of drug compatibility contraindications.

[0026] Information sharing and interaction module: based on the network characteristics of cloud computing, the drug information sharing between different medical institutions and the information interaction between the pharmacy department and the hospital other departments are realized, and the patient medication consultation service is provided. Promote the circulation and collaborative management of drug information, improve the overall efficiency and quality of medical services.

[0027] Security module: encryption is performed during data transmission, and SSL / TLS encryption protocol is used to ensure the security of data transmission; access permissions are set for stored data, and access control list, role-based access control technology is used in the data storage layer to ensure the security of data storage; regular backup of cloud data and establishment of disaster recovery mechanism, backup data storage in different geographical locations and establishment of disaster recovery exercise mechanism, prevent data loss and damage, ensure the integrity and availability of data.

[0028] User management module: manage medical staff, pharmacy staff, drug suppliers and patients, and assign different permissions to different types of users. Design user management database to record user basic information, role information and permission information, use multi-factor authentication method for user login authentication, and record user operation log to ensure the security and traceability of user operation.

[0029] The advantages of the present application compared with the prior art are: the present application realizes efficient storage and management of drug information through cloud computing technology, uses advanced data mining and machine learning algorithms to deeply analyze drug information, can accurately predict drug demand trend, timely discover drug adverse reactions and detect drug incompatibility, effectively improves the efficiency of pharmacy drug management and drug safety. At the same time, the information sharing and interaction module promotes the circulation and collaborative management of drug information, the security protection module and the user management module ensure the safety of data and the standardization of user operation, and provides reliable technical support for medical service. BRIEF DESCRIPTION OF DRAWINGS

[0030] Figure 1 is a system block diagram of a pharmacy drug information service platform based on cloud computing. DETAILED DESCRIPTION

[0031] The various exemplary embodiments of the present application will be described in detail below with reference to the accompanying drawings. It should be noted that the relative arrangement, numerical expressions and values of the components and steps set forth in these embodiments do not limit the scope of the present application unless otherwise specifically stated.

[0032] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way limiting to the scope of the application and its applications or uses.

[0033] Techniques, methods, and devices known to those of ordinary skill in the relevant art can not be discussed in detail herein, but should be considered as part of the specification, where appropriate.

[0034] In all examples shown and discussed herein, any specific values should be interpreted as merely illustrative, and not as a limitation. Thus, other examples of the exemplary embodiments can have different values.

[0035] The pharmacy drug information service platform based on cloud computing will be further described in detail below in conjunction with the accompanying drawings.

[0036] In conjunction with the accompanying Figure 1 , the specific implementation process of the pharmacy drug information service platform based on cloud computing is as follows:

[0037] A pharmacy drug information service platform based on cloud computing comprises the following modules:

[0038] Cloud computing storage module: This module is used to store the information of the pharmacy department in the cloud server, and the information of the medicine is distributedly stored in multiple physical nodes. By selecting a cloud computing service provider to configure the cloud storage resources, and regularly maintaining and optimizing the stored data, the security and reliability of data storage can be effectively improved, avoiding data loss due to single node failure, while meeting the growing storage needs of drug information.

[0039] Information collection and input module: This module is used to collect and input the basic information of the medicine, production information, pharmacological information, inventory information and procurement information. It can automatically collect drug information by connecting with the hospital information system and the drug supplier system, and also allows manual input. Connecting with the hospital information system can obtain patient medication information and medical order information, and connecting with the drug supplier system can synchronize drug procurement information and inventory information. When manually inputting, data verification function is set to ensure the accuracy and integrity of the collected and input information.

[0040] Information processing and analysis module: This module processes and analyzes the collected drug information. Data mining and machine learning algorithms are used to analyze drug use, predict drug demand trends, analyze drug adverse reaction information, and detect drug compatibility contraindications.

[0041] Predicting drug demand trends:

[0042] Using additive model for time series decomposition: Y t = T t + S t + R t , where Y t is the drug demand at time t; T t is the trend component, estimated by Holt-Winters exponential smoothing method; S t is the seasonal component, which extracts periodic characteristics through Fourier transform; R t is the residual component, representing random fluctuations;

[0043] Using ARIMA model to capture the autoregressive and moving average characteristics of drug demand, and predicting future demand through historical observations and prediction errors;

[0044] ARIMA model prediction: ARIMA model prediction value of drug demand at time t; p: autoregressive order, representing the number of historical observations used in the model; φ i : autoregressive coefficient, reflecting the influence of the i-th historical value on the current value; B: lag operator; q: moving average order, representing the number of historical prediction errors used in the model; θ j: the moving average coefficient, reflecting the influence degree of the prediction error of the jth period on the current value; ∈ t : white noise error term at time t.

[0045] With the help of LSTM network to process the long-term dependence of data, the cell state is updated through the forget gate, input gate and output gate, and the predicted value is obtained;

[0046] LSTM model prediction: h t = o t ⊙tanh(c t ), Through training the LSTM network, the predicted value is obtained, where h t : the hidden state output of the LSTM model at time t (i.e. the predicted value); o t : the activation value of the output gate, which controls the output proportion of the hidden state, and the value range is [0, 1]; ⊙: element-level multiplication (element-wise multiplication); tanh: hyperbolic tangent activation function, which maps the value to the interval [-1, 1]; c t : cell state at time t, storing long-term memory information; f t : the activation value of the forget gate, which controls the forgetting proportion of historical information in the cell state, and the value range is [0, 1]; c t-1 : cell state at time t-1; i t : the activation value of the input gate, which controls the input proportion of new information, and the value range is [0, 1]; The candidate cell state to be updated is calculated from the current input.

[0047] XGBoost model is used to build decision tree integration for comprehensive analysis and prediction of multi-dimensional features.

[0048] XGBoost model prediction: where, XGBoost model prediction value of drug demand at time t. K: the number of decision trees in XGBoost model (the number of integrated trees). f k (x i ): the prediction value of the kth decision tree for sample x i . All possible regression tree function spaces, f k belongs to the space.

[0049] The prediction results of ARIMA model, LSTM network and XGBoost model are dynamically integrated with weights, and the weights are constantly updated according to the loss function and learning rate, so that the fused prediction value is more consistent with the actual demand.

[0050] Dynamic weight integration: where is the final drug demand prediction value at time t; a t , β t , γ t are the weights of the prediction values of each model at time t, and satisfy a t + β t + γ t = 1; are the prediction values of the ARIMA model, the LSTM model, and the XGBoost model at time t, respectively;

[0051] Weight update formula: where η is the learning rate, controlling the step size of weight update; L is the loss function, used to measure the error between the final prediction value and the true value Y t , calculated by the backpropagation algorithm Continuously adjust the weights to minimize the loss function;

[0052] Evaluate the difference between the fusion prediction value and the actual value by calculating the deviation rate. When the deviation rate exceeds the threshold, dynamically adjust the correction coefficient according to the drug category, and correct the subsequent fusion prediction value, so as to realize the accurate prediction of the drug demand trend, and provide a scientific basis for drug procurement and inventory management.

[0053] Deviation rate calculation: When e t > τ0, trigger model correction: where τ0 is the deviation rate threshold; δ is the correction coefficient, dynamically adjusted according to the drug category; sgn is the sign function.

[0054] Analyze drug adverse reaction information: enumerate all possible drug combinations and adverse reaction pairs; count the probability of simultaneous occurrence of drug combinations and adverse reactions, and select combinations with a frequency higher than the threshold; use the Apriori algorithm to mine drug-adverse reaction association rules: Support(X→Y) = P(X∪Y), representing the probability of simultaneous occurrence of drug combination X and adverse reaction Y;

[0055] Evaluate the conditional probability of adverse reaction occurrence when the drug combination appears: represents the probability of occurrence of adverse reaction Y under the condition of occurrence of drug combination X;

[0056] Judge the significance of the rule and determine that there is a positive correlation between the drug combination and the adverse reaction: used to measure the significance of the rule, Lift > 1 indicates a positive correlation; finally generate a rule base to provide prior assumptions for subsequent causal analysis.

[0057] Build the adverse reaction distribution vector in the time dimension, detect the abnormality by Z-score: calculate the mean and standard deviation of the adverse reaction distribution vector in the time window; the Z-score of the current time adverse reaction distribution vector reflects its deviation from the historical baseline; when |Z-score| exceeds the preset threshold, mark the abnormal event automatically;

[0058] Build the adverse reaction time sequence feature: ADR t = [adr t1 ,adr t2 ,...,adr tn ], which represents the adverse reaction distribution at time t; adopt the time series anomaly detection algorithm: Where μ t and σ t are the mean and standard deviation in the time window; when |Z-score|>τ1, trigger the adverse reaction warning, τ1 is the warning threshold.

[0059] Calculate the probability of patients using a specific drug through logistic regression, balance the baseline characteristics of the drug group and the non-drug group, and reduce selection bias; estimate the potential outcome difference between drug and non-drug by weighted method; build the causal chain of "drug-metabolic pathway-adverse reaction".

[0060] Extract key information from electronic medical records and adverse reaction reports: identify drug names and adverse reactions through BERT model, output entity pairs; quantify the semantic intensity of the description sentence, output a severity value between 0 and 1, the larger the value, the more serious.

[0061] According to the severity and frequency of adverse reactions, generate a comprehensive risk index; when the adverse reaction monitoring index exceeds the preset threshold, push an alarm to the pharmacy department, accompanied by a list of risky drugs, so as to timely discover and handle drug adverse reactions and ensure patient medication safety.

[0062] Build the adverse reaction monitoring index: Where w i represents the weight of the ith adverse reaction; when AMI t >AMI threshold , trigger the system alarm.

[0063] Detect drug compatibility contraindications:

[0064] S1, multi-source knowledge integration and feature construction

[0065] Extract drug components, effects and interaction relationships from pharmaceutical databases, clinical guidelines and literature, and build a drug component-effect knowledge graph G=(V,E); where the node set V={V d ,V e ,V r}, including drug nodes V d , effect nodes V e and relationship nodes V r ; edges E represent the association between nodes (such as "metabolic inhibition", "synergistic effect", etc.); the drug interaction matrix I m×n = [i ij ] is established; where i ij represents the interaction intensity of drug d i and d j , and the weighted s k is the similarity function of the kth effect, and w k is the corresponding weight; the interaction intensity of the drug pair is calculated, and the molecular structure similarity and metabolic pathway overlap are fused; the drug chemical structure features and medication regimen features are extracted to provide a data basis for subsequent detection; the drug chemical structure features include molecular fingerprints, physicochemical properties (such as lipid solubility, pKa value), etc. The medication regimen feature vector includes drug categories, doses, administration routes, medication time intervals, and other features.

[0066] S2, multi-modal compatibility contraindication detection

[0067] Reasoning detection based on knowledge graph: R(d i ,d j ) = Reasoning(G,d i ,d j ), determine whether there is a contraindication relationship between the drug pair through path search and semantic matching, and output a binary result;

[0068] Input the drug structure and medication regimen features into the XGBoost model, calculate the compatibility risk probability RiskScore through the Sigmoid function, and capture the nonlinear association; based on the pre-set rule base, logically judge the drug combination, and output the rule penalty value RulePenalty;

[0069] Machine learning risk prediction model: XGBoost model is used to output the compatibility risk probability: RiskScore = σ(W·concat(F d ,P) + b), where σ is the Sigmoid function, and W and b are model parameters; contraindication verification based on rule engine:

[0070]

[0071] The rule base includes: Rule k : IF (C k1 ∧ C k2 ∧ … ∧ C kn ) THEN AlertLevelk In this context, Cki is a conditional predicate, such as "drug A and drug B have the same metabolic pathway" and "the dose exceeds the threshold".

[0072] S3. Risk Quantification and Tiered Early Warning

[0073] The knowledge graph reasoning results, machine learning risk probabilities, and rule penalty values ​​are weighted and integrated into a comprehensive risk score.

[0074] Comprehensive Risk Score: FinalRisk = α·R(d) i ,d j )+β·RiskScore+γ·RulePenalty, where α+β+γ=1, and RulePenalty is the penalty value for rule matching;

[0075] Thresholds are set based on the ROC curve of historical data, and risks are divided into four levels: none, low, medium and high.

[0076] Risk grading mechanism:

[0077] Thresholds T1, T2, and T3 were determined using ROC curves from historical data.

[0078] S4, Cloud Computing Distributed Computing Implementation

[0079] Leveraging the distributed nature of cloud computing, the drug knowledge graph is segmented into different computing nodes according to pharmacological classification and chemical structure, and then processed in parallel using the MapReduce framework: the Map phase generates detection key-value pairs for each drug combination, and the Reduce phase merges the results; computing resources are dynamically adjusted according to the real-time query load to ensure detection efficiency under large-scale data.

[0080] Data partitioning: D split ={D1,D2,…,D n};

[0081] Parallel inference framework:

[0082] Parallel rule matching and model inference are implemented using the MapReduce paradigm:

[0083] Map(d)={(k,v)|k=Key(d),v=Check(d)};

[0084] Reduce(k,{v1,v2,…,v n})=Merge(v1,v2,…,v n );

[0085] Elastic resource scheduling: R(t) = R0 + k Load(t); the computing resource R is dynamically adjusted according to the real-time query load Load(t).

[0086] S5, real-time feedback and knowledge update

[0087] When the clinical feedback new compatibility cases, update the posterior probability of the contraindication hypothesis by Bayes formula, and use incremental learning algorithm to optimize the model parameters in real time;

[0088] Bayesian update mechanism: When the clinical feedback new compatibility cases E, update the posterior probability of the contraindication hypothesis H;

[0089] Incremental learning algorithm: Model parameters θ are updated in real time according to new cases (x t , y t ), and L is the loss function;

[0090] If the confidence of the newly discovered drug relationship exceeds the threshold, the knowledge graph is automatically expanded, and the dynamic evolution of the contraindication knowledge base is realized;

[0091] Dynamic expansion of knowledge graph: G t+1 = G t ∪{(v i ,e ij ,v j )|Confidence(v i ,e ij ,v j )>τ}, the newly discovered drug relationship is added to the knowledge graph after confidence verification.

[0092] S6, multi-dimensional verification and explanation generation

[0093] Construct explanation and explanation by multi-source evidence chain, and cross-verify by heterogeneous models:

[0094] Evidence chain construction: EvidenceChain = {e1, e2, …, e n}, each evidence ei is associated with confidence ci and source si, and supports explanation generation:

[0095] Explain(d i ,d j ) = Generate(EvidenceChain);

[0096] Cross-validation mechanism: Consensus = MajorityVote(M1, M2, …, M k ), improve detection accuracy by voting results of multiple heterogeneous models (such as knowledge graph reasoning, deep learning, rule engine).

[0097] Computational result uncertainty: U = 1-max (P (y1), P (y2), …, P (y n )) When the uncertainty U exceeds the threshold value, the expert review process is triggered to ensure the accuracy of high-risk contraindication judgment, effectively avoiding the occurrence of drug compatibility contraindication.

[0098] Information sharing and interaction module: Based on the characteristics of cloud computing network, the drug information sharing between different medical institutions pharmacy departments and between pharmacy department and other departments of the hospital is realized, the information interaction between pharmacy department and drug suppliers is realized, and the patient medication consultation service is provided. Promote the circulation and collaborative management of drug information, improve the overall efficiency and quality of medical service.

[0099] Security module: Encryption is performed during data transmission, and SSL / TLS encryption protocol is used to ensure the security of data during transmission; access permission is set for stored data, access control list and role-based access control technology are used in data storage layer to ensure the security of data storage; regular backup of cloud data and establishment of disaster recovery mechanism, backup data storage in different geographical locations and establishment of disaster recovery exercise mechanism, prevent data loss and damage, ensure the integrity and availability of data.

[0100] User management module: The medical staff, pharmacy staff, drug suppliers and patients are managed, and different permissions are allocated to different types of users. The user management database records the basic information, role information and permission information of the user, the multi-factor authentication method is used for user login authentication, and the user operation log is recorded to ensure the security and traceability of user operation.

[0101] The above describes the present application and its embodiments, which are not limited, and the drawings shown are only one of the embodiments of the present application, and the actual structure is not limited thereto. In summary, if a person skilled in the art is inspired by it, without departing from the purpose of the present application, without creative design, similar structure and embodiments of the technical solution can be designed, which should belong to the protection scope of the present application.

Claims

1. A cloud-based pharmacy department drug information service platform, characterized in that, include: The cloud computing storage module is used to store the pharmacy department's drug information on a cloud server and to distribute the drug information across multiple physical nodes. The information collection and entry module is used to collect and enter basic information, production information, pharmacological information, inventory information and procurement information of drugs, and to connect with the hospital information system and drug supplier system to realize automatic collection of drug information, while also allowing manual entry; The information processing and analysis module is used to process and analyze the collected drug information, and to use data mining and machine learning algorithms to analyze drug usage, predict drug demand trends, analyze adverse drug reaction information, and detect drug incompatibilities. The information sharing and interaction module is used to realize drug information sharing between pharmacy departments of different medical institutions and between pharmacy departments and other departments of the hospital based on the characteristics of cloud computing networks, realize information interaction between pharmacy departments and drug suppliers, and provide patients with medication consultation services. The security module is used to encrypt data during transmission, set access permissions for stored data, regularly back up cloud data, and establish a disaster recovery mechanism. The user management module is used to manage medical staff, pharmacy staff, drug suppliers, and patients, and to assign different permissions to different types of users.

2. The pharmacy department drug information service platform based on cloud computing according to claim 1, characterized in that: The specific method for predicting drug demand trends is as follows: The ARIMA model is used to capture the autoregressive and moving average features of drug demand, and future demand is predicted by historical observations and prediction errors. The LSTM network is used to process the long-term dependencies of the data, and cell states are updated through forget gates, input gates, and output gates to obtain the predicted values. The XGBoost model is used to construct a decision tree ensemble to comprehensively analyze and predict multi-dimensional features. The prediction results of the ARIMA model, LSTM network, and XGBoost model are dynamically weighted and integrated, and the weights are continuously updated according to the loss function and learning rate to make the fused prediction values ​​more closely match the actual needs. The difference between the predicted and actual values ​​of fusion is assessed by calculating the deviation rate. When the deviation rate exceeds the threshold, the correction coefficient is dynamically adjusted according to the drug category to correct the subsequent predicted values ​​of fusion.

3. The pharmacy department drug information service platform based on cloud computing according to claim 2, characterized in that: The specific methods for analyzing adverse drug reaction information are as follows: The algorithm enumerates all possible drug combinations and pairings with adverse reactions; it calculates the probability of drug combinations and adverse reactions occurring simultaneously, and filters out combinations with frequencies higher than a threshold; it assesses the conditional probability of adverse reactions occurring when drug combinations occur; it determines the significance of the rules and identifies a positive correlation between drug combinations and adverse reactions; finally, it generates a rule base to provide prior hypotheses for subsequent causal analysis. An adverse reaction distribution vector is constructed based on the time dimension, and anomalies are detected by Z-score: the mean and standard deviation of the adverse reaction distribution vector within the time window are calculated; the Z-score of the adverse reaction distribution vector at the current moment reflects its deviation from the historical baseline; when |Z-score| exceeds a preset threshold, an abnormal event is automatically marked. Logistic regression was used to calculate the probability of patients using a specific drug, balancing the baseline characteristics of the drug-using and non-drug-using groups to reduce selection bias; a weighted method was used to estimate the potential difference in outcomes between drug-using and non-drug-using groups; and a causal chain of "drug-metabolic pathway-adverse reaction" was constructed. Extracting key information from electronic medical records and adverse reaction reports: Identifying drug names and adverse reactions using the BERT model, and outputting entity pairs; Perform semantic intensity quantization on the description statement and output a severity value between 0 and 1; A comprehensive risk index is generated based on the severity and frequency of adverse reactions. When the adverse reaction monitoring index exceeds the preset threshold, an alarm is sent to the pharmacy department along with a list of risky drugs.

4. A cloud computing-based pharmacy drug information service platform according to claim 3, characterized in that: The specific methods for detecting drug incompatibilities are as follows: S1. Multi-source knowledge integration and feature construction Drug components, effects, and interactions are extracted from medical databases, clinical guidelines, and literature to construct a drug component-effect knowledge graph G = (V, E), where nodes V contain drugs, effects, and relationships, and edges E represent the associations between nodes. A drug interaction matrix was established, and the interaction strength of drug pairs was calculated by weighting, integrating molecular structure similarity and metabolic pathway overlap. Extracting the chemical structural characteristics and dosing regimen characteristics of drugs provides a data foundation for subsequent detection; S2, Multimodal Compatibility Detection By using path search and semantic matching in the knowledge graph, we can determine whether there are contraindications such as metabolic competition or target overlap between drug pairs and output a binary result. The characteristics of drug structure and dosing regimen are input into the XGBoost model, and the compatibility risk probability RiskScore is calculated through the Sigmoid function to capture nonlinear correlations. Based on a pre-defined rule base, logical judgments are made on drug combinations, and a rule penalty value (RulePenalty) is output. S3. Risk Quantification and Tiered Early Warning The results of knowledge graph reasoning, machine learning risk probability, and rule penalty value are weighted and integrated into a comprehensive risk score; based on the ROC curve of historical data, a threshold is set to classify the risk into four levels: none, low, medium, and high. S4, Cloud Computing Distributed Computing Implementation Leveraging the distributed nature of cloud computing, the drug knowledge graph is segmented into different computing nodes according to pharmacological classification and chemical structure, and then processed in parallel using the MapReduce framework: the Map phase generates detection key-value pairs for each drug combination, and the Reduce phase merges the results; computing resources are dynamically adjusted according to the real-time query load to ensure detection efficiency under large-scale data. S5, Real-time Feedback and Knowledge Updates When new drug compatibility cases are reported from the clinic, the posterior probability of the contraindication hypothesis is updated using Bayesian formula, and the model parameters are optimized in real time using incremental learning algorithm; if the confidence of the newly discovered drug relationship exceeds the threshold, the knowledge graph is automatically expanded to achieve dynamic evolution of the contraindication knowledge base. S6. Multi-dimensional verification and explanation generation Explanations are constructed using multi-source evidence chains, and cross-validation is performed using heterogeneous models; The calculation results are uncertain. When the uncertainty is too high, an expert review process is triggered to ensure the accuracy of high-risk taboo judgments.

5. A cloud computing-based pharmacy drug information service platform according to claim 4, characterized in that: The cloud computing storage module selects a cloud computing service provider to configure cloud storage resources and regularly maintains and optimizes the stored data.

6. A cloud-based pharmacy drug information service platform according to claim 5, characterized in that: The information collection and entry module interfaces with the hospital information system to obtain patient medication information and medical order information, and connects with the drug supplier system to synchronize drug procurement information and inventory information. It also has a data verification function set up when manually entering data.

7. A cloud computing-based pharmacy drug information service platform according to claim 6, characterized in that: The security module employs SSL / TLS encryption at the data transmission layer and access control lists and role-based access control at the data storage layer. Backup data is stored in different geographical locations, and a disaster recovery drill mechanism is established.

8. A cloud computing-based pharmacy drug information service platform according to claim 7, characterized in that: The user management module is designed with a user management database to record user basic information, role information, and permission information. It uses a multi-factor authentication method for user login authentication and records user operation logs.

Images