Intelligent planning generation method and system for medical marketing activities

Abstract

The invention discloses an intelligent planning generation method and system for a medicine marketing activity, and the method comprises the steps: collecting medicine market dynamic data, patient portrait data, doctor behavior data, medicine circulation data, policy and regulation data and medical institution operation data through a data collection and preprocessing step, and constructing a structured data set after processing; a demand analysis and target setting step: based on the structured data set, using a pre-trained BERT model to analyze a medical enterprise marketing demand document, extracting key demand points such as a target market area, key promotion of drugs, budget limitation and time nodes, and setting a target demand point; marketing objectives such as market share improvement, medicine popularization range, compliance requirements and patient service satisfaction are set in combination with enterprise strategic planning; and a strategy generation and optimization step of generating an initial marketing strategy by using a deep Q network reinforcement learning algorithm.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent planning generation of marketing activities in the pharmaceutical industry, in particular to an intelligent planning generation method and system for pharmaceutical marketing activities. BACKGROUND

[0002] Pharmaceutical marketing activities are the core link of product promotion and market share improvement for pharmaceutical enterprises, involving multi-dimensional complex factors such as market dynamics, patient needs, physician behavior, policies and regulations, and medical institution operation. With the intensification of competition in the pharmaceutical industry and the promotion of digital transformation, how to efficiently integrate multi-source data, scientifically set marketing goals, and dynamically optimize strategies has become a key challenge to improve marketing efficiency and effectiveness. In the field of marketing in the pharmaceutical industry, marketing activity planning, as the core link connecting product promotion and terminal demand, has long been challenged by the lack of data-driven decision-making capabilities. Under the traditional mode, the marketing team mainly relies on manual experience to plan activities, analyzes historical sales data, regional market characteristics, and competitor dynamics, and formulates activity plans based on industry practices. Although this mode can maintain a certain flexibility, it gradually exposes the limitations of systematic analysis capabilities in dealing with complex and changing pharmaceutical market environments.

[0003] In existing pharmaceutical marketing activity planning technology, data collection and processing are mostly limited to a single source or a specific period, and the cross-platform and cross-period data fusion capability is insufficient, which limits the comprehensiveness and accuracy of market insight. Marketing goal setting relies on personal experience and lacks systematic integration of enterprise strategies and quantitative indicators, resulting in insufficient goal scientificity and operability. Strategy generation lacks quantitative model support, and when facing market fluctuations or policy adjustments, the adjustment period is long and the direction is blind. Customer stratification strategies are relatively extensive, and there is a mismatch between resource allocation and actual terminal demand, affecting the stability of marketing effectiveness. Therefore, an intelligent planning generation method and system for pharmaceutical marketing activities are proposed. SUMMARY

[0004] The present application provides the following technical solutions: an intelligent planning generation method for pharmaceutical marketing activities, comprising the following steps: S1, data collection and preprocessing step: Collecting pharmaceutical market dynamic data, patient portrait data, physician behavior data, drug circulation data, policy and regulation data, and medical institution operation data; cleaning, deduplicating, and standardizing the collected data to construct a structured data set; S2, demand analysis and target setting step: Based on the structured data set, the pre-trained BERT model is used to analyze the marketing demand documents of pharmaceutical enterprises, and the target market area, key promoted drugs, budget limit and time node key demand points are extracted. Combined with enterprise strategic planning, marketing goals are set, including market share improvement, drug promotion range, compliance requirements and patient service satisfaction goals; S3, strategy generation and optimization step: Using deep Q network reinforcement learning algorithm, according to the marketing target, combined with real-time market dynamic data and patient demand data, the initial marketing strategy is generated; Using non-dominated sorting genetic algorithm multi-objective optimization algorithm, the market promotion effect, cost input, compliance risk and patient service experience four targets are balanced and optimized; S4, compliance review step: Establish a compliance rule library, use regular expression matching and Word2Vec model semantic similarity calculation to match rules one by one for the optimized strategy, automatically identify and correct the contents that do not comply with regulations and policies; S5, execution and feedback adjustment step: The compliance marketing strategy is output to the marketing execution system through JSON format; Real-time collection of execution feedback data; According to the indicators of sales fluctuation amplitude, patient satisfaction score and doctor prescription amount change, automatically adjust the next stage promotion resource allocation proportion, modify the propaganda content and adjust the activity form, so as to form a closed loop optimization mechanism.

[0005] Data collection and preprocessing module execution process: Through the data collection interface, the medical market dynamic data, patient portrait data, doctor behavior data, drug circulation data, policy and regulation data and medical institution operation data are synchronously obtained; The original data is cleaned to eliminate invalid fields and outliers; Based on the hash algorithm, the duplicate data is processed for deduplication; According to the pre-set data dictionary, the unstructured data is standardized and converted, and finally the structured data set is formed.

[0006] S2, demand analysis and target setting step execution process: Load the pre-trained BERT model to the natural language processing unit; Input the marketing demand documents of pharmaceutical enterprises to the BERT model for semantic analysis; From the analysis results, the target market area, key promoted drugs, budget limit and time node key demand elements are extracted; Combined with the long-term development goals in the enterprise strategic planning database, the marketing goals including market share improvement, drug promotion range, compliance requirements and patient service satisfaction are generated.

[0007] S3, strategy generation and optimization step execution process: Initialize the parameters of the deep Q network reinforcement learning algorithm, input marketing goals and real-time market dynamic data; Generate initial marketing strategy through neural network iteration; Start the non-dominated sorting genetic algorithm multi-objective optimization engine, input market promotion effect, cost input, compliance risk and patient service experience four optimization goals; Perform selection, crossover and mutation operations in genetic algorithm, output Pareto optimal solution set.

[0008] S4, compliance review step execution process: Load the pre-constructed compliance rule library, including drug advertising law, medical insurance policy and other regulatory provisions; Use regular expressions to match keywords in strategy text; Use Word2Vec model to calculate semantic similarity between strategy clauses and rule library; Automatically correct or mark strategy clauses with a matching degree below the threshold.

[0009] S5, execution and feedback adjustment step execution process: Serializes the marketing strategy that passes the compliance review into a JSON format data packet; Push the data packet to the marketing execution system through the API interface; Real-time receive feedback data such as sales fluctuation amplitude, patient satisfaction score and doctor prescription quantity change returned by the execution system; Based on the feedback data, dynamically adjust the allocation proportion of promotion resources, promotion content and activity form in the next stage.

[0010] Preferably, in the data collection and preprocessing step, the medical institution operation data further includes hospital procurement plan, department drug preference, outpatient volume change data and drug expiration date management data.

[0011] In the data collection and preprocessing step, the collection range of medical institution operation data is further expanded, specifically including hospital procurement plan, department drug preference, outpatient volume change data and drug expiration date management data.

[0012] Preferably, in the data collection and preprocessing step, patient portrait data is collected through patient mobile APP behavior logs, medical insurance reimbursement records and wearable device health monitoring data, and K-means clustering algorithm is used to group process patient drug frequency, drug preference and health status, forming a refined patient tag system.

[0013] In the data collection and preprocessing step, the collection of patient image data is completed through the patient mobile terminal APP behavior log, medical insurance reimbursement record and wearable device health monitoring data; after collection, the K-means clustering algorithm is used to group the patient medication frequency, drug preference and health status, and finally a refined patient label system is formed.

[0014] Preferably, in the S3, strategy generation and optimization step, real-time market dynamic data is obtained by interfacing with a third-party data platform, and patient demand data is collected in real time through a patient research system.

[0015] In the S3, strategy generation and optimization step, real-time market dynamic data is obtained by interfacing with a third-party data platform, and patient demand data is collected in real time through a patient research system.

[0016] Preferably, in the S4, compliance review step, the compliance rule library is updated once a day, and the differences between historical rules and new rules are automatically compared during updating to generate an update log and push it to compliance management personnel.

[0017] In the S4, compliance review step, the compliance rule library is updated once a day; the system automatically compares the differences between historical rules and new rules during updating to generate an update log and push it to compliance management personnel.

[0018] Preferably, in the S3, strategy generation and optimization step, the deep Q network reinforcement learning algorithm introduces historical marketing strategy data to construct an experience pool during training, and adjusts the ratio of the market share improvement coefficient to the cost input coefficient in the reward function to generate an initial marketing strategy that conforms to the current market dynamics.

[0019] In the S3, strategy generation and optimization step, the deep Q network reinforcement learning algorithm introduces historical marketing strategy data to construct an experience pool during training, and adjusts the ratio of the market share improvement coefficient to the cost input coefficient in the reward function to generate an initial marketing strategy that conforms to the current market dynamics.

[0020] An intelligent planning generation system for medical marketing activities, which adopts the above-mentioned intelligent planning generation method for medical marketing activities, comprising: Data collection module: Including a network crawler sub-module, a database interface sub-module, a sensor collection sub-module and an API interface sub-module; Data processing module: The collected data is processed to construct a structured data set; multi-source data fusion is supported, and patient medication data and doctor prescription data are integrated through correlation analysis to form a unified data view; Demand analysis module: The natural language processing submodule pre-trains a BERT and BiLSTM model for marketing demand document classification and entity recognition. The strategy generation module: The strategy generation module includes a reinforcement learning algorithm submodule and an environment interaction submodule. The strategy optimization module: The strategy optimization module includes a multi-objective optimization algorithm submodule and an optimization target management submodule, which defines four optimization targets: marketing promotion effect, cost input, compliance risk, and patient service experience. The compliance review module: The compliance review module includes a rule library management submodule and a rule matching submodule. The rule library management submodule automatically extracts the latest policies from official policy platforms every day, updates them through manual verification, and records the update logs. The rule matching submodule uses regular expression matching and Word2Vec model semantic similarity calculation to compare marketing strategy text with rule text, mark and automatically correct illegal content. The execution feedback module: The execution feedback module includes an execution interface submodule and a feedback analysis submodule. The feedback analysis submodule collects drug sales data, patient satisfaction survey results, doctor prescription volume changes, doctor feedback opinions, and market research reports, and automatically adjusts the distribution ratio of promotion resources in the next stage, modifies the promotion content, and adjusts the activity form according to the sales fluctuation amplitude, patient satisfaction score, and doctor prescription volume change indicators.

[0021] The data acquisition module performs multi-source data acquisition operations: The network crawler submodule targets the Internet medical information; The database interface submodule connects the enterprise's own database to extract historical marketing data; The sensor acquisition submodule obtains real-time participation data from offline activity sites; The API docking submodule accesses third-party medical information platforms to obtain industry dynamic data.

[0022] The data processing module performs structured processing on the obtained raw data: The structured data set includes patient medication records, doctor prescription behavior, and drug circulation information; Through entity alignment technology, the patient medication data and doctor prescription data are associated and analyzed; A multi-dimensional data view is generated, including time and space dimension features.

[0023] The demand analysis module performs marketing demand analysis: The natural language processing submodule uses a pre-trained BERT model for demand document classification; A bidirectional long short-term memory network (BiLSTM) model is used to extract key entities in the demand document; The strategic planning docking submodule maps the analysis result to the enterprise annual marketing strategy framework; The strategy generation module builds a decision model based on a reinforcement learning framework: The environment interaction submodule simulates market response to the environment; The reinforcement learning algorithm submodule generates an initial set of marketing strategies through a trial-and-error mechanism; The strategy optimization module performs multi-dimensional optimization: The multi-objective optimization algorithm submodule uses a Pareto frontier analysis method; The optimization target management submodule synchronously balances the optimization targets of market promotion effect, cost input, compliance risk, and patient service experience in four dimensions; The compliance review module implements a double verification mechanism: The rule library management submodule automatically captures official policy platform updates daily; After manual verification, the rule library is updated and a version control log is generated; The rule matching submodule performs format compliance checks using regular expressions; The Word2Vec model is used to calculate the semantic similarity between marketing text and rule text; The execution feedback module completes the strategy closed-loop management: The execution interface submodule pushes the compliance strategy to various execution terminals; The feedback analysis submodule continuously collects drug sales data, patient satisfaction survey results, doctor prescription volume change data, doctor feedback, and market research reports; According to the preset non-numerical index system, the resource allocation ratio is automatically adjusted, the promotion content is optimized, and the activity form is restructured.

[0024] Preferably, the sensor acquisition submodule of the data acquisition module further comprises a cold chain transportation temperature sensor for monitoring temperature changes during drug logistics to ensure compliance with drug storage requirements.

[0025] In the sensor acquisition submodule of the data acquisition module, the cold chain transportation temperature sensor performs temperature monitoring operations during drug logistics by collecting temperature data in real time to continuously monitor the temperature of the drug storage environment.

[0026] Preferably, in the multi-source data fusion process of the data processing module, a graph neural network model is used to construct a patient-doctor-drug association graph, learn patient medication characteristics, doctor prescription characteristics, and drug attribute characteristics through node embedding, calculate the association strength of patient medication behavior and doctor prescription behavior using edge weights, realize cross-domain data association analysis, and generate a unified data view.

[0027] In the multi-source data fusion process of the data processing module, a graph neural network model is used to construct a patient-doctor-drug association graph, patient medication characteristics, doctor prescription characteristics and drug attribute characteristics are learned through node embedding, and the association strength between patient medication behavior and doctor prescription behavior is calculated by using edge weight, so that cross-domain data association analysis is realized and a unified data view is generated.

[0028] Preferably, the rule matching sub-module of the compliance review module, when automatically correcting the violation content, preferentially retains the market share improvement and patient service satisfaction core indicators in the marketing target, adjusts the drug indication expression mode in the propaganda content and modifies the patient participation qualification conditions in the activity form, to ensure that the corrected strategy meets the requirements of the compliance rule library and the original marketing target setting.

[0029] The rule matching sub-module of the compliance review module, when automatically correcting the violation content, preferentially retains the market share improvement and patient service satisfaction core indicators in the marketing target, adjusts the drug indication expression mode in the propaganda content and modifies the patient participation qualification conditions in the activity form, to execute the violation content correction operation.

[0030] Compared with the prior art, the present application provides an intelligent planning generation method and system for medical marketing activities, which has the following beneficial effects: The present application collects medical market dynamic data, patient portrait data, doctor behavior data, drug circulation data, policy and regulation data and medical institution operation data through a multi-source data acquisition and preprocessing mechanism, constructs a structured data set after cleaning, deduplication and standardization, and supports multi-source data fusion and association analysis, breaks through the limitations of the prior art data utilization dimension single and cross-platform cross-cycle data fusion difficulty, significantly improves the comprehensiveness and accuracy of market insight, and provides a more reliable data basis for subsequent decision-making; The pre-trained BERT model is used to analyze the marketing demand documents of medical enterprises, automatically extract key demand points such as target market area, key promoted drugs, budget limit and time node, and set quantitative marketing targets such as market share improvement, drug promotion range, compliance requirements and patient service satisfaction in combination with enterprise strategic planning, which changes the traditional mode of relying on personal experience to set targets, makes the target setting more scientific and operable, and provides a clear direction for strategy generation; An initial marketing strategy is generated by using a deep Q network reinforcement learning algorithm, and a non-dominated sorting genetic algorithm is used for multi-objective optimization of market promotion effect, cost input, compliance risk and patient service experience, which solves the problem that the existing technology lacks quantitative model support in the decision-making process, and can quickly generate and optimize the strategy when facing sudden market fluctuations or regional policy adjustments, shorten the adjustment cycle, ensure the scientific nature of the adjustment direction, and avoid the experience-driven hysteresis and blindness; By executing the feedback module, multi-dimensional feedback data such as drug sales data, patient satisfaction survey results and doctor prescription quantity changes are collected in real time, and based on indicators such as sales fluctuation amplitude, patient satisfaction score and doctor prescription quantity change, the distribution proportion of promotion resources in the next stage is automatically adjusted, the propaganda content is modified, and the activity form is adjusted, forming a closed-loop optimization mechanism, solving the problem that the existing technology causes the resource allocation to be mismatched with the actual demand of the terminal due to the extensive customer stratification strategy, making the marketing activity more in line with the real needs of the market and patients, and improving the resource utilization efficiency and marketing effect stability. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1 is a schematic diagram of the method of the present application.

[0032] Figure 2 is a schematic diagram of the system of the present application. DETAILED DESCRIPTION

[0033] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0034] Please refer to Figure 1 The present application provides a technical solution, an intelligent planning generation method for medical marketing activities, comprising the following steps: S1, data acquisition and preprocessing step: Acquire medical market dynamic data, patient portrait data, doctor behavior data, drug circulation data, policy and regulation data and medical institution operation data; clean, deduplicate and standardize the collected data to construct a structured data set; S2, demand analysis and target setting step: Based on the structured data set, the pre-trained BERT model is used to analyze the marketing demand document of the pharmaceutical enterprise, extract the key demand points of the target market area, the key promotion drug, the budget limit and the time node, and set the marketing target according to the enterprise strategic planning, including market share improvement, drug promotion range, compliance requirement and patient service satisfaction target; S3, strategy generation and optimization step: Using deep Q network reinforcement learning algorithm, according to the marketing target, combined with real-time market dynamic data and patient demand data, generate initial marketing strategy; Use non-dominated sorting genetic algorithm multi-objective optimization algorithm to balance and optimize the four targets of market promotion effect, cost input, compliance risk and patient service experience; S4, compliance review step: Establish a compliance rule library, use regular expression matching and Word2Vec model semantic similarity calculation to match rules one by one for the optimized strategy, automatically identify and correct the content that does not comply with regulations and policies; S5, execution and feedback adjustment step: The compliance marketing strategy is output to the marketing execution system in JSON format; Real-time collection of execution feedback data; According to the indicators of sales fluctuation range, patient satisfaction score and doctor prescription amount change, automatically adjust the distribution ratio of promotion resources in the next stage, modify the publicity content and adjust the activity form to form a closed-loop optimization mechanism.

[0035] S1, data collection and preprocessing step execution process: Through the data collection interface, the medical market dynamic data, patient portrait data, doctor behavior data, drug circulation data, policy and regulation data and medical institution operation data are synchronously obtained; Perform cleaning operation on raw data to eliminate invalid fields and outliers; Based on the hash algorithm, the duplicate data is processed for deduplication; According to the pre-set data dictionary, the unstructured data is standardized and converted, and finally the structured data set is formed.

[0036] S2, demand analysis and target setting step execution process: Load the pre-trained BERT model to the natural language processing unit; Input the marketing demand document of pharmaceutical enterprises to the BERT model for semantic analysis; Extract key demand elements such as target market area, key promotion drugs, budget limit and time node from the analysis results; Combined with the long-term development target in the enterprise strategic planning database, generate marketing targets including market share improvement, drug promotion range, compliance requirements and patient service satisfaction.

[0037] S3, strategy generation and optimization step execution process: Initialize the parameters of deep Q network reinforcement learning algorithm, input marketing target and real-time market dynamic data; Generate initial marketing strategy through neural network iteration; Start the non-dominated sorting genetic algorithm multi-objective optimization engine, input the market promotion effect, cost investment, compliance risk and patient service experience as four optimization objectives; Perform selection, crossover and mutation operations in the genetic algorithm, and output the Pareto optimal solution set.

[0038] S4, compliance review step execution process: Load the pre-constructed compliance rule library, including drug advertising law, medical insurance policy and other regulatory provisions; Use regular expressions to match keywords in the strategy text; Use the Word2Vec model to calculate the semantic similarity between the strategy provisions and the rule library; Automatically correct or mark the strategy provisions with a matching degree below the threshold.

[0039] S5, execution and feedback adjustment step execution process: Serialize the marketing strategy that passes the compliance review into a JSON format data packet; Push the data packet to the marketing execution system through the API interface; Real-time receive feedback data such as sales fluctuation amplitude, patient satisfaction score and doctor prescription quantity change returned by the execution system; Based on the feedback data, dynamically adjust the allocation proportion of promotion resources, promotion content and activity form in the next stage; Through multi-source data collection and standardized processing, ensure the integrity and consistency of the basic data of marketing decision, and improve the reliability of subsequent analysis; Use a pre-trained language model to analyze the demand document, which can accurately extract key marketing elements and make the target setting highly consistent with the enterprise strategy; Combine reinforcement learning and multi-objective optimization algorithm to generate balanced strategies that take into account effect, cost, compliance and experience in dynamic market environment; Establish an automated compliance review mechanism to effectively reduce policy risks caused by human errors and ensure the legality of marketing activities; Build a closed-loop feedback adjustment system to realize dynamic adaptation from strategy formulation to execution optimization, and continuously improve the effectiveness of marketing activities.

[0040] In the data collection and preprocessing step S1, the medical institution operation data also includes hospital procurement plan, department drug preference, outpatient volume change data and drug expiration date management data.

[0041] In the data collection and preprocessing step S1, the medical institution operation data is further expanded, including hospital procurement plan, department drug preference, outpatient volume change data and drug expiration date management data; Through supplementing the hospital procurement plan, department drug preference and other operation data, the comprehensiveness of the operation information of the medical institutions is improved, more complete basic data support is provided for subsequent analysis, and the accuracy and pertinence of data preprocessing are enhanced.

[0042] In the data collection and preprocessing step S1, patient portrait data is collected through patient mobile terminal APP behavior logs, medical insurance reimbursement records and wearable device health monitoring data, and K-means clustering algorithm is used to group process the patient medication frequency, drug preference and health status, forming a refined patient label system.

[0043] In the data collection and preprocessing step S1, patient portrait data is collected through patient mobile terminal APP behavior logs, medical insurance reimbursement records and wearable device health monitoring data; after collection, K-means clustering algorithm is used to group process the patient medication frequency, drug preference and health status, and finally a refined patient label system is formed. Through multi-source data collection and clustering algorithm processing, a refined patient label system covering medication behavior, health status and other dimensions is constructed, the accuracy of patient portrait is improved, and a more reliable data basis is provided for personalized strategy formulation.

[0044] In the strategy generation and optimization step S3, real-time market dynamic data is obtained by interfacing with a third-party data platform, and patient demand data is collected in real time through a patient research system.

[0045] In the strategy generation and optimization step S3, real-time market dynamic data is obtained by interfacing with a third-party data platform, and patient demand data is collected in real time through a patient research system. By interfacing with a third-party platform and a real-time patient research system, the timeliness of market dynamic data and patient demand data is ensured, providing more current market environment input for strategy generation, and improving the real-time adaptability of the strategy.

[0046] In the compliance review step S4, the update frequency of the compliance rule library is once a day, and the differences between historical rules and new rules are automatically compared during update, an update log is generated and pushed to compliance management personnel.

[0047] In the compliance review step S4, the update frequency of the compliance rule library is once a day; during update, the system automatically compares the differences between historical rules and new rules, generates an update log and pushes it to compliance management personnel. Through daily update of the rule library and automatic generation of the update log, the timeliness and traceability of the compliance rules are ensured, which facilitates compliance management personnel to quickly master rule changes and improves the efficiency and accuracy of compliance review.

[0048] In the strategy generation and optimization step S3, the deep Q network reinforcement learning algorithm introduces historical marketing strategy data to build an experience pool during training, and generates an initial marketing strategy that meets the current market dynamics by adjusting the ratio of the market share improvement coefficient to the cost input coefficient in the reward function.

[0049] In the strategy generation and optimization step S3, the deep Q network reinforcement learning algorithm introduces historical marketing strategy data to build an experience pool during training, and generates an initial marketing strategy that meets the current market dynamics by adjusting the ratio of the market share improvement coefficient to the cost input coefficient in the reward function. By introducing the historical experience pool and adjusting the reward function coefficient, the reinforcement learning algorithm can more accurately balance the relationship between market share improvement and cost input, generate an initial strategy that better fits the current market dynamics, and improve the effectiveness and economy of the strategy.

[0050] Please refer to Figure 2 An intelligent planning generation system for medical marketing activities, which adopts the above-mentioned intelligent planning generation method for medical marketing activities, comprising: A data acquisition module: Including a web crawler sub-module, a database interface sub-module, a sensor acquisition sub-module, and an API docking sub-module. A data processing module: Process the collected data to build a structured data set; support multi-source data fusion, integrate patient medication data and doctor prescription data through correlation analysis, and form a unified data view. A demand analysis module: Including a natural language processing sub-module and a strategic planning docking sub-module, the natural language processing sub-module pre-trains BERT and BiLSTM models for marketing demand document classification and entity recognition. A strategy generation module: Including a reinforcement learning algorithm sub-module and an environment interaction sub-module. A strategy optimization module: Including a multi-objective optimization algorithm sub-module and an optimization target management sub-module, the optimization target management sub-module defines four optimization targets: market promotion effect, cost input, compliance risk, and patient service experience. A compliance review module: Including a rule library management sub-module and a rule matching sub-module, the rule library management sub-module automatically extracts the latest policies from official policy platforms every day, updates them through manual verification, and records the update logs; the rule matching sub-module uses regular expression matching and Word2Vec model semantic similarity calculation to compare marketing strategy text with rule text, mark and automatically correct illegal content. An execution feedback module: The interface execution submodule and the feedback analysis submodule are included. The feedback analysis submodule collects drug sales data, patient satisfaction survey results, doctor prescription volume changes, doctor feedback opinions and market research reports. According to the sales fluctuation amplitude, patient satisfaction score and doctor prescription volume change index, the distribution proportion of the next stage of promotion resources is automatically adjusted, the propaganda content is modified, and the activity form is adjusted.

[0051] The multi-source data acquisition operation is performed through the data acquisition module: The network crawler submodule targets the Internet medical information; The database interface submodule connects the enterprise's own database to extract historical marketing data; The sensor acquisition submodule obtains real-time participation data of offline activities on site; The API docking submodule accesses the third-party medical information platform to obtain industry dynamic data.

[0052] The data processing module performs structured processing on the obtained raw data: A structured data set containing patient medication records, doctor prescription behavior and drug circulation information is constructed; The association analysis of patient medication data and doctor prescription data is realized through entity alignment technology; A multi-dimensional data view containing spatio-temporal dimension features is generated.

[0053] The demand analysis module performs marketing demand analysis: The natural language processing submodule uses the pre-trained BERT model for demand document classification; The bidirectional long short-term memory network model is used to extract key entities in the demand document; The strategy planning docking submodule maps the analysis results to the enterprise's annual marketing strategy framework; The strategy generation module constructs a decision model based on the reinforcement learning framework: The environment interaction submodule simulates the market response environment; The reinforcement learning algorithm submodule generates an initial marketing strategy group through a trial-and-error mechanism; The strategy optimization module performs multi-dimensional optimization: The multi-objective optimization algorithm submodule uses the Pareto frontier analysis method; The optimization target management submodule synchronously balances the optimization targets of market promotion effect, cost input, compliance risk and patient service experience in four dimensions; The compliance review module implements a double verification mechanism: The rule library management submodule automatically crawls official policy platform updates every day; After manual verification, the rule library is updated and a version control log is generated; The rule matching submodule performs format compliance checks through regular expressions. The Word2Vec model is used to calculate the semantic similarity between marketing texts and rule texts. The feedback module completes the policy closed-loop management: The execution interface submodule pushes compliance policies to various execution terminals. The feedback analysis submodule continuously collects drug sales data, patient satisfaction survey results, doctor prescription volume change data, doctor feedback, and market research reports. According to the preset non-numerical index system, the resource allocation ratio is automatically adjusted, the propaganda content is optimized, and the activity form is reconstructed. The data acquisition module solves the problem of single data dimension in traditional systems through the integration of four types of heterogeneous data sources, improving the comprehensiveness of marketing decisions. The multi-dimensional data view constructed by the data processing module realizes the correlation analysis of patient medication behavior and doctor prescription behavior, providing data support for precision marketing. The demand analysis module uses a pre-trained language model to improve the accuracy of marketing demand document analysis, ensuring that the generated strategy aligns with the strategic objectives. The reinforcement learning framework of the strategy generation module can dynamically adapt to changes in the market environment, providing stronger environmental adaptability compared to traditional fixed algorithms. The strategy optimization module uses a multi-objective balancing mechanism to avoid decision bias caused by single optimization targets, achieving collaborative optimization of marketing effectiveness and cost control. The automatic rule update and double verification mechanism of the compliance review module builds real-time policy response capabilities, effectively reducing the compliance risks of marketing activities. The dynamic adjustment mechanism based on non-numerical indicators of the execution feedback module breaks through the limitations of traditional systems relying on fixed thresholds, achieving the continuous evolution capability of marketing strategies.

[0054] The sensor acquisition submodule of the data acquisition module also includes a cold chain transportation temperature sensor for monitoring temperature changes during drug logistics to ensure compliance with drug storage requirements.

[0055] In the sensor acquisition submodule of the data acquisition module, the cold chain transportation temperature sensor performs temperature monitoring operations during drug logistics by collecting temperature data in the logistics process in real time, continuously monitoring the temperature of the drug storage environment. By monitoring temperature changes during drug logistics, it ensures that drug storage meets the specified requirements and avoids drug quality risks caused by temperature abnormalities.

[0056] In the multi-source data fusion process of the data processing module, a graph neural network model is used to construct a patient-doctor-drug association graph, patient medication characteristics, doctor prescription characteristics and drug attribute characteristics are learned through node embedding, and the association strength between patient medication behavior and doctor prescription behavior is calculated using edge weight to realize cross-domain data association analysis and generate a unified data view.

[0057] In the multi-source data fusion process of the data processing module, a graph neural network model is used to construct a patient-doctor-drug association graph, patient medication characteristics, doctor prescription characteristics and drug attribute characteristics are learned through node embedding, and the association strength between patient medication behavior and doctor prescription behavior is calculated using edge weight to realize cross-domain data association analysis and generate a unified data view. By constructing a patient-doctor-drug association graph and performing cross-domain data association analysis, a unified data view is generated, the correlation of multi-source data is improved, and the analysis efficiency is improved, supporting more comprehensive data processing and decision-making.

[0058] When the rule matching submodule of the compliance review module automatically corrects the violation content, the market share improvement and patient service satisfaction core indicators in the marketing target are preferentially retained, the drug indication expression method in the propaganda content is adjusted, and the patient participation qualification condition in the activity form is modified to ensure that the corrected strategy meets the requirements of the compliance rule library and the original marketing target setting.

[0059] When the rule matching submodule of the compliance review module automatically corrects the violation content, the market share improvement and patient service satisfaction core indicators in the marketing target are preferentially retained, the drug indication expression method in the propaganda content is adjusted, and the patient participation qualification condition in the activity form is modified to ensure that the corrected strategy meets the requirements of the compliance rule library and the original marketing target setting. By preferentially retaining core marketing indicators and adjusting violation content, the corrected strategy meets the requirements of the compliance rule library and the original marketing target setting, achieving effective balance between compliance and marketing goals.

[0060] The scheme: through the data acquisition interface, medical market dynamic data, patient portrait data, doctor behavior data, drug circulation data, policy and regulation data and medical institution operation data are synchronously acquired; wherein the medical institution operation data includes hospital procurement plan, department drug preference, outpatient volume change data and drug expiration date management data; the patient portrait data is collected through patient mobile terminal APP behavior log, medical insurance reimbursement record and wearable device health monitoring data; Perform cleaning operation on original data, eliminate invalid fields and abnormal values; Based on the hash algorithm, the repeated data is processed for deduplication; According to the preset data dictionary, the unstructured data is standardized converted; The K-means clustering algorithm is used to cluster the patient's medication frequency, drug preference and health status, and form a refined patient tag system; Finally, a structured data set is formed; The pre-trained BERT model is loaded into the natural language processing unit; Input the marketing demand document of the pharmaceutical enterprise into the BERT model for semantic analysis; Extract the key demand elements such as target market area, key promoted drugs, budget limit and time node from the analysis results; Combined with the long-term development goals in the enterprise strategic planning database, the marketing goals including market share improvement, drug promotion range, compliance requirements and patient service satisfaction are generated.

[0061] It should be noted that in this paper, relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the term "includes", "contains" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment.

[0062] Although the embodiments of the present application have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for generating intelligent planning for pharmaceutical marketing activities, characterized in that: The steps include: S1. Data collection and preprocessing steps: Collect pharmaceutical market dynamics data, patient profile data, doctor behavior data, drug circulation data, policy and regulatory data, and medical institution operation data; clean, deduplicate, and standardize the collected data to construct a structured data set; S2. Demand analysis and goal setting steps: Based on structured datasets, we use a pre-trained BERT model to analyze pharmaceutical companies' marketing requirements documents, extracting key market areas, key drugs to promote, budget constraints, and timeframes. Combined with the company's strategic planning, we set marketing goals, including market share growth, drug promotion scope, compliance requirements, and patient satisfaction targets. S3. Strategy generation and optimization steps: A deep Q-network reinforcement learning algorithm is used to generate an initial marketing strategy based on marketing objectives, combined with real-time market dynamics data and patient demand data. A non-dominated sorting genetic algorithm multi-objective optimization algorithm is used to balance and optimize the four objectives of marketing effectiveness, cost input, compliance risk, and patient service experience. S4. Compliance review steps: Establish a compliance rule library, use regular expression matching and Word2Vec model semantic similarity calculation to match the optimized strategy rule by rule, and automatically identify and correct content that does not comply with regulations and policies; S5. Execution and feedback adjustment steps: Output compliant marketing strategies to the marketing execution system in JSON format; collect execution feedback data in real time; and automatically adjust the next stage's promotional resource allocation ratio, modify promotional content, and adjust activity formats based on indicators such as sales fluctuations, patient satisfaction scores, and changes in physician prescription volume, to form a closed-loop optimization mechanism.

2. The method for generating an intelligent plan for pharmaceutical marketing activities according to claim 1, characterized in that: In the step S1, data collection and preprocessing, the medical institution operation data also includes hospital procurement plans, department medication preferences, outpatient volume change data, and drug expiration date management data.

3. The method for generating an intelligent plan for pharmaceutical marketing activities according to claim 1, characterized in that: In the step S1, data collection and preprocessing, patient portrait data is collected through the patient's mobile APP behavior log, medical insurance reimbursement records and wearable device health monitoring data, and the K-means clustering algorithm is used to group the patient's medication frequency, drug preference and health status to form a refined patient label system.

4. The method for generating an intelligent plan for pharmaceutical marketing activities according to claim 1, characterized in that: In the above-mentioned S3, strategy generation and optimization step, real-time market dynamic data is obtained by connecting with a third-party data platform, and patient demand data is collected in real time through a patient survey system.

5. The method for generating an intelligent plan for pharmaceutical marketing activities according to claim 1, characterized in that: In the step S4, compliance review, the compliance rule base is updated once a day. During the update, the differences between the historical rules and the new rules are automatically compared, and an update log is generated and pushed to the compliance management personnel.

6. The method for generating an intelligent plan for pharmaceutical marketing activities according to claim 1, characterized in that: In the above-mentioned S3, strategy generation and optimization step, the deep Q network reinforcement learning algorithm introduces historical marketing strategy data to construct an experience pool during training, and generates an initial marketing strategy that conforms to current market dynamics by adjusting the ratio of the market share improvement coefficient to the cost input coefficient in the reward function.

7. An intelligent planning and generation system for pharmaceutical marketing activities, using an intelligent planning and generation method for pharmaceutical marketing activities according to any one of claims 1 to 6, characterized in that: include: Data acquisition module: Including web crawler submodule, database interface submodule, sensor acquisition submodule and API docking submodule; Data processing module: Process the collected data and construct a structured data set; Support multi-source data fusion, integrate patient medication data and doctor prescription data through correlation analysis to form a unified data view; Requirements Analysis Module: It includes a natural language processing submodule and a strategic planning docking submodule. The natural language processing submodule pre-trains BERT and BiLSTM models for marketing demand document classification and entity recognition; Strategy generation module: Includes reinforcement learning algorithm submodule and environment interaction submodule; Strategy optimization module: It includes a multi-objective optimization algorithm submodule and an optimization target management submodule. The optimization target management submodule defines four optimization targets: marketing effect, cost input, compliance risk, and patient service experience; Compliance Review Module: It includes a rule base management submodule and a rule matching submodule. The rule base management submodule automatically captures the latest policies from the official policy platform every day, manually verifies and updates them, and records update logs. The rule matching submodule uses regular expression matching and Word2Vec model semantic similarity calculation to compare marketing strategy text with rule text, mark and automatically correct illegal content. Execution feedback module: It includes an execution interface submodule and a feedback analysis submodule. The feedback analysis submodule collects drug sales data, patient satisfaction survey results, changes in doctors' prescription volume, doctors' feedback and market research reports. According to the sales fluctuation range, patient satisfaction score and doctors' prescription volume change indicators, it automatically adjusts the promotion resource allocation ratio for the next stage, modifies the promotion content and adjusts the activity form.

8. The intelligent planning and generation system for pharmaceutical marketing activities according to claim 7, characterized in that: The sensor acquisition submodule of the data acquisition module also includes a cold chain transport temperature sensor, which is used to monitor temperature changes during the drug logistics process to ensure that drug storage requirements are met.

9. The intelligent planning and generation system for pharmaceutical marketing activities according to claim 7, characterized in that: During the multi-source data fusion process of the data processing module, a graph neural network model is used to construct a patient-doctor-drug association graph, and the patient's medication characteristics, doctor's prescription characteristics and drug attribute characteristics are learned through node embedding. The edge weight is used to calculate the correlation strength between the patient's medication behavior and the doctor's prescription behavior, to achieve cross-domain data association analysis and generate a unified data view.

10. The intelligent planning and generation system for pharmaceutical marketing activities according to claim 7, characterized in that: When automatically correcting illegal content, the rule matching submodule of the compliance review module gives priority to retaining the core indicators of market share improvement and patient service satisfaction in the marketing objectives. By adjusting the expression of drug indications in the promotional content and modifying the patient participation eligibility conditions in the activity form, it ensures that the revised strategy meets both the compliance rule library requirements and the original marketing goal settings.

Images