Marketing strategy analysis optimization method based on big data and edge calculation

Abstract

The invention provides a marketing strategy analysis optimization method based on big data and edge calculation. The method comprises the steps of constructing a data acquisition model to acquire data; generating a global supply chain scheduling scheme and a global precision marketing strategy according to the local supply chain adjustment and demand matching strategy and the output of the user demand prediction model; the global supply chain scheduling scheme and the global precision marketing strategy are adjusted to be cooperatively executed on each edge device; feedback data of each edge device is collected, analyzed and globally aggregated, and the global supply chain scheduling scheme and the global precision marketing strategy are adaptively adjusted according to a global feedback matrix; and collecting dual feedback data of the global supply chain scheduling scheme and the global precision marketing strategy after adaptive adjustment, and performing dynamic closed-loop optimization according to the dual feedback data. The response speed and the operation efficiency of the medical instrument supply chain are improved, and the precision marketing effect and the customer satisfaction are remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the field of big data and edge computing, and in particular to a marketing strategy analysis and optimization method based on big data and edge computing. Background Art

[0002] With the continuous development of big data and Internet technology, more and more industries have begun to adopt digital transformation to improve operational efficiency, especially in the medical device industry. The production, circulation and sales of medical devices are complex and closely linked. Inefficiency in any link will lead to unsalable products, supply chain blockages or failure to meet terminal demand in a timely manner. At present, many companies have begun to use big data, cloud computing and artificial intelligence technologies to optimize supply chain management and marketing strategies. However, the use of existing technologies still has many shortcomings and fails to fully meet the unique needs of the medical device industry.

[0003] Existing supply chain management systems mostly rely on a centralized cloud computing model, which requires uploading a large amount of data from different supply chain nodes to a central server for processing and analysis. Although this centralized model can provide a relatively comprehensive supply chain scheduling analysis, it faces the following key problems in practical applications:

[0004] Data transmission delay and network dependence: Centralized processing requires frequent data transmission from various supply chain nodes to the cloud, which not only relies on the stability and transmission speed of the network, but also leads to large time delays, affecting the real-time adjustment of the supply chain. Especially in scenarios that require rapid response to market changes, network instability or bandwidth limitations will seriously weaken the efficiency of the system, resulting in delays in inventory management and logistics scheduling.

[0005] Concentrated processing load: The centralized processing of all data will bring a huge burden to the central server, especially in complex scenarios with numerous supply chain nodes and terminal devices. The amount of data increases dramatically, making it difficult for the system to maintain efficient operation. In this case, the central server can easily become a bottleneck, causing the response speed of the entire system to decrease, affecting the overall collaborative efficiency of the supply chain.

[0006] Precision marketing is not effective: In the medical device industry, customer needs are highly diverse and dynamically changing, and traditional marketing strategies cannot accurately capture and predict changes in demand at each terminal. Existing marketing systems mostly rely on historical data for push, lacking real-time and personalization, which makes recommended products and services out of touch with customer needs, greatly reducing marketing effectiveness and resulting in low conversion rates.

[0007] In summary, although existing technologies have utilized technologies such as big data and cloud computing, they still face many challenges in supply chain optimization and precision marketing, especially in terms of real-time responsiveness, system load distribution, and meeting personalized customer needs. Summary of the invention

[0008] The purpose of this invention is to design a marketing strategy analysis and optimization method based on big data and edge computing, which disperses computing pressure and speeds up response speed through edge computing technology, and combines real-time analysis of user behavior to achieve intelligent collaborative optimization of supply chain and marketing strategy.

[0009] In order to achieve the above object, the present invention provides a marketing strategy analysis and optimization method based on big data and edge computing, the method comprising:

[0010] S1. Construct a data collection model to collect data, wherein the data includes supply chain data and end-user data, and pre-process the collected data and perform weighted integration on the pre-processed data to obtain integrated features, dynamically adjust the local supply chain according to the integrated features to obtain a supply chain dynamic adjustment model, and construct a user demand prediction model according to the end-user data, and finally combine the outputs of the supply chain dynamic adjustment model and the user demand prediction model to generate a local supply chain adjustment and demand matching strategy;

[0011] The data collection model is constructed as follows:

[0012] Install edge devices at supply chain edge devices and end-user data collection points to collect supply chain data and end-user data;

[0013] The supply chain data includes inventory data, order data and logistics data; the terminal user data includes user browsing behavior and user purchasing behavior;

[0014] S2. Generate a global supply chain scheduling plan and a global precision marketing strategy based on the output of the local supply chain adjustment and demand matching strategy and the user demand prediction model. At the same time, the generated global supply chain scheduling plan and the generated global precision marketing strategy are designed to dynamically adjust the marketing push model based on user preferences during operation, and make adaptive adjustments based on user feedback to cope with real-time changes;

[0015] S3, combining the supply and demand situation, logistics conditions and global priority of each edge device with the global supply chain scheduling plan and global precision marketing strategy adjustment for coordinated execution at each edge device;

[0016] S4. Collect feedback data from each edge device, analyze and globally aggregate it to obtain an aggregated global feedback matrix, and adaptively adjust the global supply chain scheduling plan and global precision marketing strategy according to the global feedback matrix;

[0017] S5. Collect dual feedback data of the adaptively adjusted global supply chain scheduling plan and the global precision marketing strategy, and perform dynamic closed-loop optimization based on the dual feedback data. At the same time, during the dynamic closed-loop optimization process, build a multi-dimensional feedback convergence model to ensure that the supply chain data and end-user data can reach a convergence state in all dimensions. At the same time, design a feedback iterative convergence mechanism to ensure that convergence is maintained after iteration and to maintain the global optimal state.

[0018] Further, in S1, the preprocessing is: using local deviation detection to denoise the inventory data, and then using a behavior pattern detection model to filter abnormal user behavior data; extracting features from the supply chain data and the end user data, including:

[0019] Inventory Features F I , calculated as follows:

[0020]

[0021] in, is the inventory data after denoising. is the inventory change rate, calculated as:

[0022]

[0023] Order feature F O :

[0024] F O = {O sc,t ,ΔO sc,t}

[0025] Among them, sc Indicates order data, ΔO sc,t represents the order change rate, ΔO sc,t Also based on changes in order data;

[0026] Logistics Features F L :

[0027] F L ={L sc,t ,ΔL sc,t}

[0028] User behavior characteristics F B :

[0029] F B = {B u ,ΔB u} and F P = {P u ,ΔP u}

[0030] Among them, Bu represents the user's browsing behavior, P u represents the user's purchasing behavior, ΔB u and ΔP u Represent the change rates of user browsing and purchasing behaviors respectively;

[0031] The weighted integration of the preprocessed data is expressed as follows:

[0032] F agg =α I F I +α O F O +α L F L +β B F B +β P F P

[0033] Among them, F agg represents the integrated supply chain and user behavior characteristics, and the weight parameter α I , α O , α L They represent the importance weights of inventory, order, and logistics data respectively, and the weight of user behavior characteristics is β B and β P .

[0034] Furthermore, the supply chain dynamic adjustment model is expressed as follows:

[0035]

[0036] Among them, I sc,t Indicates the current inventory level; O sc,t Indicates the order increment at the current moment, reflecting the impact of external orders on inventory; L sc,t Represents the current logistics data, and adjusts the impact of logistics on inventory through time delay T; γ u represents the user demand adjustment coefficient, which is used to directly incorporate the results of user behavior prediction into inventory adjustment; D u,t It represents the predicted value of user demand, obtained using the user demand model;

[0037] The user demand prediction model is constructed based on user behavior data, which is expressed as follows:

[0038] D u,t+1 =β B ·ΔB u +β P ΔP u -λ·R u

[0039] Among them, Du,t+1 Indicates the predicted value of user demand at the next moment; ΔB u Indicates the rate of change of user browsing behavior; ΔP u represents the rate of change of user purchase behavior); λ represents the regularization coefficient, which controls the noise or extreme behavior in user behavior modeling; R u Inverse function of the last purchase time This item is used to suppress user demand forecasts that purchase too frequently;

[0040] According to the results of the supply chain dynamic adjustment model and the user demand prediction model, a local supply chain adjustment and demand matching strategy is generated, which is expressed as follows:

[0041] S opt =I sc,t+1 -D u,t+1

[0042] Among them, S opt Indicates the difference between local inventory and demand, which is used to indicate whether it is necessary to increase orders, adjust logistics, or adjust inventory;

[0043] When S opt >0, indicating excess inventory, it is necessary to reduce purchases or increase promotion efforts; when S opt <0, it means that the inventory is insufficient and it is necessary to speed up replenishment or reallocate logistics resources.

[0044] Furthermore, the global supply chain scheduling scheme integrates the supply and demand differences and demand forecasts of each supply chain node to generate a global supply and demand matrix S global , and then according to the global supply and demand matrix S global Build a scheduling priority model to determine the resource allocation priority of each node by adjusting the allocation of logistics, production and inventory;

[0045] Among them, the global supply and demand matrix S global It is expressed as follows:

[0046]

[0047] Among them, S global represents the global supply and demand matrix, which is used to describe the overall supply and demand balance of each node in the system; w i Represents the weight coefficient, which represents the priority of the i-th node in the supply chain system. opt,i represents the local supply-demand difference of the i-th node; η represents the global adjustment coefficient, which controls the impact of user demand changes on the supply-demand matrix; ΔD u,i represents the user demand change rate of the i-th node, that is, the prediction of future user demand;

[0048] The scheduling priority model is expressed as follows:

[0049]

[0050] Among them, P log,i represents the logistics scheduling priority of the i-th node, which determines the urgency of resource allocation at this node; S global,i represents the global supply-demand difference value of the i-th node, representing the current supply-demand imbalance state of the node; T i represents the logistics time delay of the i-th node; μ i Represents the uncertainty penalty term, which indicates the uncertainty degree of the supply chain of the i-th node.

[0051] Furthermore, the global precision marketing strategy combines user behavior data with global demand prediction to design a marketing push model that is dynamically adjusted based on user preferences, as shown below:

[0052]

[0053] Among them, M push,i represents the precision marketing push strength of the i-th node, which is used to determine the frequency of advertising or product recommendations to the region; β j Represents the regional weight coefficient, which is used to assign different priorities to different regions based on historical needs; D u,i represents the user demand forecast value of the i-th region; γ j Represents the user behavior weight, which adjusts the user response preference in this area; R u Indicates the user's browsing or purchasing frequency; R max Indicates the maximum reasonable frequency set by the system, used for normalized calculations.

[0054] Further, in S3, the edge device is based on the global supply and demand matrix S global and the logistics priority P of the i-th node log,i Dynamically allocate logistics resources, as shown below:

[0055]

[0056] Among them, L dist,i represents the number of logistics resources allocated to the i-th node; κ i represents the node specificity coefficient; S global,i is the global supply-demand difference;

[0057] According to the global supply and demand matrix S global , each node design will build a dynamic inventory adjustment model to adjust its inventory level to ensure that demand is met or to avoid excess inventory; the dynamic inventory adjustment model is expressed as follows:

[0058]

[0059] Among them, I adj,i Represents the adjusted inventory of the i-th node; I sc,i represents the current inventory level of the i-th node; L dist,i Indicates the number of logistics resources received by the node; λ indicates the inventory warning coefficient, which is used to indicate the adjustment in the case of abnormal inventory; represents the inverse function of transportation delay;

[0060] According to the global precision marketing strategy, a dynamic user behavior feedback push model is constructed, and push adjustment is performed in combination with real-time user behavior feedback, as shown below:

[0061] A push,i =M push,i ·(1+θ·ΔR u,i )

[0062] Among them, A push,i represents the actual push amount of the i-th node; M push,i represents the marketing push strategy, indicating the push frequency and intensity; θ represents the push feedback coefficient, indicating the sensitivity of user response; ΔR u,i Indicates the rate of change of user response, based on real-time feedback of user clicks, browsing, purchases, etc.

[0063] Furthermore, the S4 specifically includes:

[0064] S401, designing a weighted feedback model on the edge device to dynamically collect feedback data;

[0065] S402: Globally aggregate the feedback data of edge devices to generate a global feedback matrix F global ;

[0066] S403, according to the global feedback matrix F global Perform anomaly detection and filter abnormal data;

[0067] S404, design a feedback-driven global strategy adjustment model based on the processed feedback data to adjust the global supply and demand difference value S global,i Conduct feedback optimization.

[0068] 8. The marketing strategy analysis and optimization method based on big data and edge computing according to claim 7 is characterized in that the weighted feedback model is expressed as follows:

[0069] F weighted,i =α i ·(I adj,i -I sc,i )+β i ·(A push,i -A real,i)+γ i ΔL dist,i

[0070] Among them, F weighted,i Represents the weighted feedback data of the i-th node, which is used for feedback analysis and subsequent adjustment; adj,i Represents the adjusted inventory; I sc,i Indicates the inventory quantity before execution, used to calculate inventory differences; A push,i Indicates the expected amount of push. real,i Indicates the actual response behavior of the user; ΔL dist,i Indicates the difference in logistics execution, indicating the difference between expected and actual logistics resource arrival; α i ,β i ,γ i The weighting coefficients representing the feedback types, which represent the weights of inventory, marketing, and logistics respectively;

[0071] The feedback data of the edge devices are globally aggregated to generate a global feedback matrix F global , which is expressed as follows:

[0072]

[0073] Among them, F global represents the global feedback matrix, which is used to represent the overall performance of each node’s feedback; w i represents the weight of the i-th node, based on the strategic importance and historical data of the node; F weighted,i Represents the weighted feedback data of the i-th node; ∈·R urg,i represents the emergency feedback correction term, where R urg,i represents the emergency demand score of the i-th node;

[0074] The anomaly detection is represented as follows:

[0075]

[0076] Among them, E i Indicates the abnormal flag of the i-th node. When E i =1, it indicates that the node feedback is abnormal; represents the mean of the global feedback matrix; represents the standard deviation of the global feedback; λ represents the adaptive anomaly detection parameter, which is used to adjust the sensitivity of anomaly detection.

[0077] Furthermore, the S5 specifically includes:

[0078] The dual feedback data of the adaptively adjusted global supply chain scheduling plan and the global precision marketing strategy are collected, and dynamic closed-loop optimization is performed based on the dual feedback data, as shown below:

[0079]

[0080] Among them, S opt,i represents the supply chain scheduling strategy after optimization of the i-th node; S adjust,i represents the supply chain scheduling strategy after the previous step adjustment; μ i It represents the feedback adjustment coefficient, which indicates the weight of the impact of feedback on the scheduling strategy; Represents global feedback data related to inventory, reflecting inventory deviation; It represents the global feedback data related to logistics, reflecting the deviation of logistics execution; It represents the global feedback data related to marketing, reflecting the execution deviation of precision marketing;

[0081] The multi-dimensional feedback convergence model is expressed as follows:

[0082] ΔS opt,i =λ1·(I adj,i -I sc,i )+λ2·(L dist,i -L real,i )+λ3·(A push,i -A real,i )

[0083] Among them, ΔS opt,i I represents the multi-dimensional feedback optimization adjustment of the i-th node; adj,i represents the adjusted inventory of the i-th node; I sc,i Indicates the actual inventory data; L dist,i Indicates the execution volume of the first logistics scheduling; L real,i Indicates the actual execution result of the logistics; A push,i Indicates the execution volume of marketing push strategy; A real,i represents the actual response result of the user; λ1, λ2, λ3 are the weights of inventory, logistics and push feedback respectively;

[0084] The feedback iterative convergence mechanism is expressed as follows:

[0085]

[0086] in, represents the supply chain scheduling adjustment after the kth round of feedback optimization; represents the supply chain scheduling adjustment after the k-1th round of feedback optimization; ∈ represents the convergence threshold, the standard for the control system to enter a stable state.

[0087] Furthermore, after the supply chain scheduling strategy is dynamically closed, a marketing feedback adjustment model is designed to dynamically adjust the push frequency and content according to the actual response data of the user; wherein the marketing feedback adjustment model is expressed as follows:

[0088] M opt,i =M push,i +ν i ·(A real,i -A push,i )+θ i ΔA u,i

[0089] Among them, M opt,i represents the marketing push strategy after the optimization of the i-th node; M push,i Indicates the previous round of push plan; ν i A represents the push feedback coefficient, which controls the impact of user feedback on push adjustment; real,i Indicates the result of the user's actual response; A push,i Indicates the actual push volume in the last push plan; θ i ΔA u,i An adjustment item that indicates the rate of change of user behavior, indicating the adjustment of push strategy when user behavior changes significantly.

[0090] The beneficial technical effects of the present invention are at least as follows:

[0091] (1) To solve the data transmission delay and network dependency problems caused by centralized processing, the present invention deploys edge computing devices at various supply chain nodes (such as manufacturers, distributors, hospitals, etc.) so that each node can independently complete data collection, processing and optimization. Through the local processing capabilities of edge computing, supply chain nodes can analyze inventory, orders and logistics information in real time and generate local optimization strategies. This mechanism significantly reduces data transmission delays, reduces dependence on network stability, and improves the dynamic response capabilities of the supply chain.

[0092] At the same time, edge devices disperse the computing load and solve the problem of centralized processing load. The edge devices on each node take on the task of local data processing and only need to upload simplified analysis results to the central server when necessary. This greatly reduces the load on the central server, avoids the bottleneck problem of centralized processing, and ensures the efficient operation of the entire system.

[0093] (2) The core reason why precision marketing in the prior art is not effective is the lack of capturing and analyzing real-time user behavior. To solve this problem, the present invention uses edge AI devices to capture browsing, purchasing and other behavioral data of medical institutions in real time at user terminals (such as hospitals, clinics, etc.), and uses the AI ​​model in the edge device to analyze user behavior and extract personalized needs, thereby generating targeted precision marketing strategies. This personalized push mechanism based on real-time behavior analysis can significantly improve the accuracy and conversion rate of marketing, and avoid the problem of disconnection between push content and user needs that is common in traditional marketing.

[0094] In addition, edge AI also continuously optimizes marketing strategies through feedback loops, further solving the problem of lack of dynamic response and continuous optimization capabilities. Edge devices can adjust subsequent recommended content in real time based on the user's response to the pushed content (such as clicks, purchases, etc.), forming an adaptive marketing closed loop to ensure that the system can dynamically meet the user's personalized needs.

[0095] (3) The present invention realizes a complete collaborative optimization system from supply chain scheduling, inventory management to end-user demand capture and product recommendation by closely combining supply chain optimization with precision marketing strategies. Supply chain optimization no longer relies solely on inventory and logistics data, but also combines user demand data from the marketing end to ensure that inventory adjustment and product supply can better match the actual needs of the terminal market and avoid inventory backlogs and shortages. At the same time, precision marketing strategies are also optimized in real time according to the dynamic adjustment of the supply chain, making product recommendations more accurate and timely. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] The present invention is further described using the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.

[0097] Figure 1 This is a flow chart of the marketing strategy analysis and optimization method based on big data and edge computing of the present invention. DETAILED DESCRIPTION

[0098] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0099] In one or more embodiments, Figure 1 As shown, the marketing strategy analysis and optimization method based on big data and edge computing of the present invention is disclosed, and the method includes steps S1-S5:

[0100] S1. Construct a data collection model to collect data, wherein the data includes supply chain data and end-user data, and pre-process the collected data and perform weighted integration on the pre-processed data to obtain integrated features, dynamically adjust the local supply chain according to the integrated features to obtain a supply chain dynamic adjustment model, and construct a user demand prediction model according to the end-user data, and finally combine the outputs of the supply chain dynamic adjustment model and the user demand prediction model to generate a local supply chain adjustment and demand matching strategy;

[0101] The data collection model is constructed as follows:

[0102] Install edge devices at supply chain edge devices and end-user data collection points to collect supply chain data and end-user data;

[0103] The supply chain data includes inventory data, order data and logistics data; the terminal user data includes user browsing behavior and user purchasing behavior.

[0104] Specifically, edge devices are installed at data collection points of supply chain nodes (manufacturers, distributors, hospitals, etc.) and end users (such as hospitals, clinics, etc.). The present invention defines three types of data collected from each supply chain node:

[0105] I sc : Inventory data, including current inventory and historical inventory data.

[0106] O sc : Order data, representing the current order status, arrival time of historical orders, etc.

[0107] L sc :Logistics data, including real-time tracking information and transportation time of logistics.

[0108] For end-user behavior data:

[0109] B u : User browsing behavior, including the frequency and duration of product page visits.

[0110] P u : User’s purchase behavior, indicating the user’s purchase record.

[0111] Real-time data collection is achieved through sensors and communication modules of edge devices. The collection frequency is determined by the dynamic requirements of the system, for example, an update every 5 minutes.

[0112] Furthermore, in the process of data collection, noise and abnormal data will inevitably appear. The present invention first performs denoising on the collected data. In order to remove extreme data and abnormal fluctuations, the present invention introduces a window-based dynamic local anomaly detection model. sc Denoising using local deviation detection:

[0113]

[0114] in, is the processed inventory data, I sc,tis the inventory data at the current time, n is the window size, and the average inventory within 7 days is usually selected as the benchmark. The function of this formula is to eliminate the noise and extreme values ​​of short-term inventory data and retain the long-term trend of inventory.

[0115] For user behavior data B u and P u , use the behavior pattern detection model to filter abnormal user behavior data. For example, records with short browsing time (e.g. less than 2 seconds) or abnormal purchase frequency are directly marked as abnormal.

[0116] Furthermore, for each type of data (inventory, order, logistics and user behavior data), the present invention performs feature extraction and aggregation to prepare input for subsequent supply chain scheduling and marketing strategy generation. The extracted features include:

[0117] Inventory Features: in is the inventory change rate, calculated as:

[0118]

[0119] Order characteristics: F O = {O sc,t ,ΔO sc,t}, represents the order volume and order change rate, where ΔO sc,t Also based on changes in order volume.

[0120] Logistics characteristics: F L ={L sc,t ,ΔL sc,t}, describing the real-time status of logistics.

[0121] User behavior characteristics: F B = {B u ,ΔB u} and F P = {P u ,ΔP u}, where ΔB u and ΔP u Represent the change rates of user browsing and purchasing behaviors respectively.

[0122] These feature values ​​will be updated in real time through edge computing devices and aggregated in segments. On a daily basis, inventory fluctuations, order increases and decreases, logistics delays, etc. will be summarized to ensure the timeliness and real-time nature of the data. The aggregated data will be further used for subsequent optimization of supply chains and marketing strategies.

[0123] Furthermore, since the supply chain data and user behavior data have different weights for subsequent optimization, the present invention designs a weighted integration model for integrating feature data. Define the weight parameter αI , α O , α L Represent the importance weights of inventory, order, and logistics data, respectively, while the weight of user behavior features is β B and β P .

[0124] F agg =α I F I +α O F O +α L F L +β B F B +β P F P

[0125] These weights can be set according to the system's priorities, such as when dealing with inventory pressure, α I Weight greater than α O The integrated F agg As the final output of this step, it is provided to the next step for the generation of supply chain scheduling and precision marketing strategies.

[0126] Furthermore, the integrated supply chain and user behavior characteristics F agg , which contains the inventory data of the supply chain nodes (F I ), order data (F O ), logistics data (F L ) and the user's browsing behavior (F B ) and purchasing behavior (F P ). agg The specific meaning can be expressed as:

[0127] F agg =α I F I +α O F O +α L F L +β B F B +β P F P

[0128] Among them, α I ,α O ,α L ,β B ,β P They are the weights of the corresponding features (dynamically adjusted according to different needs), F I represents the inventory data characteristics, F O Represents the order data characteristics, F LRepresents the characteristics of logistics data, F B and F P Represents user behavior characteristics (browsing and purchasing behavior). This step uses these characteristics to dynamically adjust local inventory and logistics and predict user demand.

[0129] Furthermore, the present invention needs to optimize the inventory and logistics scheduling of supply chain nodes based on real-time inventory, orders, and logistics information, so that the local supply chain can adaptively adjust inventory supply to cope with fluctuations in user demand. An innovative inventory adjustment formula is proposed, combining the current status of the supply chain and expected future demand, and designing a special adjustment item to cope with supply chain uncertainty. The formula is as follows:

[0130]

[0131] Among them, I sc,t Indicates the current inventory level. sc,t Indicates the order increment at the current moment, reflecting the impact of external orders on inventory. sc,t Represents the current logistics data, and adjusts the impact of logistics on inventory through time delay T. γ u Represents the user demand adjustment coefficient, which is used to directly incorporate the results of user behavior prediction into inventory adjustment. u,t Represents the predicted value of user demand (calculated by the next formula).

[0132] The innovation in this formula is the addition of γ u D u,t This item embeds the direct impact of user demand on inventory into the inventory adjustment process by considering the user's future demand. In this way, the local supply chain can not only adjust inventory according to existing orders and logistics status, but also predict and respond to potential user demand fluctuations in advance.

[0133] Furthermore, for the user's behavior data (browsing behavior F B and purchasing behavior P ), the present invention establishes a user behavior prediction model to capture future demand in advance. By comparing the change rate of historical behavior with current behavior, the present invention can infer the user's future purchasing needs. Demand prediction formula:

[0134] D u,t+1 =β B ·ΔB u +β P ΔP u -λ·R u

[0135] Among them, D u,t+1 Indicates the predicted value of user demand at the next moment. ΔB uIndicates the rate of change of user browsing behavior (e.g., the frequency of visiting similar medical device products increases within a short period of time). ΔP u Indicates the rate of change of user purchase behavior (e.g., indicates an increase in the purchase history of a certain type of product). λ represents the regularization term coefficient, which controls the noise or extreme behavior in user behavior modeling. u Inverse function of the last purchase time This item is used to suppress user demand forecasts that purchase too frequently.

[0136] In this formula, by introducing λ·R u This regularization term can effectively avoid the excessive impact of extreme user behaviors (such as frequent clicks or invalid browsing) on ​​the overall demand forecast. This is an innovation designed specifically for medical device marketing scenarios and can ensure the stability and accuracy of the forecast results.

[0137] Furthermore, based on the inventory adjustment formula and the user demand modeling results, the present invention generates a local supply chain adjustment and demand matching strategy. The key to this strategy is to balance supply and demand in real time, ensure that there will be no overstock or shortage of inventory at the supply chain node, and prepare in advance to meet the potential needs of users. The final adjustment strategy formula is:

[0138] S opt =I sc,t+1 -D u,t+1

[0139] Among them, S opt Indicates the difference between local inventory and demand, used to indicate whether it is necessary to increase orders, adjust logistics, or adjust inventory. opt >0, indicating excess inventory, it is necessary to reduce purchases or increase promotion efforts; when S opt <0, it means that the inventory is insufficient and it may be necessary to speed up replenishment or reallocate logistics resources. Through this real-time adjustment, supply chain nodes can dynamically respond to changes in market demand.

[0140] Furthermore, the final output regulation strategy S opt And the user demand prediction result D u,t+1 It will be used as input to generate global supply chain scheduling and precision marketing strategies in the next step. opt The supply-demand difference value not only helps the global system to conduct macro-control, but also provides user demand forecast information to facilitate the generation of accurate push strategies and ensure that subsequent marketing strategies are coordinated with supply chain optimization.

[0141] S2. Generate a global supply chain scheduling plan and a global precision marketing strategy based on the output of the local supply chain adjustment and demand matching strategy and the user demand prediction model. At the same time, the generated global supply chain scheduling plan and the generated global precision marketing strategy are designed to dynamically adjust the marketing push model based on user preferences during operation, and perform adaptive adjustments based on user feedback to cope with real-time changes.

[0142] Specifically, each supply chain node generates a local supply-demand difference value S opt and user demand forecast value D u,t+1 These data will be used as input in this step for global optimization.

[0143] Furthermore, the present invention designs a global supply and demand optimization model, which integrates the supply and demand differences and demand forecasts of each supply chain node to generate a global supply and demand matrix S global . This matrix can describe the supply and demand relationship of each node in the entire supply chain system. Global supply and demand matrix formula:

[0144]

[0145] Among them, S global represents the global supply and demand matrix, which is used to describe the overall supply and demand balance of each node in the system. i It represents the weight coefficient, which represents the priority of the i-th node in the supply chain system, usually determined based on inventory, geographical location, and strategic importance. opt,i represents the local supply and demand difference of the i-th node, which comes from the output of the previous step. η represents the global adjustment coefficient, which controls the impact of user demand changes on the supply and demand matrix. ΔD u,i represents the user demand change rate of the i-th node, that is, the prediction of future user demand, which comes from the previous step. The special design of this formula is to combine the local supply and demand difference with the user demand prediction and use the weight w i This ensures that the system not only considers the current inventory status, but also optimizes in advance based on future demand forecasts.

[0146] Furthermore, the global supply and demand matrix S global After generation, the present invention uses the matrix to optimize resource scheduling. The goal of the scheduling model is to balance the supply and demand of each node by adjusting the allocation of logistics, production and inventory. Here, the present invention introduces a new scheduling priority model to determine the resource allocation priority of each node. Global scheduling priority formula:

[0147]

[0148] Among them, P log,iIt represents the logistics scheduling priority of the i-th node, which determines the urgency of resource allocation at this node. global,i Represents the global supply-demand difference value of the i-th node, representing the current supply-demand imbalance state of the node. i Represents the logistics time delay of the i-th node, which is usually determined by historical data or real-time logistics information. i represents the uncertainty penalty term, which indicates the uncertainty degree of the supply chain of the i-th node. For nodes with large historical fluctuations, this value should be higher to increase the scheduling priority.

[0149] In the traditional logistics scheduling formula, this paper introduces the uncertainty penalty term μ i This item is used to model historical volatility or possible future demand fluctuations to solve resource allocation errors in the logistics chain caused by large changes in demand. This is particularly applicable to the medical device industry, because medical demand can change dramatically during epidemics or special circumstances.

[0150] Furthermore, in addition to supply chain scheduling, global user demand forecast data will also be used to generate precision marketing strategies. Combining user behavior data with global demand forecasts, the present invention designs a marketing push model that is dynamically adjusted based on user preferences. This model can adjust advertising strategies or product recommendations based on changes in user demand. Precision marketing push model:

[0151]

[0152] Among them, M push,i Indicates the precision marketing push strength of the i-th node, which is used to determine the frequency of advertising or product recommendations to the area. j Represents the regional weight coefficient, which is used to assign different priorities to different regions based on historical needs. u,i Represents the user demand forecast value of the i-th region, which comes from the previous step. j Represents the user behavior weight, which adjusts the user response preference in this area. u Indicates the user's browsing or purchasing frequency (such as a user's recent visit history). max Represents the maximum reasonable frequency set by the system, which is used for normalized calculation. This formula innovatively combines user demand prediction and dynamic feedback of user behavior. By adjusting the regular term of browsing / purchase frequency, it can prevent users from being disturbed by frequent marketing push, thereby improving user participation and conversion rate. This model is particularly suitable for personalized marketing in the medical device industry, because medical needs are usually cyclical and urgent.

[0153] Furthermore, the global system needs to continuously adjust the supply chain and marketing strategies during operation to cope with real-time changes. Therefore, the system will make adaptive adjustments based on the execution results of each node and market feedback. The present invention introduces a global feedback adjustment mechanism to dynamically adjust the global scheduling and marketing push efforts by monitoring the supply and demand status of each node and user behavior response.

[0154] Global feedback adjustment formula:

[0155]

[0156] Among them, F adj,i Represents the feedback adjustment amount for the i-th node. global,i and Represents the deviation between the current supply and demand difference of the i-th node and the global predicted supply and demand. push,i and M real,i It represents the difference between the current marketing push strength and the actual user response. α,δ represent the feedback adjustment coefficients, which are used to balance the proportion of supply and demand adjustment and marketing adjustment. Through this feedback adjustment formula, the system can make adaptive adjustments to the supply chain and marketing strategies according to the actual execution situation, ensuring that the system can quickly respond to market changes and maintain the efficiency of the supply chain and marketing push.

[0157] Finally, the global supply chain scheduling plan and precision marketing strategy generated by the system will serve as input for the next step of execution. log,i And precision marketing push strength M push,i It will be sent to each edge node for execution, thereby ensuring the rational allocation of supply chain resources and accurate response to user needs.

[0158] In this step, the present invention realizes the dynamic integration of global supply chain and marketing strategy through the global supply and demand balance formula, logistics priority model and precision marketing push model. The supply and demand adjustment terms and user behavior regularization terms are innovatively introduced to ensure that the system can balance resource allocation and accurately push according to the real needs of users. At the same time, the global feedback mechanism ensures the flexible adaptation of the system and the dynamic optimization of supply chain and marketing push. This step effectively solves the complex supply chain management and personalized marketing problems in the medical device industry, and provides strong technical support for global coordination and efficient operation.

[0159] S3. Combine the supply and demand situation, logistics conditions and global priorities of each edge device with the global supply chain scheduling plan and global precision marketing strategy adjustment to be executed collaboratively on each edge device.

[0160] Specifically, according to P log,i and S global,i, edge devices need to perform dynamic allocation of logistics resources. The allocation of logistics resources needs to consider the supply and demand of each node, logistics conditions (such as transportation time, cost, etc.) and global priority. Here, the present invention proposes an innovative logistics allocation model based on dynamic priority:

[0161] Logistics distribution formula:

[0162]

[0163] Among them, L dist,i Represents the number of logistics resources allocated to the i-th node (such as transport vehicles, transportation capacity, storage space, etc.). global,i Represents the global supply-demand difference of the i-th node, representing the urgency of the demand at the current node. log,i Represents the logistics priority of the i-th node, which comes from the global scheduling priority model in step 3. i It represents the node specificity coefficient. Considering that some nodes (such as nodes in urgent need of medical care) may have special needs (such as epidemic prevention and control, hospitals in urgent need of medical equipment), these nodes need to be given priority in logistics resource allocation. This formula adds the node specificity coefficient κ to the traditional logistics allocation model. i , ensuring that certain nodes in the medical device industry have higher logistics response priority. Through this allocation model, logistics resources can be dynamically adjusted according to the actual needs and special circumstances of each node, which is particularly suitable for quickly responding to high-priority nodes when resources are tight.

[0164] Furthermore, based on the global supply and demand balance solution S global , each node will adjust its inventory level to ensure that demand is met or to avoid overstocking. The core of inventory adjustment is how to dynamically balance local inventory and global demand and respond to sudden demand (such as a sudden shortage of a medical device) in a timely manner. This invention introduces an innovative dynamic inventory adjustment model:

[0165] Inventory reconciliation formula:

[0166]

[0167] Among them, I adj,i Represents the adjusted inventory of the i-th node. sc,i Indicates the current inventory level of the i-th node. dist,i Indicates the amount of logistics resources received by the node, which comes from the above logistics allocation formula. global,i Represents the global supply-demand difference, which is used to adjust the supply-demand balance of the node. λ represents the inventory warning coefficient, which is used to indicate the adjustment in abnormal inventory conditions (such as the need for rapid replenishment when the inventory is very low). The inverse function of transportation delay is used to indicate that the shorter the logistics time, the faster the inventory adjustment speed. This formula introduces the inventory warning term λ and the logistics time delay correction term Ensure that the system can respond quickly in special scenarios. For example, when the inventory of a certain node approaches the warning line, the system will accelerate the deployment of logistics resources to avoid supply chain disruptions, which is particularly suitable for medical equipment inventory management.

[0168] Furthermore, for the execution of marketing strategies, the present invention designs a dynamic user behavior feedback push model to ensure that users receive personalized product recommendations. This model is based on the precise marketing strategy M generated in the previous step. push,i , and push adjustments based on real-time user behavior feedback:

[0169] Precision push formula:

[0170] A push,i =M push,i ·(1+θ·ΔR u,i )

[0171] Among them, A push,i Indicates the actual push volume (advertising volume or recommended product volume) of the i-th node. push,i represents the marketing push strategy from step 3, indicating the push frequency and intensity. θ represents the push feedback coefficient, indicating the sensitivity of user response. ΔR u,i Indicates the rate of change of user response, based on real-time feedback of user clicks, browsing, purchases, etc. This push model dynamically adjusts the push frequency and content by combining the user's real-time response data to ensure that pushes are not too many or too few, thus avoiding user fatigue. At the same time, due to the particularity of medical devices, accurate push can increase product exposure and conversion rate.

[0172] Furthermore, during the execution process, edge devices are not only responsible for logistics scheduling, inventory adjustment and precise push implementation, but also need to collect feedback data in real time. These feedbacks include inventory changes, logistics arrival status, user clicks and purchase behaviors, etc. The data will be uploaded to the central system as the basis for the next step of global optimization. Feedback data formula:

[0173] F exec,i =α·(I adj,i -I sc,i )+β·(A push,i -A real,i )

[0174] Among them, F exec,i Represents the feedback data of the i-th node, which is used to measure the execution deviation. α and β represent the weights of inventory and marketing push feedback respectively. adj,i Indicates the result after inventory adjustment. sc,iIndicates the actual inventory status. push,i Indicates the execution result of the push plan. real,i Represents the actual response of the user. Through this feedback formula, the system can measure the deviation in execution of each node, adjust the logistics and push strategy in real time, and ensure that the execution process of each node can be quickly corrected. This is especially important in the medical device supply chain, because any delay or execution deviation may directly affect the timely supply of medical equipment.

[0175] Furthermore, during the execution process, edge devices generate feedback data based on the actual results of logistics scheduling, inventory adjustment, and precise push notifications. These data will be returned to the central system in real time for subsequent global optimization and adjustment. Ultimately, through the feedback loop, the system can further improve scheduling efficiency and precise push notification effects in the next round of supply chain and marketing optimization.

[0176] S4. Collect feedback data from each edge device and perform analysis and global aggregation to obtain an aggregated global feedback matrix, and adaptively adjust the global supply chain scheduling plan and global precision marketing strategy according to the global feedback matrix.

[0177] Specifically, in order to ensure the accuracy and consistency of feedback data, the present invention designs a dynamic weighted feedback model to pre-process the feedback data to ensure the credibility of the data. This model not only eliminates noise and anomalies, but also performs weighted processing according to the priorities of different nodes and feedback types. Dynamic weighted feedback formula:

[0178] F weighted,i =α i ·(I adj,i -I sc,i )+β i ·(A push,i -A real,i )+γ i ΔL dist,i

[0179] Among them, F weighted,i Represents the weighted feedback data of the i-th node, which is used for feedback analysis and subsequent adjustment. adj,i Represents the adjusted inventory level. sc,i Indicates the inventory quantity before execution, used to calculate inventory differences. push,i Indicates the expected amount of push. real,i Indicates the actual response behavior of the user. ΔL dist,i Indicates the difference in logistics execution, indicating the difference between expected and actual logistics resource arrival. i ,β i ,γ iThe weighted coefficients represent the feedback types, which represent the weights of inventory, marketing, and logistics. This formula ensures that different types of feedback can be reasonably considered in the system by weighting the feedback data. Especially in the medical device supply chain, the priorities of logistics, inventory, and user needs may be different, so dynamic adjustment of weights is necessary. For example, when the urgent inventory demand of a node takes priority, α can be increased. i The weight of .

[0180] Furthermore, the present invention globally aggregates the feedback data of each node to generate a global feedback matrix F global , providing a basis for subsequent global optimization. To ensure the system's sensitivity to global data, the present invention introduces a regional priority correction model for feedback data, assigning different feedback weights according to node importance. Global feedback aggregation formula:

[0181]

[0182] Among them, F global Represents the global feedback matrix, which is used to represent the overall performance of each node’s feedback. i represents the weight of the ith node, based on the strategic importance of the node and historical data (for example, important nodes such as hospitals and emergency medical centers have higher weights). weighted,i Represents the weighted feedback data of the i-th node. ∈·R urg,i represents the emergency feedback correction term, where R urg,i R represents the emergency demand score of the i-th node (e.g. emergency dispatch of medical equipment). urg,i The addition of this item enables the medical device supply chain to better respond to unexpected demands. Especially during epidemics or medical emergencies, the urgent demand for certain medical equipment can be quickly identified by the system and given priority through this amendment item.

[0183] Furthermore, in the global feedback data, the present invention needs to perform anomaly detection to filter out abnormal data that may be caused by logistics delays, atypical user behavior, etc. The present invention proposes an adaptive threshold feedback anomaly detection model, which automatically adjusts the anomaly detection threshold according to the fluctuation of feedback data. Anomaly detection formula:

[0184]

[0185] Among them, E i Indicates the abnormal flag of the i-th node. When E i =1, it indicates that the node feedback is abnormal. global Represents the mean of the global feedback matrix. represents the standard deviation of global feedback. λ represents the adaptive anomaly detection parameter, which is used to adjust the sensitivity of anomaly detection. This anomaly detection model dynamically adjusts the threshold to ensure that the system can effectively filter outliers in the feedback and avoid the impact of abnormal data on global optimization. Especially in the case of frequent logistics delays or large fluctuations in user behavior, this formula can ensure the robustness of feedback data.

[0186] Furthermore, by adjusting the global feedback matrix F global Based on the analysis, the system can optimize the next round of supply chain scheduling and marketing push strategy. To this end, the present invention designs a feedback-driven global strategy adjustment model, which makes adaptive adjustments to the supply chain and marketing push based on feedback data. Feedback-driven strategy optimization formula:

[0187] S adjust,i =S global,i +δ i ·F global,i

[0188] Among them, S adjust,i S represents the adjusted supply chain scheduling strategy of the i-th node. global,i Represents the global initial supply chain scheduling solution. δ i Represents the feedback response coefficient, which controls the adjustment strength of the feedback data on the next round of scheduling. global,i Represents the feedback data of the i-th node in the global feedback matrix. This model can adjust the supply chain scheduling strategy in real time according to the feedback data of each node, which is especially suitable for the scenario of medical device supply chain. Through feedback-driven strategy optimization, the system can automatically adjust the inventory adjustment, logistics distribution and marketing push of each node to ensure the flexibility of the system in a complex environment.

[0189] The final adjusted strategy S adjust,i It will serve as the input for the next round of supply chain scheduling and marketing push. The results of the feedback analysis directly affect the next execution plan. The system continuously optimizes the supply chain and marketing strategies through a feedback closed loop to ensure real-time response to dynamic changes in the medical device supply chain.

[0190] This step realizes effective analysis of logistics scheduling, inventory adjustment and marketing push execution results by designing a dynamic weighted feedback model, a global feedback aggregation formula and anomaly detection mechanism. The system ensures the accuracy of feedback through dynamic weighting and global aggregation of feedback data, and filters abnormal data through an innovative anomaly detection model. Ultimately, through the feedback-driven strategy optimization model, the system can adaptively adjust the supply chain and marketing strategies to achieve closed-loop optimization, which is particularly important in the medical device industry and can cope with sudden changes in demand and emergency medical scenarios.

[0191] S5. Collect dual feedback data of the adaptively adjusted global supply chain scheduling plan and the global precision marketing strategy, and perform dynamic closed-loop optimization based on the dual feedback data. At the same time, during the dynamic closed-loop optimization process, build a multi-dimensional feedback convergence model to ensure that the supply chain data and end-user data can reach a convergence state in all dimensions. At the same time, design a feedback iterative convergence mechanism to ensure that convergence is maintained after iteration and to maintain the global optimal state.

[0192] Specifically, in this step, the present invention designs a dynamic feedback closed-loop optimization model, which processes the dual feedback data of supply chain scheduling and marketing strategy, generates an optimized strategy, and uses it for the next round of execution. Closed-loop feedback optimization formula:

[0193]

[0194] Among them, S opt,i Represents the optimized supply chain scheduling strategy of the i-th node, which is used to guide the next round of execution. adjust,i Represents the supply chain scheduling strategy adjusted in the previous step. i It represents the feedback adjustment coefficient, which indicates the weight of the impact of feedback on the scheduling strategy. Represents global feedback data related to inventory, reflecting inventory deviation. It represents the global feedback data related to logistics, reflecting the deviation of logistics execution. Represents the global feedback data related to marketing, reflecting the execution deviation of precision marketing. This formula innovatively combines supply chain scheduling, logistics scheduling and marketing feedback together. i Control the impact of feedback data on the next round of supply chain scheduling. This comprehensive feedback can effectively handle the interdependence between supply chain and marketing strategies, ensuring that the system can optimize logistics, inventory and marketing push at the same time. This model is particularly suitable for the medical device supply chain because the supply chain in this industry must flexibly respond to complex market demands, logistics delays and changes in user behavior.

[0195] In addition to optimizing supply chain scheduling, the system also needs to optimize the feedback of precision marketing push strategies. The present invention designs a marketing feedback adjustment model, which dynamically adjusts the frequency and content of push based on the actual response data of users. Marketing feedback optimization formula:

[0196] M opt,i =M push,i +ν i ·(A real,i -A push,i )+θ i ΔA u,i

[0197] Among them, M opt,irepresents the marketing push strategy after the optimization of the i-th node. push,i Indicates the last round of push plan. i Represents the push feedback coefficient, which controls the impact of user feedback on push adjustment. real,i Indicates the actual response of users, based on clicks, browsing, purchases, etc. push,i Indicates the actual push volume in the last push plan. i ΔA u,i The adjustment item representing the user behavior change rate indicates the push strategy adjustment when the user behavior changes significantly. This formula combines the user's actual response data (A real,i -A push,i ) and the rate of change of user behavior ΔA u,i , dynamically optimize the next round of push strategies. Marketing push in the field of medical devices is different from general consumer products. The accuracy of push content and the control of push frequency are particularly critical. Through this formula system, adaptive adjustments can be made according to changes in user needs.

[0198] Since the system performs feedback optimization in multiple dimensions (such as inventory, logistics, marketing, etc.), the present invention needs to ensure that the optimization between these dimensions is coordinated and consistent to avoid local optimality or over-adjustment. To this end, the present invention proposes a multi-dimensional feedback convergence model to ensure that the system can reach a convergence state in all dimensions. Multi-dimensional feedback convergence formula:

[0199] ΔS opt,i =λ1·(I adj,i -I sc,i )+λ2·(L dist,i -L real,i )+λ3·(A push,i -A real,i )

[0200] Among them, ΔS opt,i Represents the multi-dimensional feedback optimization adjustment of the i-th node. adj,i Represents the adjusted inventory of the i-th node. sc,i Indicates actual inventory data. dist,i Indicates the execution volume of logistics scheduling. real,i Indicates the actual execution result of logistics. push,i Indicates the execution volume of marketing push strategies. real,iRepresents the actual response result of the user. λ1,λ2,λ3 represent the weights of inventory, logistics and push feedback respectively. Through the multi-dimensional adjustment of this formula, the system can find the global optimal solution in inventory, logistics and precision marketing push. Especially in the medical device supply chain, inventory shortages, logistics delays and changes in user demand often affect each other, and the system needs to integrate these dimensions for linkage optimization. By setting different weights, the system can adjust the focus of optimization according to the current situation.

[0201] Furthermore, during multiple feedback adjustments, the system may frequently adjust. To avoid over-adjustment or feedback oscillation, the present invention designs a feedback iterative convergence mechanism to ensure that the system can converge stably after multiple iterations and remain in the global optimal state. Convergence judgment formula:

[0202]

[0203] in, Represents the supply chain scheduling adjustment after the kth round of feedback optimization. Represents the supply chain scheduling adjustment after the k-1th round of feedback optimization. ∈ represents the convergence threshold, which is the standard for the control system to enter a stable state. This formula determines whether the system has entered a convergence state by judging whether the change in each round of adjustment is less than the threshold ∈. After convergence, the system will enter a stable state and will no longer make large adjustments to avoid over-adjustment. The supply chain and marketing push strategies of the medical device industry require high stability, especially after an emergency, the system needs to return to stable operation as soon as possible.

[0204] Finally, the supply chain scheduling strategy S after feedback optimization opt,i and marketing push strategy M opt,i This will be used as the execution input for the next round and will continue into the next round of closed-loop optimization. Each feedback and iteration ensures the continuous optimization and self-adaptation of the system, especially in the scenario of the medical device supply chain. This closed-loop mechanism can best cope with market changes, inventory fluctuations, and changes in user demand.

[0205] In this step, the present invention realizes the continuous optimization of the medical device supply chain and precision marketing push by designing a dynamic feedback closed-loop optimization model, a multi-dimensional feedback convergence model, and a feedback convergence control mechanism. Combining logistics, inventory, and user behavior feedback data, the system can gradually find the global optimal solution to ensure the accuracy and efficiency of supply chain scheduling and marketing push. Through multiple iterative optimizations, the system can not only respond to sudden market changes, but also ensure long-term operational stability, so that the entire supply chain closed loop can achieve dynamic adaptive optimization, which is particularly suitable for complex supply chain management scenarios in the field of medical devices.

[0206] The above disclosures are only some preferred embodiments of the present invention, which certainly cannot be used to limit the scope of rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiments and equivalent changes made according to the claims of the present invention are still within the scope of the invention.

Claims

1. A marketing strategy analysis and optimization method based on big data and edge computing, characterized in that: The method comprises: S1. Construct a data collection model to collect data, wherein the data includes supply chain data and end-user data, and pre-process the collected data and perform weighted integration on the pre-processed data to obtain integrated features, dynamically adjust the local supply chain according to the integrated features to obtain a supply chain dynamic adjustment model, and construct a user demand prediction model according to the end-user data, and finally combine the outputs of the supply chain dynamic adjustment model and the user demand prediction model to generate a local supply chain adjustment and demand matching strategy; The data collection model is constructed as follows: Install edge devices at supply chain edge devices and end-user data collection points to collect supply chain data and end-user data; The supply chain data includes inventory data, order data and logistics data; the terminal user data includes user browsing behavior and user purchasing behavior; S2. Generate a global supply chain scheduling plan and a global precision marketing strategy based on the output of the local supply chain adjustment and demand matching strategy and the user demand prediction model. At the same time, the generated global supply chain scheduling plan and the generated global precision marketing strategy are designed to dynamically adjust the marketing push model based on user preferences during operation, and make adaptive adjustments based on user feedback to cope with real-time changes; S3, combining the supply and demand situation, logistics conditions and global priority of each edge device with the global supply chain scheduling plan and global precision marketing strategy adjustment for coordinated execution at each edge device; S4. Collect feedback data from each edge device, analyze and globally aggregate it to obtain an aggregated global feedback matrix, and adaptively adjust the global supply chain scheduling plan and global precision marketing strategy according to the global feedback matrix; S5. Collect dual feedback data of the adaptively adjusted global supply chain scheduling plan and the global precision marketing strategy, and perform dynamic closed-loop optimization based on the dual feedback data. At the same time, during the dynamic closed-loop optimization process, build a multi-dimensional feedback convergence model to ensure that the supply chain data and end-user data can reach a convergence state in all dimensions. At the same time, design a feedback iterative convergence mechanism to ensure that convergence is maintained after iteration and to maintain the global optimal state.

2. The marketing strategy analysis and optimization method based on big data and edge computing according to claim 1 is characterized in that: In S1, the preprocessing is: using local deviation detection to denoise the inventory data, and then using a behavior pattern detection model to filter abnormal user behavior data; Extract features from supply chain data and end-user data, including: Inventory Features F I , calculated as follows: F I ={I sc,t ,ΔI sc,t } Among them, I sc,t is the inventory data after denoising, ΔI sc,t is the inventory change rate, calculated as: Order feature F O : F O ={O sc,t ,ΔO sc,t } Among them, sc Indicates order data, ΔO sc,t represents the order change rate, ΔO sc,t Also based on changes in order data; Logistics Features F L : F L ={L sc,t ,ΔL sc,t } User behavior characteristics F B : F B = {B u ,ΔB u } and F P = {P u ,ΔP u } Among them, B u represents the user's browsing behavior, P u represents the user's purchasing behavior, ΔB u and ΔP u Represent the change rates of user browsing and purchasing behaviors respectively; The weighted integration of the preprocessed data is expressed as follows: F agg =a I F I +a O F O +a L F L +b B F B +b P F P Among them, F agg represents the integrated supply chain and user behavior characteristics, and the weight parameter α I , α O , α L They represent the importance weights of inventory, order, and logistics data respectively, and the weight of user behavior characteristics is β B and β P .

3. The marketing strategy analysis and optimization method based on big data and edge computing according to claim 2 is characterized in that: The supply chain dynamic adjustment model is expressed as follows: Among them, I sc,t Indicates the current inventory level; O sc,t Indicates the order increment at the current moment, reflecting the impact of external orders on inventory; L sc,t Represents the current logistics data, and adjusts the impact of logistics on inventory through time delay T; γ u represents the user demand adjustment coefficient, which is used to directly incorporate the results of user behavior prediction into inventory adjustment; D u,t It represents the predicted value of user demand, obtained using the user demand model; The user demand prediction model is constructed based on user behavior data, which is expressed as follows: D u,t+1 =b B ·D u +b P ·ΔP u -λ·R u Among them, D u,t+1 Indicates the predicted value of user demand at the next moment; ΔB u Indicates the rate of change of user browsing behavior; ΔP u represents the rate of change of user purchase behavior); λ represents the regularization coefficient, which controls the noise or extreme behavior in user behavior modeling; R u Inverse function of the last purchase time This item is used to suppress user demand forecasts that purchase too frequently; According to the results of the supply chain dynamic adjustment model and the user demand prediction model, a local supply chain adjustment and demand matching strategy is generated, which is expressed as follows: S opt =I sc,t+1 -D u,t+1 Among them, S opt Indicates the difference between local inventory and demand, which is used to indicate whether it is necessary to increase orders, adjust logistics, or adjust inventory; When S opt >0, indicating excess inventory, it is necessary to reduce purchases or increase promotion efforts; when S opt <0, it means that the inventory is insufficient and it is necessary to speed up replenishment or reallocate logistics resources.

4. The marketing strategy analysis and optimization method based on big data and edge computing according to claim 1 is characterized in that: The global supply chain scheduling scheme integrates the supply and demand differences and demand forecasts of each supply chain node to generate a global supply and demand matrix S global , and then according to the global supply and demand matrix S global Build a scheduling priority model to determine the resource allocation priority of each node by adjusting the allocation of logistics, production and inventory; Among them, the global supply and demand matrix S global It is expressed as follows: Among them, S global represents the global supply and demand matrix, which is used to describe the overall supply and demand balance of each node in the system; w i Represents the weight coefficient, which represents the priority of the i-th node in the supply chain system. opt,i represents the local supply-demand difference of the i-th node; η represents the global adjustment coefficient, which controls the impact of user demand changes on the supply-demand matrix; ΔD u,i represents the user demand change rate of the i-th node, that is, the prediction of future user demand; The scheduling priority model is expressed as follows: Among them, P log,i represents the logistics scheduling priority of the i-th node, which determines the urgency of resource allocation at this node; S global,i represents the global supply-demand difference value of the i-th node, representing the current supply-demand imbalance state of the node; T i represents the logistics time delay of the i-th node; μ i Represents the uncertainty penalty term, which indicates the uncertainty degree of the supply chain of the i-th node.

5. The marketing strategy analysis and optimization method based on big data and edge computing according to claim 1 is characterized in that: The global precision marketing strategy combines user behavior data with global demand forecasts to design a marketing push model that is dynamically adjusted based on user preferences, as shown below: Among them, M push,i represents the precision marketing push strength of the i-th node, which is used to determine the frequency of advertising or product recommendations to the region; β j Represents the regional weight coefficient, which is used to assign different priorities to different regions based on historical needs; D u,i represents the user demand forecast value of the i-th region; γ j Represents the user behavior weight, which adjusts the user response preference in this area; R u Indicates the user's browsing or purchasing frequency; R max Indicates the maximum reasonable frequency set by the system, used for normalized calculations.

6. The marketing strategy analysis and optimization method based on big data and edge computing according to any one of claims 4-5, characterized in that: In S3, the edge device is based on the global supply and demand matrix S global and the logistics priority P of the i-th node log,i Dynamically allocate logistics resources, as shown below: Among them, L dist,i represents the number of logistics resources allocated to the i-th node; κ i represents the node specificity coefficient; S global,i is the global supply-demand difference; According to the global supply and demand matrix S global , each node design will build a dynamic inventory adjustment model to adjust its inventory level to ensure that demand is met or to avoid excess inventory; the dynamic inventory adjustment model is expressed as follows: Among them, I adj,i Represents the adjusted inventory of the i-th node; I sc,i represents the current inventory level of the i-th node; L dist,i Indicates the number of logistics resources received by the node; λ indicates the inventory warning coefficient, which is used to indicate the adjustment in the case of abnormal inventory; represents the inverse function of transportation delay; According to the global precision marketing strategy, a dynamic user behavior feedback push model is constructed, and push adjustment is performed in combination with real-time user behavior feedback, as shown below: A push,i =M push,i ·(1+θ·ΔR u,i ) Among them, A push,i represents the actual push amount of the i-th node; M push,i represents the marketing push strategy, indicating the push frequency and intensity; θ represents the push feedback coefficient, indicating the sensitivity of user response; ΔR u,i Indicates the rate of change of user response, based on real-time feedback of user clicks, browsing, purchases, etc.

7. The marketing strategy analysis and optimization method based on big data and edge computing according to claim 1 is characterized in that: The S4 specifically includes: S401, designing a weighted feedback model on the edge device to dynamically collect feedback data; S402: Globally aggregate the feedback data of edge devices to generate a global feedback matrix F global ; S403, according to the global feedback matrix F global Perform anomaly detection and filter abnormal data; S404, design a feedback-driven global strategy adjustment model based on the processed feedback data to adjust the global supply and demand difference value S global,i Conduct feedback optimization.

8. The marketing strategy analysis and optimization method based on big data and edge computing according to claim 7 is characterized in that: The weighted feedback model is expressed as follows: F weighted,i =a i ·(I adj,i -I sc,i )+b i ·(A push,i -A real,i )+c i ·ΔL dist,i Among them, F weighted,i Represents the weighted feedback data of the i-th node, which is used for feedback analysis and subsequent adjustment; adj,i Represents the adjusted inventory; I sc,i Indicates the inventory quantity before execution, used to calculate inventory differences; A push,i Indicates the expected amount of push. real,i Indicates the actual response behavior of the user; ΔL dist,i Indicates the difference in logistics execution, indicating the difference between expected and actual logistics resource arrival; α i ,β i ,γ i The weighting coefficients representing the feedback types, which represent the weights of inventory, marketing, and logistics respectively; The feedback data of the edge devices are globally aggregated to generate a global feedback matrix F global , which is expressed as follows: Among them, F global represents the global feedback matrix, which is used to represent the overall performance of each node’s feedback; w i represents the weight of the i-th node, based on the strategic importance and historical data of the node; F weighted,i Represents the weighted feedback data of the i-th node; ∈·R urg,i represents the emergency feedback correction term, where R urg,i represents the emergency demand score of the i-th node; The anomaly detection is represented as follows: Among them, E i Indicates the abnormal flag of the i-th node. When E i =1, it indicates that the node feedback is abnormal; represents the mean of the global feedback matrix; represents the standard deviation of the global feedback; λ represents the adaptive anomaly detection parameter, which is used to adjust the sensitivity of anomaly detection.

9. The marketing strategy analysis and optimization method based on big data and edge computing according to claim 1 is characterized in that: The S5 specifically includes: The dual feedback data of the adaptively adjusted global supply chain scheduling plan and the global precision marketing strategy are collected, and dynamic closed-loop optimization is performed based on the dual feedback data, as shown below: Among them, S opt,i represents the supply chain scheduling strategy after optimization of the i-th node; S adjust,i represents the supply chain scheduling strategy after the previous step adjustment; μ i It represents the feedback adjustment coefficient, which indicates the weight of the impact of feedback on the scheduling strategy; Represents global feedback data related to inventory, reflecting inventory deviation; It represents the global feedback data related to logistics, reflecting the deviation of logistics execution; It represents the global feedback data related to marketing, reflecting the execution deviation of precision marketing; The multi-dimensional feedback convergence model is expressed as follows: ΔS opt,i =λ1·(I adj,i -I sc,i )+λ2·(L dist,i -L real,i )+λ3·(A push,i -A real,i ) Among them, ΔS opt,i I represents the multi-dimensional feedback optimization adjustment of the i-th node; adj,i represents the adjusted inventory of the i-th node; I sc,i Indicates the actual inventory data; L dist,i Indicates the execution volume of the first logistics scheduling; L real,i Indicates the actual execution result of the logistics; A push,i Indicates the execution volume of the marketing push strategy; A real,i represents the actual response result of the user; λ1, λ2, λ3 are the weights of inventory, logistics and push feedback respectively; The feedback iterative convergence mechanism is expressed as follows: in, represents the supply chain scheduling adjustment after the kth round of feedback optimization; represents the supply chain scheduling adjustment after the k-1th round of feedback optimization; ∈ represents the convergence threshold, the standard for the control system to enter a stable state.

10. The marketing strategy analysis and optimization method based on big data and edge computing according to claim 9 is characterized in that: After the supply chain scheduling strategy is dynamically closed, a marketing feedback adjustment model is also designed to dynamically adjust the push frequency and content according to the user's actual response data; wherein the marketing feedback adjustment model is expressed as follows: M opt,i =M push,i +n i ·(A real,i -A push,i )+θ i ·DA u,i Among them, M opt,i represents the marketing push strategy after the optimization of the i-th node; M push,i Indicates the previous round of push plan; ν i A represents the push feedback coefficient, which controls the impact of user feedback on push adjustment; real,i Indicates the result of the user's actual response; A push,i Indicates the actual push volume in the last push plan; θ i ΔA u,i An adjustment item that indicates the rate of change of user behavior, indicating the adjustment of push strategy when user behavior changes significantly.

Images