A big data-based cigarette retail customer brand cultivation capability analysis system

Abstract

The application relates to a big data-based cigarette retail customer brand cultivation capability analysis system in the fields of business, finance or management, and the system comprises a data collection and feedback collection module, a blockchain data storage module and an intelligent marketing strategy generation module, wherein the data collected by the data collection and feedback collection module comprises tobacco monopoly system data, offline consumer behavior data and customer feedback data. Compared with the prior art, the application has the advantages that the data quality and compliance are ensured, the credibility is improved, the marketing accuracy is significantly improved, the customer feedback is automatically classified, sentiment analysis and trend prediction are realized, the labor cost is reduced and the decision-making scientificity is enhanced, the overall architecture considers data safety, real-time response and strategy flexibility, and efficient and reliable technical support is provided for digital transformation of the tobacco industry.

Description

Technical Field

[0001] This application relates to the fields of commerce, finance, or management, and in particular to a big data-based system for analyzing the brand cultivation capabilities of cigarette retail customers. Background Technology

[0002] With the rapid development of information technology, the application of big data and artificial intelligence technologies in the tobacco industry's marketing field is gradually deepening. Traditional tobacco marketing models have long relied on experience-based judgment and manual intervention, making it difficult to cope with challenges such as diversified consumer demands and intensified market competition. In recent years, solutions based on data platforms as their core architecture have emerged in the industry, achieving initial digital transformation of marketing strategies by integrating customer profiles, sales behavior, and market trend data. For example, customer relationship management based on CRM systems (such as Fenxiang Sales) improves the accuracy of personalized services through tagging and customer segmentation; while tobacco big data platforms (such as the Dongfang Jinxin solution) enhance the data-driven capabilities of the entire chain through supply chain optimization and production monitoring. Furthermore, the introduction of blockchain technology provides new possibilities for data storage and traceability, and some companies have already attempted to put sales data on the blockchain to ensure immutability. However, existing technologies still have significant shortcomings in areas such as the dynamic adaptability of data collection, privacy protection mechanisms, intelligent feedback processing, and the real-time nature of strategy generation.

[0003] The core problems currently facing tobacco marketing systems lie in the fragmentation of data governance and the limitations of algorithmic models. Traditional data collection strategies mostly employ static rules, which struggle to cope with dynamic changes in regional markets, leading to significant fluctuations in data quality and a lack of adaptive adjustment mechanisms for key indicators such as missing rates and response rates. Customer feedback processing still relies heavily on manual annotation, resulting in low classification efficiency and strong subjectivity, failing to meet the real-time analysis needs of large-scale text data and impacting the accuracy of customer satisfaction predictions. Therefore, a comprehensive solution capable of integrating multi-source data and achieving dynamic optimization is urgently needed to overcome existing technological bottlenecks. Summary of the Invention

[0004] This application provides a big data-based analysis system for cigarette retail customers' brand cultivation capabilities, which addresses the problems of existing technologies such as single data sources, static and incomplete customer profiles, lack of real-time feedback mechanisms, and weak data security measures.

[0005] This application provides a big data-based analysis system for cigarette retail customer brand cultivation capabilities, including a data collection and feedback module, a blockchain data storage module, and an intelligent marketing strategy generation module. The data collected by the data collection and feedback module includes tobacco monopoly system data, offline consumer behavior data, and customer feedback data.

[0006] As an improvement, the data collection and feedback collection module includes:

[0007] Data collection optimization involves dynamically adjusting the collection path and efficiency by simulating individual adaptive behavior.

[0008] Information analysis and organization, integrating multi-source data, and generating a unified tagging system;

[0009] User feedback analysis and tag classification automatically categorize customer feedback into tags such as "product improvement", "service optimization", and "policy consultation".

[0010] Data anonymization and compliance optimization: Federated learning avoids directly transmitting raw data by training models locally, thus meeting privacy protection requirements.

[0011] As an improvement, the data collection optimization is divided into information collection phase optimization and information filtering and evaluation optimization. The information collection phase optimization involves dynamically adjusting the data collection strategy, with the weight of the strategy adjustment being θ, where θ is a random number from 0 to 1, and the current collection strategy for the i-th data source being x. iter The current optimal acquisition strategy is x best The calculation method is as follows:

[0012]

[0013] Information filtering and evaluation optimization removes invalid data by dynamically adjusting filtering rules. The subjective influence factor is [I], and the information quality factor is Φ. The calculation method is as follows:

[0014] ,

[0015]

[0016] The formula for calculating the information quality factor Φ is:

[0017]

[0018] δ is a random number between 0 and 1, simulating data quality fluctuations;

[0019] Γ is a dynamic adjustment factor combining the sine and logarithmic functions;

[0020] Low-quality data is dynamically filtered using [I].

[0021] As an improvement, the information analysis and organization calculation method is as follows:

[0022]

[0023] Where Λ represents the analysis weight, and D represents the data dimension, which includes customer attributes, behavior, and feedback.

[0024] As an improvement, the user feedback analysis and tag classification are divided into user feedback classification processing, sentiment analysis processing, and real-time feedback optimization processing.

[0025] The feedback classification process uses the K-means clustering algorithm, and the calculation formula is as follows:

[0026]

[0027] Where the number of categories is k, and the centroid of the i-th category is μ i Customer feedback is categorized by topic to reduce manual labeling costs;

[0028] Sentiment analysis is performed using a logistic regression model, and the calculation formula is as follows:

[0029]

[0030] Where β is the model weight, x is the text feature vector used to predict customer satisfaction trends, and P(y=1|x) is the prediction probability;

[0031] Real-time feedback optimization processing employs a time series analysis differential autoregressive moving average model algorithm, with the calculation formula as follows:

[0032]

[0033] Where φ(L) and θ(L) are the autoregressive and moving average coefficients, and d is the difference order, which predicts the trend of customer satisfaction.

[0034] As an improvement, the data anonymization and compliance optimization denotes the learning law as η, and the i-th local model parameter as θ. i The parameter update formula is:

[0035]

[0036] The customer profile model is trained locally at the retail store, and only the aggregated model parameters are uploaded.

[0037] As an improvement, the steps of the data collection and feedback collection module are as follows:

[0038] Set basic data collection strategies based on historical data and monitor data quality in real time;

[0039] The strategy is dynamically adjusted according to the formula. If the missing data rate for a certain area is high, then θ for that area is increased. If the feedback response rate is low, then the questionnaire distribution area is expanded by x. best Updated to a broader coverage strategy;

[0040] In the initial stage, basic filtering rules are set, [I] and Φ are calculated in real time, and the filtering threshold is dynamically updated.

[0041] By integrating multi-source data from the tobacco monopoly system, offline consumer behavior data, and customer feedback data, and calculating a weighted average using a formula, a unified labeling system is generated.

[0042] As an improvement, the steps for user feedback analysis and tag classification are as follows:

[0043] The feedback text is vectorized using TF-IDF or word embedding, and the optimal k value is selected using the elbow rule. The feedback is then automatically categorized into preset labels.

[0044] Train a logistic regression model using labeled feedback data, input new feedback text, and output the probability of positive and negative sentiment.

[0045] The time series data is differentially processed and fitted with an ARIMA model to predict the satisfaction trend for the next three months.

[0046] As an improvement, the steps for data anonymization and compliance optimization are as follows:

[0047] Retailers train customer profile models locally using private data and upload local model parameters θ. i The central server aggregates all θ i Update global model parameters θ global ;

[0048] Identify sensitive fields, name fields, and numeric fields, and check whether the anonymized data meets the requirements.

[0049] As an improvement, the blockchain data storage module stores the data on the blockchain to ensure that the data is tamper-proof and traceable;

[0050] The intelligent marketing strategy generation module generates personalized marketing strategies based on customer tags and market trends.

[0051] Compared with existing technologies, the advantages of this invention are as follows: the data collection module adopts an adaptive optimization strategy, dynamically adjusts the collection path, and combines federated learning to achieve privacy protection, ensuring data quality and compliance; the blockchain notarization module guarantees data immutability and enhances credibility; the intelligent marketing module integrates customer tags and market trends to generate personalized strategies, significantly improving marketing accuracy. Furthermore, the system uses algorithms such as K-means clustering, logistic regression, and ARIMA models to achieve automated classification, sentiment analysis, and trend prediction of customer feedback, reducing manual costs and enhancing the scientific nature of decision-making. The overall architecture balances data security, real-time response, and strategy flexibility, providing efficient and reliable technical support for the digital transformation of the tobacco industry. Attached Figure Description

[0052] The accompanying drawings are provided to further understand the technical solutions of the present invention and constitute a part of the specification. They are used together with the embodiments of this application to explain the technical solutions of the present invention and do not constitute a limitation on the technical solutions of the present invention.

[0053] Figure 1 System block diagrams provided for embodiments of this application;

[0054] Figure 2 A flowchart provided for embodiments of this application;

[0055] Figure 3 Timing diagrams provided for embodiments of this application. Detailed Implementation

[0056] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0057] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indication will also change accordingly.

[0058] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0059] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "connected" and "linked" should be interpreted broadly, for example, as a fixed connection, a detachable connection, or an integral connection. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances. Furthermore, when describing pipelines, the terms "connected" and "linked" as used in this application have the meaning of establishing electrical connection. The specific meaning needs to be understood in conjunction with the context.

[0060] In the embodiments of this application, the terms "exemplary" or "for example" are used to indicate that something is an example, illustration, or description. Any embodiment or design that is described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design. Specifically, the use of the terms "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0061] like Figures 1-3 A big data-based analysis system for cigarette retail customer brand cultivation capabilities includes a data collection and feedback module, a blockchain data storage module, and an intelligent marketing strategy generation module. The data collected by the data collection and feedback module includes tobacco monopoly system data, offline consumer behavior data, and customer feedback data.

[0062] As an improvement, the data collection and feedback collection module includes:

[0063] Data collection optimization involves dynamically adjusting the collection path and efficiency by simulating individual adaptive behavior.

[0064] Information analysis and organization, integrating multi-source data, and generating a unified tagging system;

[0065] User feedback analysis and tag classification automatically categorize customer feedback into tags such as "product improvement", "service optimization", and "policy consultation".

[0066] Data anonymization and compliance optimization: Federated learning avoids directly transmitting raw data by training models locally, thus meeting privacy protection requirements.

[0067] As an improvement, the data collection optimization is divided into information collection phase optimization and information filtering and evaluation optimization. The information collection phase optimization involves dynamically adjusting the data collection strategy, with the weight of the strategy adjustment being θ, where θ is a random number from 0 to 1. The current collection strategy for the i-th data source is xitter, and the current optimal collection strategy is xbest, calculated as follows:

[0068]

[0069] Information filtering and evaluation optimization removes invalid data by dynamically adjusting filtering rules. The subjective influence factor is [I], and the information quality factor is Φ. The calculation method is as follows:

[0070] ,

[0071]

[0072] The formula for calculating the information quality factor Φ is:

[0073]

[0074] δ is a random number between 0 and 1, simulating data quality fluctuations;

[0075] Γ is a dynamic adjustment factor combining the sine and logarithmic functions;

[0076] Low-quality data is dynamically filtered using [I].

[0077] As an improvement, the information analysis and organization calculation method is as follows:

[0078]

[0079] Where Λ represents the analysis weight, and D represents the data dimension, which includes customer attributes, behavior, and feedback.

[0080] As an improvement, the user feedback analysis and tag classification are divided into user feedback classification processing, sentiment analysis processing, and real-time feedback optimization processing.

[0081] The feedback classification process uses the K-means clustering algorithm, and the calculation formula is as follows:

[0082]

[0083] Where the number of categories is k, and the centroid of the i-th category is μi, customer feedback is classified by topic to reduce manual annotation costs;

[0084] Sentiment analysis is performed using a logistic regression model, and the calculation formula is as follows:

[0085]

[0086] Where β is the model weight, x is the text feature vector used to predict customer satisfaction trends, and P(y=1|x) is the prediction probability;

[0087] Real-time feedback optimization processing employs a time series analysis differential autoregressive moving average model algorithm, with the calculation formula as follows:

[0088]

[0089] Where φ(L) and θ(L) are the autoregressive and moving average coefficients, and d is the difference order, which predicts the trend of customer satisfaction.

[0090] As an improvement, the data anonymization and compliance optimization denotes the learning law as η, the i-th local model parameter as θi, and the parameter update formula is:

[0091]

[0092] The customer profile model is trained locally at the retail store, and only the aggregated model parameters are uploaded.

[0093] As an improvement, the steps of the data collection and feedback collection module are as follows:

[0094] Set basic data collection strategies based on historical data and monitor data quality in real time;

[0095] The strategy is dynamically adjusted according to the formula. If the missing data rate of QR code scanning is high in a certain area, then θ for that area is increased. If the feedback response rate is low, then the questionnaire distribution scope is expanded and xbest is updated to a broader coverage strategy.

[0096] In the initial stage, basic filtering rules are set, [I] and Φ are calculated in real time, and the filtering threshold is dynamically updated.

[0097] By integrating multi-source data from the tobacco monopoly system, offline consumer behavior data, and customer feedback data, and calculating a weighted average using a formula, a unified labeling system is generated.

[0098] As an improvement, the steps for user feedback analysis and tag classification are as follows:

[0099] The feedback text is vectorized using TF-IDF or word embedding, and the optimal k value is selected using the elbow rule. The feedback is then automatically categorized into preset labels.

[0100] Train a logistic regression model using labeled feedback data, input new feedback text, and output the probability of positive and negative sentiment.

[0101] The time series data is differentially processed and fitted with an ARIMA model to predict the satisfaction trend for the next three months.

[0102] As an improvement, the steps for data anonymization and compliance optimization are as follows:

[0103] Retailers train customer profile models locally using private data, upload local model parameters θi, and the central server aggregates all θi and updates the global model parameters θglobal.

[0104] Identify sensitive fields, name fields, and numeric fields, and check whether the anonymized data meets the requirements.

[0105] As an improvement, the blockchain data storage module stores the data on the blockchain to ensure that the data is tamper-proof and traceable;

[0106] The intelligent marketing strategy generation module generates personalized marketing strategies based on customer tags and market trends.

[0107] Example:

[0108] Based on the big data analysis system for cigarette retail customer brand cultivation capabilities, the system's equipment workflow is as follows.

[0109] This embodiment describes in detail the overall workflow and principle of a big data-based cigarette retail customer brand cultivation capability analysis system, covering the collaborative operation of the data collection and feedback module, the blockchain data storage module, and the intelligent marketing strategy generation module.

[0110] I. System Architecture and Equipment Composition

[0111] The overall system architecture is divided into the following three core modules, each containing specific equipment components:

[0112] Data collection and feedback collection module (device number: DCM-100)

[0113] Sensor units: Deployed at offline retail terminals (such as barcode scanners and RFID readers) to collect consumer behavior data (such as scanning frequency and purchase volume).

[0114] Data acquisition unit: Integrated into the tobacco monopoly system server (model: Tobacco-Server V3.0), it receives data from the sensor units and stores it in a local database (model: MySQL 8.0).

[0115] Feedback collection terminals include a mobile app (model: Tobacco-App V2.5) and an offline questionnaire machine (model: Survey-Kiosk V1.2), used to collect customer feedback texts and satisfaction ratings.

[0116] Data preprocessing unit: Based on the federated learning framework (model: Federated-Learning-Engine V1.0), it performs data anonymization and compliance optimization.

[0117] Blockchain data storage module (device number: BCM-200)

[0118] Blockchain nodes: Using the Hyperledger Fabric V2.4 platform, deployed on a central server cluster (model: Hyperledger-Node V2.4), responsible for data uploading and evidence storage.

[0119] Hash Calculation Unit: Integrated into the blockchain node, it generates hash values ​​for preprocessed data and stores them on the blockchain.

[0120] Audit log unit: Records data storage timestamps and operator identities to ensure traceability.

[0121] Intelligent Marketing Strategy Generation Module (Device No.: SGM-300)

[0122] Analysis server: Based on K-means clustering algorithm (model: K-Means-Cluster V3.2) and logistic regression model (model: Logistic-Regression V2.8), it performs customer tag classification and sentiment analysis.

[0123] ARIMA Prediction Unit: Deployed on the analytics server, it uses a differential autoregressive moving average model (model: ARIMA-Model V1.5) to predict customer satisfaction trends.

[0124] Strategy generation engine: Generates personalized marketing strategies (such as coupon distribution rules and product recommendation lists) based on the tag system and market trends.

[0125] II. Equipment Working Process and Principle

[0126] 1. Data Collection and Feedback Collection Module (DCM-100)

[0127] Data acquisition phase (equipment: sensor unit, data acquisition device)

[0128] The sensor unit collects offline consumer behavior data in real time (such as scan records and purchase volume) and transmits it to the data collector via API interface.

[0129] The data collector adjusts the weights θ (θ∈[0,1]) and the optimal strategy xbest according to a dynamic strategy to optimize the data collection path. For example, if the data loss rate of a certain area is high, the value of θ is increased to expand the collection range.

[0130] Data preprocessing stage (equipment: data preprocessing unit)

[0131] The data preprocessing unit calculates the subjective influence factor [I] and the information quality factor Φ (Φ = δ·Γ+ log(Φ)) using the formula, where δ is a random number (δ∈[0,1]) and Γ is a sine-logarithmic function combination factor.

[0132] Low-quality data (such as invalid questionnaires and duplicate scan records) is dynamically filtered using [I] and Φ, and only data that meets the threshold is retained.

[0133] Customer feedback processing stage (equipment: feedback collection terminal, analysis server)

[0134] The feedback collection terminal vectorizes customer feedback text using the TF-IDF algorithm or word embedding model (such as Word2Vec) and inputs it into the analysis server.

[0135] The analysis server uses the K-means clustering algorithm (formula: The feedback was categorized into tags such as "product improvement" and "service optimization" (k=3).

[0136] Federated learning and data anonymization (equipment: data preprocessing unit, blockchain node)

[0137] The data preprocessing unit trains the customer profile model (parameter θ) locally at the retailer. i Only the aggregated parameter θ is uploaded. global To the blockchain node.

[0138] Blockchain node pair θ global Hash the data and store it on the blockchain to ensure that the data cannot be tampered with.

[0139] Customer tag generation (device: analytics server)

[0140] The analytics server integrates multi-source data (customer attributes, behaviors, feedback) and generates a unified label system using a weighted average formula (Λ⋅D), where Λ is the analysis weight (Λ=0.4:0.5:0.1) and D is the data dimension (attribute:behavior:feedback).

[0141] Sentiment analysis and prediction (equipment: analysis server, ARIMA prediction unit)

[0142] The analysis server uses a logistic regression model (formula: ), predicting the probability of customer satisfaction.

[0143] 2. Blockchain Data Storage Module (BCM-200)

[0144] Data on-chain storage and evidence preservation (equipment: blockchain node, hash calculation unit)

[0145] The hash calculation unit generates a hash digest of the preprocessed data (such as customer tags and satisfaction ratings) and stores it on-chain through the Hyperledger Fabric V2.4 platform.

[0146] The audit log unit records the operation timestamp and the operator's identity "Administrator A-xxxx-xxxx" to ensure data traceability.

[0147] 3. Intelligent Marketing Strategy Generation Module (SGM-300)

[0148] The ARIMA prediction unit performs differencing on time series data (d=1), fits an ARIMA model (φ(L)=0.8, θ(L)=0.5), and predicts the satisfaction trend for the next 3 months.

[0149] Policy generation and distribution (Device: Policy generation engine)

[0150] The strategy generation engine generates personalized marketing strategies based on the tagging system and market trends (such as "high-value customers" needing exclusive discounts), and distributes them to retail terminals via API (such as coupon push and product recommendation).

[0151] III. Equipment Collaboration Logic and Technical Implementation

[0152] Data Flow Path

[0153] Sensor unit → Data acquisition unit → Data preprocessing unit → Blockchain node → Analysis server → Strategy generation engine.

[0154] Privacy protection mechanism

[0155] The Federated-Learning-Engine V1.0 framework ensures that the raw data is processed only locally, and only the aggregated θ is uploaded. global .

[0156] Real-time guarantee

[0157] The ARIMA forecasting unit updates customer satisfaction trends hourly, while the strategy generation engine dynamically adjusts marketing plans.

[0158] Table 1: Correspondence between Equipment Model and Function

[0159]

[0160] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A big data-based analysis system for cigarette retail customer brand cultivation capabilities, comprising a data collection and feedback module, a blockchain data storage module, and an intelligent marketing strategy generation module, characterized in that: The data collection and feedback module collects data including tobacco monopoly system data, offline consumer behavior data, and customer feedback data. The data collection and feedback collection module includes: Data collection optimization involves dynamically adjusting the collection path and efficiency by simulating individual adaptive behavior. Information analysis and organization, integrating multi-source data, and generating a unified tagging system; User feedback analysis and tag classification automatically categorize customer feedback into tags such as "product improvement", "service optimization" and "policy consultation". Data anonymization and compliance optimization: Federated learning avoids direct transmission of raw data by training models locally, thus meeting privacy protection requirements; The data collection optimization is divided into information collection phase optimization and information filtering and evaluation optimization. The information collection phase optimization involves dynamically adjusting the data collection strategy, with the weight of the strategy adjustment being θ, where θ is a random number between 0 and 1, and the current collection strategy for the i-th data source is x. iter The current optimal acquisition strategy is x best The calculation method is as follows: Information filtering and evaluation optimization removes invalid data by dynamically adjusting filtering rules. The subjective influence factor is [I], and the information quality factor is Φ. The calculation method is as follows: ， The formula for calculating the information quality factor Φ is: δ is a random number between 0 and 1, simulating data quality fluctuations; Γ is a dynamic adjustment factor combining the sine and logarithmic functions; Low-quality data is dynamically filtered using [I]; The user feedback analysis and tag classification are divided into user feedback classification processing, sentiment analysis processing, and real-time feedback optimization processing. The feedback classification process uses the K-means clustering algorithm, and the calculation formula is as follows: Where the number of categories is k, and the centroid of the i-th category is μ i Customer feedback is categorized by topic to reduce manual labeling costs; Sentiment analysis is performed using a logistic regression model, and the calculation formula is as follows: Where β is the model weight, x is the text feature vector used to predict customer satisfaction trends, and P(y=1|x) is the prediction probability; Real-time feedback optimization processing employs a time series analysis differential autoregressive moving average model algorithm, with the calculation formula as follows: Where φ(L) and θ(L) are the autoregressive and moving average coefficients, and d is the difference order, which predicts the trend of customer satisfaction.

2. The big data-based cigarette retail customer brand cultivation capability analysis system according to claim 1, characterized in that: The information analysis and organization calculation method is as follows: Where Λ represents the analysis weight, and D represents the data dimension, which includes customer attributes, behavior, and feedback.

3. The big data-based cigarette retail customer brand cultivation capability analysis system according to claim 1, characterized in that: The data anonymization and compliance optimization process denotes the learning law as η, and the i-th local model parameter as θ. i The parameter update formula is: The customer profile model is trained locally at the retail store, and only the aggregated model parameters are uploaded.

4. The big data-based cigarette retail customer brand cultivation capability analysis system according to claim 1, characterized in that, The steps of the data collection and feedback collection module are as follows: Set basic data collection strategies based on historical data and monitor data quality in real time; The strategy is dynamically adjusted according to the formula. If the missing data rate of a certain area is high, then θ for that area is increased. If the response rate is low, expand the scope of the questionnaire distribution by x best Updated to a broader coverage strategy; In the initial stage, basic filtering rules are set, [I] and Φ are calculated in real time, and the filtering threshold is dynamically updated. By integrating multi-source data from the tobacco monopoly system, offline consumer behavior data, and customer feedback data, and calculating a weighted average using a formula, a unified labeling system is generated.

5. The big data-based cigarette retail customer brand cultivation capability analysis system according to claim 1, characterized in that, The steps for user feedback analysis and tag classification are as follows: The feedback text is vectorized using TF-IDF or word embedding, and the optimal k value is selected using the elbow rule. The feedback is then automatically categorized into preset labels. Train a logistic regression model using labeled feedback data, input new feedback text, and output the probability of positive and negative sentiment. The time series data is differentially processed and fitted with an ARIMA model to predict the satisfaction trend for the next three months.

6. The big data-based cigarette retail customer brand cultivation capability analysis system according to claim 1, characterized in that, The steps for data anonymization and compliance optimization are as follows: Retailers train customer profile models locally using private data and upload local model parameters θ. i The central server aggregates all θ i Update global model parameters θ global ; Identify sensitive fields, name fields, and numeric fields, and check whether the anonymized data meets the requirements.

7. The big data-based cigarette retail customer brand cultivation capability analysis system according to claim 1, characterized in that: The blockchain data storage module stores data on the blockchain to ensure that the data is tamper-proof and traceable. The intelligent marketing strategy generation module generates personalized marketing strategies based on customer tags and market trends.

Images