Multi-dimensional dynamic evaluation method and device based on multi-source heterogeneous data and electronic equipment

Abstract

This invention belongs to the field of big data analysis, market forecasting, and decision support technology, and relates to a multi-dimensional dynamic evaluation method, device, electronic device, and storage medium based on multi-source heterogeneous data. The method includes: fusing multi-source heterogeneous market data; constructing a multi-dimensional dynamic evaluation index for assessing market status based on the fused multi-source heterogeneous market data; constructing a long-term market forecasting model based on the multi-dimensional dynamic evaluation index; performing closed-loop fusion of intelligent algorithms and business scenarios based on the long-term market forecasting model; constructing an intelligent early warning and dynamic control system; and quantitatively evaluating the business value and technical effectiveness brought by the market perception center system itself, and continuously iterating and optimizing based on the evaluation data and feedback. By opening up data resource collection paths, it achieves large-scale data resource sorting, integration, and management; and overall promotes a fundamental transformation of market perception from experience-based and fragmented to scientific and systematic.

Description

Technical Field

[0001] This invention relates to the fields of big data analysis, market forecasting and decision support technology, and in particular to a multidimensional dynamic evaluation method, device and electronic device based on multi-source heterogeneous data. Background Technology

[0002] With the arrival of the "smart marketing" era, the tobacco industry faces higher demands for accurate market situation perception and scientific management of economic operations. The current cigarette market status evaluation system has the following main shortcomings: Single data dimension: Traditional methods rely too much on internal business data such as wholesale volume and retail price, and fail to effectively integrate external multi-source data such as macroeconomics, consumer behavior, and terminal operations, resulting in fragmented and one-sided market perception.

[0003] The models are rigid and outdated: They generally use static indicator systems and fixed weights, which cannot be automatically adjusted according to dynamic changes in the market. The evaluation results are not timely and cannot truly reflect the market status.

[0004] Disconnected from business scenarios: The evaluation results are mostly macro-level indicator analyses, not deeply linked to specific business decision-making scenarios such as "brand introduction and exit, supply control, and terminal management", which are not very practical and cannot form a closed loop of "perception-decision-execution".

[0005] Therefore, there is an urgent need for a multi-dimensional dynamic evaluation method that can integrate multi-source data, dynamically adjust models, and deeply integrate with business operations. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a multidimensional dynamic evaluation method based on multi-source heterogeneous data, employing the following technical solution, including the following steps: Integrate heterogeneous market data from multiple sources; Based on the fused multi-source heterogeneous market data, a multi-dimensional dynamic evaluation index for assessing market status is constructed. Based on the aforementioned multidimensional dynamic evaluation index, a long-term market forecasting model is constructed. Based on the aforementioned long-cycle market forecasting model, a closed-loop fusion of intelligent algorithms and business scenarios is performed. Construct an intelligent early warning and dynamic control system; The business value and technical effectiveness of the market perception center system itself are quantitatively evaluated, and continuous iterative optimization is carried out based on the evaluation data and feedback.

[0007] Preferably, the step of fusing multi-source heterogeneous market data specifically includes: Perform targeted crawling and structured mapping of macroeconomic data; Compliant access and anonymization processing of mobile operator population data; Align and integrate industry data with data from the middle platform.

[0008] Preferably, the step of constructing a multidimensional dynamic evaluation index for assessing market status based on the fused multi-source heterogeneous market data specifically includes: Scientific selection and construction of a multidimensional evaluation index system; Dynamic weight assignment is performed based on the tiered approach; The standardized indicator values ​​are combined with dynamic weights to calculate the specific value of the multidimensional dynamic evaluation index.

[0009] Preferably, the step of constructing a long-term market forecasting model based on the multidimensional dynamic evaluation index specifically includes: A sales potential prediction model that integrates multi-source big data; Construct a predictive model for structural upgrading driven by the macroeconomy; The sales potential prediction model and the structural improvement prediction model are subjected to self-learning and dynamic optimization.

[0010] Preferably, the step of performing closed-loop integration of intelligent algorithms and business scenarios based on the long-cycle market prediction model specifically includes: Provide a multi-dimensional evaluation method for the introduction of new brands or new product specifications; Establish an early warning system based on hard indicators to identify weak product specifications that should be considered for withdrawal and optimize the brand ecosystem; By applying macro-level indices and forecasts to specific business scenarios, personalized strategy recommendations can be automatically generated.

[0011] Preferably, the steps for constructing the intelligent early warning and dynamic control system specifically include: Dynamic threshold early warning is based on the 3σ rule; Develop a scenario-based multi-dimensional operation and control system; Automatic distribution and tracking of intelligent early warnings.

[0012] Preferably, the step of quantitatively evaluating the business value and technical effectiveness of the market perception center system itself, and conducting continuous iterative optimization based on the evaluation data and feedback, specifically includes: Establish and track a multi-dimensional performance evaluation indicator system; Feedback-based iterative optimization process for models and algorithms; Plan for the continuous evolution of technical architecture and data assets.

[0013] To address the aforementioned technical problems, this invention also provides a multidimensional dynamic evaluation device based on multi-source heterogeneous data, employing the following technical solution, including: The data fusion module is used to integrate heterogeneous market data from multiple sources. The index construction module is used to construct a multi-dimensional dynamic evaluation index for assessing market status based on the fused multi-source heterogeneous market data. The model building module is used to build a long-term market forecasting model based on the multidimensional dynamic evaluation index. The closed-loop fusion module is used to perform closed-loop fusion of intelligent algorithms and business scenarios based on the long-cycle market prediction model. The control module is used to construct an intelligent early warning and dynamic control system. The optimization module is used to quantitatively evaluate the business value and technical effectiveness of the market perception center system itself, and to continuously iterate and optimize it based on the evaluation data and feedback.

[0014] To address the aforementioned technical problems, the present invention also provides an electronic device that employs the technical solution described below, comprising a memory and a processor. The memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of the aforementioned multidimensional dynamic evaluation method based on multi-source heterogeneous data.

[0015] To address the aforementioned technical problems, the present invention also provides a computer-readable storage medium, which employs the technical solution described below. The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the aforementioned multidimensional dynamic evaluation method based on multi-source heterogeneous data.

[0016] Compared with existing technologies, this invention has the following main advantages: by opening up data resource collection paths, it achieves large-scale sorting, integration, and management of data resources; by constructing a market potential prediction model, it achieves thorough, accurate, and far-reaching potential exploration; by creating market status evaluation standards, it achieves three-dimensional, quantitative, and visual status evaluation; by establishing a market monitoring and early warning mechanism, it achieves proactive, sensitive, and automated business monitoring; by building an intelligent economic operation control system, it achieves clear, stable, and controllable operation regulation; and it successfully integrates big data and artificial intelligence technologies with the specific business scenarios of tobacco industry marketing, providing a feasible, assessable, and evolving methodology for the industry's digital transformation, and overall promoting a fundamental shift in market perception from experience-based and fragmented to scientific and systematic. Attached Figure Description

[0017] To more clearly illustrate the solutions in this invention, the accompanying drawings used in the description of the embodiments of this invention will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 This is a flowchart of an embodiment of the multidimensional dynamic evaluation method based on multi-source heterogeneous data of the present invention; Figure 2 This is a schematic diagram of the structure of one embodiment of the market perception and decision support device of the present invention; Figure 3 This is a schematic diagram of the structure of an embodiment of the electronic device of the present invention.

[0019] Figure reference numerals: 70-Multidimensional dynamic evaluation device based on multi-source heterogeneous data, 71-Data fusion module, 72-Index construction module, 73-Model construction module, 74-Closed-loop fusion module, 75-Control module, 76-Optimization module, 8-Electronic device, 81-Memory, 82-Processor, 83-Network interface. Detailed Implementation

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein in the specification is for the purpose of describing particular embodiments only and is not intended to limit the invention; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings are used to distinguish different objects and not to describe a particular order.

[0021] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0022] To enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0023] It should be noted that the multidimensional dynamic evaluation method based on multi-source heterogeneous data provided in the embodiments of the present invention is generally executed by a server / terminal device, and correspondingly, the market perception and decision support device is generally set in the server / terminal device.

[0024] It should be understood that the number of terminal devices, networks, and servers is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be used.

[0025] Example 1 Please refer to Figure 1 The flowchart illustrates an embodiment of the multidimensional dynamic evaluation method based on multi-source heterogeneous data according to the present invention. The multidimensional dynamic evaluation method based on multi-source heterogeneous data includes the following steps: Step S1: Integrate multi-source heterogeneous market data.

[0026] In this embodiment, the electronic device (e.g., a server / terminal device) running on the multi-dimensional dynamic evaluation method based on multi-source heterogeneous data can receive market perception and decision support requests via wired or wireless connections. It should be noted that the aforementioned wireless connection methods may include, but are not limited to, 3G / 4G / 5G connections, WiFi connections, Bluetooth connections, WiMAX connections, Zigbee connections, UWB (ultra-wideband) connections, and other currently known or future-developed wireless connection methods.

[0027] In this embodiment, step S1, fusing multi-source heterogeneous market data, may specifically include the following steps: S11 performs targeted crawling and structured mapping of macroeconomic data.

[0028] Develop a custom web crawler based on the Python Scrapy or Selenium framework. Configure crawling rules (such as URL patterns and data update frequency of monthly / quarterly) for specific data sources such as the National Bureau of Statistics and official websites of various provinces and cities. The crawler should automatically identify data tables or API interfaces on the target pages, and retrieve data on indicators such as GDP, per capita disposable income, and CPI on a scheduled and incremental basis, automatically recording the data source and timestamp.

[0029] Develop an integrated data parsing engine. For PDF reports, use OCR (Optical Character Recognition) technology combined with the PDFplumber or PyPDF2 library for text extraction, and utilize regular expressions to match key indicators and values. For Excel files, use the pandas library to directly read structured data. For API interfaces, use the requests library to call and parse the returned data in JSON / XML format. This engine uniformly transforms unstructured and semi-structured data into structured DataFrame objects.

[0030] Create a macroeconomic data standard mapping table in the database. This table defines each standard indicator field (e.g., gdp_growth_rate) and maintains its mapping relationship with different names in various data sources (e.g., "GDP year-on-year" and "Gross Domestic Product Growth Rate"). Simultaneously, attach metadata information to each data entry, including data_source (source institution), statistical_period (statistical period, e.g., 2023Q1), unit (unit of measurement, e.g., %), and update_time (update time). Through an ETL (Extract, Transform, Load) process, the parsed data is aligned to the mapping table, and values ​​are transformed before being stored in a unified data warehouse.

[0031] The purpose of step S11 is to automatically and standardize the access of scattered and inconsistent official macroeconomic data into the system, eliminate the errors and inefficiencies of manual collection, and provide standardized input for analyzing the correlation between macroeconomics and cigarette consumption.

[0032] S12 involves compliant access and anonymization of mobile operator population data.

[0033] Technical integration with mobile operators (such as China Mobile and China Unicom) is conducted, and a dedicated data access microservice is developed using Java (Spring Boot) or Python (FastAPI) based on their provided API documentation (such as the Swagger specification). This service implements identity authentication (such as OAuth2.0), parameter encapsulation (such as specifying regional codes and time ranges), request sending, and data receiving functions, and supports daily incremental retrieval of district and county-level data such as population size, age structure, and inflow / outflow trends.

[0034] Deploy anonymization components at the data access layer. Use deterministic encryption or hash algorithms (such as SHA-256) to irreversibly encrypt direct identifiers (such as phone numbers) in the original data. For quasi-identifiers (such as detailed age or precise geographic location), use generalization techniques to convert age into ranges (such as 20-30 years old) and blur geographic locations to the district / county level. Anonymization rules strictly adhere to the Data Security Law and the Personal Information Protection Law, ensuring that the output data is an aggregated statistical result and cannot be traced back to individuals.

[0035] The transmission channel is encrypted using the HTTPS protocol. Before being stored, the incoming data undergoes format verification and integrity checks again. Anonymized demographic data is stored in a partitioned table on a big data platform (such as Hive), and linked to business data using regional codes and timestamps, laying the foundation for subsequent integrated analysis.

[0036] The purpose of step S12 is to safely and compliantly introduce real-time, fine-grained population dynamics data to compensate for the shortcomings of traditional business data in consumer perception and provide key input for market capacity and potential analysis.

[0037] S13, align and integrate industry and middle platform data.

[0038] Establish a core indicator definition library. Clearly define the unified unit of measurement for metrics such as "sales volume" as "box," and write conversion rules: Sales volume (boxes) = Sales volume (units) / 250 = Sales volume (ten thousand units) / 5. Develop a conversion engine that automatically calls the corresponding rules to perform the conversion when it detects that the unit of input data is not a standard unit, and records the conversion process in the data quality log.

[0039] Construct a product specification label mapping dictionary knowledge base. Utilize Natural Language Processing (NLP) technology to calculate the similarity (e.g., cosine similarity) between product specification names in industry downstream data (e.g., "Soft Yuxi") and names in provincial bureau central platform data (e.g., "Yuxi (Soft)"). Combined with manual verification, establish "identical names with different names" mapping pairs. Develop a label management module that automatically queries the dictionary when data flows in, replacing non-standard names with standard names to resolve data association breakage issues.

[0040] Design a data cleaning pipeline with multi-level validation rules. Level 1: Format validation: Use regular expressions (e.g., dddd-dd-dd) to enforce a uniform date format; check numeric fields for non-numeric characters. Level 2: Logical validation: Check if the inventory-to-sales ratio is greater than 0 and if the price is within a reasonable range. Level 3: Outlier handling: For missing values, imputation is performed based on time series data or the average of similar product specifications; for obvious outliers (e.g., values ​​outside the 3σ rule), they are marked and alerts are triggered. The cleaned data is then loaded into the integrated data layer of the "Data into One Database" system.

[0041] The purpose of step S13 is to solve the problem of "data mismatch" between industry-wide data and provincial data platforms due to inconsistencies in statistical standards, naming conventions, and data formats, and to achieve seamless connection and integration of cross-level data.

[0042] Step S1 aims to address the shortcomings of fragmented, inconsistent, and missing data in traditional market perception systems. By integrating heterogeneous data from multiple sources, such as macroeconomic data, operator population data, and industry downturn data, a standardized, high-quality, and correlated data foundation is constructed to provide reliable data for subsequent analysis.

[0043] Step S2: Based on the fused multi-source heterogeneous market data, construct a multi-dimensional dynamic evaluation index to assess the market status.

[0044] In this embodiment, step S2, constructing a multidimensional dynamic evaluation index for assessing market status based on the fused multi-source heterogeneous market data, may specifically include the following steps: S21, to conduct scientific selection and construction of a multi-dimensional evaluation index system.

[0045] A database of academic literature was built using web crawlers to automatically extract key indicators from domestic and international papers on market conditions, consumer behavior, and retail management. Simultaneously, an expert survey questionnaire (Delphi method) was designed to collect scores from industry experts on the importance of potential indicators through an online platform. The indicators mined from the literature and those nominated by experts were merged and deduplicated to form an initial indicator pool (e.g., containing 50+ candidate indicators).

[0046] For the indicators in the initial indicator pool, correlation analysis and collinearity diagnosis were performed using historical data (variance inflation factor VIF was calculated). Principal component analysis (PCA) or factor analysis was used to identify the core original indicators corresponding to the principal components with the highest contribution rates. The importance scores of the indicators were calculated using machine learning algorithms such as random forest. Based on the combined statistical test results and business interpretability, redundant (highly correlated) and noisy (low-contribution) indicators were removed, and finally, 26 core indicators as shown in the table were refined and reasonably categorized into five dimensions.

[0047] Clearly define whether each indicator is a "positive indicator" (higher values ​​are better, such as sales rate) or a "negative indicator" (lower values ​​are better, such as inventory-to-sales ratio). Create a precise calculation formula document for each indicator. For example, the "absolute value of the wholesale-retail structure difference" can be defined as ABS (average wholesale price - average retail price) to measure the smoothness of price transmission. This provides unambiguous rules for subsequent data preprocessing and index calculation.

[0048] The purpose of step S21 is to establish a comprehensive, objective, and quantifiable indicator system to ensure that the five dimensions (external environment, market prosperity, brand health, terminal confidence, and consumer activity) can reflect the key aspects of the cigarette market without repetition or omission.

[0049] S22, based on the grade separation method, performs dynamic weight assignment.

[0050] Implement three processing functions in the code: For indicators with negative values ​​(such as premium rates): ,in, Original index values The minimum value of this indicator in the sample. : The processed index value (eliminating negative values).

[0051] For negative indicators (such as inventory-to-sales ratio): ,in, The maximum value of this indicator in the sample. After processing, the smaller the original value (the better), the better. The larger.

[0052] Standardization process: ,in, The maximum value of the indicator after the first two steps. Standardized indicator values, ranging from [0, 1]. Ensure that all indicator values ​​are within the range of [0, 1] and in the same direction (the larger the value, the better the condition).

[0053] The method of differentiating levels is adopted, and the weighting coefficients of the indicators are selected to maximize the differences between the evaluated objects. Assume... This is the comprehensive evaluation function for the object being evaluated. Wherein, ,See An undetermined positive vector, Let i be the state vector of the object being evaluated. Let the i-th object be... Observations Attached From this, we can obtain: .

[0054] Right now:

[0055] but: .

[0056] Determine the weight coefficient vector The criterion is to maximize the representation of the differences between evaluated objects of varying "quality," that is... The variance should be as large as possible. The variance is: .

[0057] Will Substituting into the above formula, we can obtain .

[0058] beg The maximum value of the variance is the key to choosing a suitable... , so that: .

[0059] Assume that after preprocessing there is One evaluation object (e.g.) (Cities and prefectures) One indicator.

[0060] Constructing an evaluation matrix Its elements For the first The object in the first Standardized values ​​for each indicator.

[0061] Calculate matrix ( (Real symmetric matrix).

[0062] Solve the matrix Maximum eigenvalue and its corresponding eigenvectors .

[0063] eigenvectors Normalization (making the sum of its components equal to 1) yields the optimal weight vector. . That is, the first The weight of each indicator.

[0064] This method requires no prior knowledge; the weights are determined entirely by the data itself, making the comprehensive evaluation function... The variance is the largest, which fully widens the gap between different evaluation objects, thus enabling a better distinction between good and bad.

[0065] Establish a weight update scheduling task. Set a cycle of quarterly or semi-annual periods. The system will automatically use the new data from the previous cycle to re-execute the "gradient method" to calculate the weights of indicators within each dimension. The updated weights will be applied to the index calculation for the next cycle, enabling the evaluation system to dynamically adapt to changes in market structure and development stage, and avoiding evaluation distortion caused by fixed weights.

[0066] The purpose of step S22 is to abandon subjective and static weight allocation methods and use objective mathematical methods to automatically determine the weight of each indicator in the index, so that the final calculated index can distinguish the differences in the market conditions of different regions and periods to the greatest extent.

[0067] S23, combine the standardized index values ​​with dynamic weights to calculate the specific value of the multidimensional dynamic evaluation index.

[0068] The external environment index is an index that reflects the macroeconomic prosperity level of different regions. The higher the index value, the higher the prosperity level of the macroeconomic and social environment in the region.

[0069] External Environment Index (EEI): ,in, Population (after standardization) Population year-on-year, GDP year-on-year, Year-on-year comparison of social electricity consumption Year-on-year growth of total retail sales of consumer goods Year-on-year growth in per capita disposable income Year-on-year growth of industrial added value above designated size : The dynamic weight of the corresponding indicator.

[0070] Market Sentiment Index (MCI): ,in, Regional inventory-to-sales ratio (after negative processing) Sales rate Retail price index The absolute value of the difference between wholesale and retail structure. : The dynamic weight of the corresponding indicator.

[0071] It reflects the health and activity level of regional market sales; the higher the index value, the better the market sales status.

[0072] Brand Health Index (BHI): First, calculate the... Health index of individual product specifications: ,in, : No. Wholesale sales growth rate of individual product specifications : No. Changes in market share for individual product specifications at the same price point : No. Inventory-to-sales ratio for each product specification (after negative processing). : No. Premium rate for each product specification : No. The order fulfillment rate of each product specification : The dynamic weight of the corresponding indicator.

[0073] Then calculate the regional brand health index: ,in, The total number of product specifications available for sale in this region. The first in this region The cumulative sales share of each product specification.

[0074] It reflects the layout and health status of the regional market brand ecosystem; the higher the index value, the healthier the brand as a whole.

[0075] Terminal Confidence Index (TCI): ,in, Order fulfillment rate Demand fulfillment rate Gross profit margin : The dynamic weight of the corresponding indicator.

[0076] It reflects the business confidence and level of regional cigarette retailers. The higher the index value, the more confident and capable the retailers are.

[0077] Consumer Activity Index (CAI): ,in, Average spending per person Average number of times per person to consume : Number of items consumed per person Repurchase rate : The dynamic weight of the corresponding indicator.

[0078] It reflects the vibrancy of the end-consumer market; the higher the index value, the more active the consumers and the more solid the market demand base.

[0079] Using OLAP technology (such as Apache Kylin or Druid), the calculated five-dimensional index, along with its underlying indicator data, is pre-calculated and stored according to multiple dimensions such as "time (year-month-day)," "region (province-city-county)," and "product specification," forming a data cube. This enables front-end applications to perform rapid drill-down, roll-up, and slice analysis on any dimension (such as "viewing the external environment index trend of Zhejiang Province in each month of 2023").

[0080] Develop a "Market Status Overview" dashboard using front-end visualization libraries (such as ECharts and D3.js). A five-dimensional radar chart simultaneously displays the values ​​of five indices for a region, intuitively showing the balance of development; an index trend comparison line chart shows index changes in the same region at different times, or in different regions at the same time; a geographic heat map maps index values ​​onto a map, intuitively displaying the spatial distribution of market status in each region. Users can click on charts to query detailed data.

[0081] The purpose of step S23 is to combine the standardized indicator values ​​with dynamic weights to calculate five specific index values, and then present them intuitively through visualization technology, making the complex market situation clear at a glance.

[0082] Step S2 aims to break through the limitations of traditional methods that rely on single, static indicators to evaluate market conditions, and to construct a dynamic index model that covers five dimensions: macro, meso, and micro, so as to achieve a systematic, quantitative, and visual assessment of market health.

[0083] Step S3: Based on the multidimensional dynamic evaluation index, construct a long-term market forecasting model.

[0084] In this embodiment, step S3, constructing a long-term market forecasting model based on the multidimensional dynamic evaluation index, may specifically include the following steps: S31 is a sales potential prediction model that integrates multi-source big data.

[0085] Establish four core sub-models and connect them for computation: Based on operator population data, time series models (such as Prophet or LSTM) are used to predict changes in the permanent and floating population of each region over the next 1-3 years.

[0086] Based on the "thick data" of historical consumption surveys, a logistic regression model is constructed to predict future smoking rates by using population structure (age, gender) and socioeconomic factors (education level) as features.

[0087] Based on historical data, the changing trend of per capita smoking volume (cartridges / year) is analyzed, and a smoothing forecasting method (such as Holt-Winters) is used for prediction.

[0088] Market Capacity Sub-warehouse: Combining the outputs of the above three sub-warehouses, the predicted sales potential is calculated: Predicted Market Capacity = Predicted Population * Predicted Smoking Rate * Predicted Per Capita Smoking Amount. This model upgrades the traditional model of predicting the future based on historical sales to a fundamental prediction driven by population and consumer behavior.

[0089] Process anonymous mobile signaling data provided by operators. Dwell point identification: Identify users' workplaces, residences, and frequently visited locations through spatiotemporal density clustering of signaling (such as the DBSCAN algorithm). Migration path extraction: Construct user migration trajectories between different regions based on the temporal changes of dwell points. Group migration pattern mining: Aggregate and analyze the trajectories of a large number of users, and use complex network theory to identify regular population flow patterns during specific periods such as holidays and weekends (e.g., flow from first-tier cities to third- and fourth-tier cities), quantify the scale of the flow, and use this as a key input to correct short-term forecasts in the "population trend sub-warehouse".

[0090] For the provincial bureau's annual sampling survey data of 220,000 samples, statistical expansion methods such as "post-hoc stratified weighting" were adopted. Based on auxiliary information such as population census data, the samples were stratified according to dimensions such as region, age, and gender, and a weight was assigned to each stratum, so that the weighted sample structure could accurately reflect the overall structure. The expanded consumer data (such as category preferences and price sensitivity) was cross-validated and calibrated with the macro-prediction results of the "market capacity sub-warehouse" to improve the micro-accuracy of the prediction.

[0091] The purpose of step S31 is to accurately predict the long-term growth potential of cigarette sales in the future market, providing a quantitative basis for capacity planning and supply allocation.

[0092] S32, Construct a structural improvement forecasting model driven by the macroeconomy.

[0093] Historical data on over 20 macroeconomic indicators, including per capita GDP, per capita disposable income, and total retail sales of consumer goods, were collected from various regions over the past decade, along with data on the sales proportion of high-end cigarettes (e.g., those with a retail price of over 300 yuan per carton) during the same period. Pearson correlation coefficient and Granger causality test were used to quantitatively analyze which macroeconomic indicators showed significant and leading correlations with the improvement in cigarette structure. For example, it was found that "year-on-year growth in per capita disposable income" typically leads "the proportion of high-end cigarettes" by 1-2 years.

[0094] Using the selected key macroeconomic indicators as independent variables Using the proportion of high-end cigarettes as the dependent variable Construct a multiple linear regression model: ,in, Sales volume of high-end cigarettes (dependent variable). : p key macroeconomic indicators (independent variables) selected from the data. : Regression intercept term, : Regression coefficients corresponding to each independent variable Random error term This model quantifies the historical impact of various macroeconomic indicators on the improvement of cigarette structure. (Value). Regression coefficients are fitted using historical data. Then, by combining the future forecasts of macroeconomic indicators with the equation, the predicted future proportion of high-end cigarettes can be calculated. The model can also output standardized regression coefficients, explaining the relative contribution of each driving factor.

[0095] Establish a hierarchical forecasting system. First, at the regional level, predict the overall structural improvement potential (e.g., the proportion of high-end cigarettes will increase from 30% to 35%). Then, at the price segment level, analyze the historical market share transfer matrix to predict which price segment will be the primary driver of growth (e.g., the 300-500 yuan range). Finally, at the product specification level, combine the "brand health index" and the "10A evaluation method" to predict which specific product specifications are most likely to meet the growth demand within that price segment. This penetrating forecasting translates macro trends into actionable product specification strategies.

[0096] The purpose of step S32 is to predict the potential and path for future improvement in the cigarette consumption structure (the proportion of high-end cigarettes), and to guide brand cultivation and the layout of high-end sources.

[0097] S33, perform self-learning and dynamic optimization on the sales potential prediction model and the structural improvement prediction model.

[0098] Establish a forecast monitoring dashboard. Whenever new actual sales and structural data are generated, the system automatically compares them with the historical forecast values ​​for the same period, calculating indicators such as Mean Absolute Percentage Error (MAPE) and Root Mean Square Error (RMSE). When the error exceeds a preset threshold for several consecutive periods, a model optimization warning is automatically triggered.

[0099] For time series models (such as LSTM) used in sales forecasting, an online learning model is employed. New real-world data is used as training samples, and the model parameters are fine-tuned with a small learning rate, allowing the model to quickly absorb the latest market information. For regression models used in structural forecasting, the model is retrained periodically (e.g., quarterly) using a rolling time window (e.g., data from the most recent 5 years), with new coefficients. Replace the old coefficient.

[0100] For the same prediction target (e.g., sales volume), multiple different types of models (e.g., ARIMA, Prophet, gradient boosting trees) can be run in parallel. A model ensemble module can be developed to dynamically allocate weights based on the performance of each model on the recent validation set, perform a weighted average, and generate the final ensemble prediction result. This mechanism can effectively reduce the risk of single model failure and improve the robustness of predictions.

[0101] The purpose of step S33 is to enable the prediction model to automatically iterate and optimize as new data accumulates and the market environment changes, maintain prediction accuracy, and overcome the performance degradation problem caused by the "once and for all" nature of traditional models.

[0102] Step S4: Based on the long-cycle market prediction model, perform closed-loop integration of intelligent algorithms and business scenarios.

[0103] In this embodiment, step S4, which involves the closed-loop fusion of intelligent algorithms and business scenarios based on the long-term market prediction model, may specifically include the following steps: S41 provides a multi-dimensional evaluation method for the introduction of new brands or new product specifications.

[0104] The quantification of product specifications and attributes, along with their status, includes: Attribute Dimension (5A): Establish a rule engine to automatically score the five attributes of candidate product specifications from 0 to 5A. For example: "Key Category": Is it an innovative category that the industry strongly encourages for development (such as slim, medium, and short cigarettes)? If yes, assign 1A. "Low Tar Content": Assign 1A if tar content ≤ 6mg. "Low Price": Is the price point positioned to meet the basic needs of the mass market? If yes, assign 1A. "Co-developing Brand": Is it a product specification that is strategically coordinated with key industrial companies? If yes, assign 1A. "Innovative Category": As mentioned above, if it conforms to the innovation direction, assign 1A. Status Dimension (5A): Connect to the evaluation data from the second step.

[0105] Calculate the "market share" and "sales growth rate" of this product specification in the regions where it has been sold. Define four quadrants: high market share and high growth (growth stage, 2A), high market share and low growth (maturity stage, 3A), low market share and high growth (launch stage, 1A), and low market share and low growth (decline stage, 0A).

[0106] Calculate the single-box structure value for this product specification. If it is higher than the overall single-box structure value of the introduced region, assign 1A; if it is higher than the per-acre structure value of its price category, assign another 1A.

[0107] Develop a "10A Evaluation" service. Input a list of candidate product specifications, and the system automatically retrieves relevant attribute data and market status data from the database, applies the aforementioned rules to score each specification, and calculates the total "A" score (0-10A) for each specification. A visual report is generated, displayed in descending order of "A" score. The report clearly lists the score and basis for each "A" item, providing transparent decision support for decision-makers.

[0108] Establish a tracking file for each introduced product specification. Record its actual market performance after introduction (e.g., market share, growth rate, and health index after 6 months). Perform correlation analysis between the performance results and the initial "A number" prediction to calculate the prediction accuracy. For evaluation rules that consistently show deviations (e.g., the "low price" attribute does not bring the expected high growth), initiate a rule optimization process. Business experts and data analysts should jointly review and adjust the scoring logic to ensure the evaluation method continuously evolves in practice.

[0109] The purpose of step S41 is to provide a set of objective, quantitative, and multi-dimensional evaluation standards for the introduction of new brands (product specifications), replacing subjective experience-based judgments and reducing the risk of introduction failure.

[0110] S42 establishes an early warning system based on hard indicators to identify weak product specifications that should be considered for withdrawal, thereby optimizing the brand ecosystem.

[0111] Thresholding definitions and calculations for four types of hard metrics: Low (low market share): Calculates the sales share of this product specification among all product specifications of the "belonging industrial company", sets a threshold (e.g., <1%), and triggers when it is below the threshold.

[0112] Lag (growth gap): Calculate the year-on-year growth rate of retail sales for this product category, rank it among all product categories currently on sale, set a threshold (e.g., bottom 20%), and trigger it.

[0113] High (high inventory-to-sales ratio): Obtain the "inventory-to-sales ratio" of this product specification, rank it among all product specifications, set a threshold (such as the top 20%, i.e., the batch with the highest inventory-to-sales ratio), and trigger it.

[0114] Loss (Price Inversion): Calculates the "market price inversion rate" (the frequency or extent to which the retail price is lower than the wholesale price), and triggers it if it ranks high (e.g., in the top 20%).

[0115] Four-color card-based visual alert interface: In the brand management dashboard, create a "card" for each product category currently on sale. The system calculates daily performance metrics and colors the cards based on trigger events. Red card: Simultaneous triggering of "high" and "loss" indicates huge inventory and price pressures, and exiting is the highest priority.

[0116] Orange card: Triggers three of the indicators.

[0117] Yellow card: Two of the indicators have been triggered.

[0118] Blue cards: Normal or only one item triggered. The interface supports color filtering and sorting, allowing managers to focus on the specifications with the most serious problems at a glance.

[0119] Clicking on the alert card displays not only the specific triggering indicators and data, but also auxiliary decision-making information. For example: a) Correlation analysis: It suggests that the withdrawal of this product specification might free up market space for a healthier product specification (with a higher brand health index) at the same price point. b) Historical cases: It shows the market impact of similar "low stagnation, high loss" product specification withdrawals in the past. c) Simulation: It provides a simple "hypothesis-outcome" simulation tool, allowing managers to estimate the impact of withdrawing this product specification on overall sales and product structure. This transforms the withdrawal decision from a passive "cleanup" to a proactive "optimization."

[0120] The purpose of step S42 is to establish an early warning system based on hard indicators, proactively and objectively identify weak product specifications that should be considered for withdrawal, optimize the brand ecosystem, and release market resources.

[0121] S43 applies macro-level indices and forecasts to specific business scenarios, automatically generating personalized strategy recommendations.

[0122] In collaboration with business experts, we analyzed the five key indices and linked them to specific marketing actions, creating an "if-then" rule base. For example: If the “Consumer Activity Index” declines and the “Terminal Confidence Index” remains stable, then the recommended strategy is: “Initiate consumer promotional activities (such as prizes for scanning QR codes) and check whether terminal inventory is sufficient.”

[0123] If the "Market Sentiment Index" is high and the "External Environment Index" is predicted to decline, then the recommended strategy is to "appropriately tighten the pace of supply release to prevent a rapid cooling of the market after it overheats."

[0124] If the “Brand Health Index” indicates that a brand in a certain price segment is aging AND the “Structure Improvement Model” predicts that the price segment still has potential, then the recommended strategy is: “Introduce new products that meet the “10A evaluation method” in this price segment, or increase resources for cultivating existing healthy product specifications.”

[0125] Develop a strategy report generation engine. Regularly (e.g., weekly), generate a personalized weekly market report for each regional manager. The first part of the report displays the current status and forecast of the region's five-dimensional indices; the second part, the engine calls upon the aforementioned rule base, matches the region's index status, and automatically generates 3-5 core strategy recommendations; the third part lists the key products requiring attention (four-color cards). The report supports one-click export to PDF or sending to the workbench.

[0126] On the mobile work platform for marketers, a simple strategy execution feedback function is designed. When the system recommends a strategy (such as "conduct a terminal display competition for product specification A"), the implementer can mark it as "executed" and optionally fill in simple effect feedback (such as "after execution, the sales rate of this product specification increased by X%)." This feedback data is collected to evaluate the effectiveness of the strategy rules, thereby continuously optimizing the "index-scenario" rule knowledge base.

[0127] The purpose of step S43 is to apply macro-level indices and forecasts to specific business scenarios, automatically generate personalized strategy recommendations, and improve the execution efficiency of frontline marketing personnel.

[0128] Step S4 aims to break the disconnect between algorithm models and business practices. By creatively designing scenario-based algorithm rules such as the "10A evaluation method" and "less stagnation and higher losses," the evaluation index from the second step and the prediction results from the third step are directly transformed into decision-making suggestions for specific business actions such as brand introduction, exit, and cultivation, forming a closed loop of "perception-decision-execution."

[0129] Step S5: Construct an intelligent early warning and dynamic control system.

[0130] In this embodiment, step S5, constructing an intelligent early warning and dynamic control system, may specifically include the following steps: S51, based on the 3σ rule, performs dynamic threshold early warning.

[0131] For the 64 indicators requiring early warning (such as the increase in sales of high-priced cigarettes), the statistical characteristics of their historical data over the past 2-3 years were calculated: mean (μ) and standard deviation (σ). This was used to establish a "normal fluctuation baseline" for the indicator.

[0132] Apply the "3σ rule" to set the warning threshold. Typically, the warning threshold is set as follows: .in, : The average of historical data for the indicator. Standard deviation of historical data for the indicator. Adjustable parameters, set according to business risk tolerance (e.g.) This is a general warning. (This is a serious warning).

[0133] This threshold is set based on the historical statistical patterns of the indicator itself. When the indicator value fluctuates beyond... When the range is exceeded, it is considered that a statistically rare event has occurred, which may indicate an anomaly. The system automatically recalculates monthly or quarterly using the latest rolling historical data. and This updates the threshold to adapt to the trend changes in the indicator itself (such as overall growth leading to an increase in the mean).

[0134] In the early warning rule configuration backend, configure each indicator individually. The value can be further personalized by region. For example, for Province A, which experiences significant economic fluctuations, the early warning value for "year-on-year growth of total retail sales of consumer goods" can be adjusted. The value can be set slightly higher than that of the economically stable province B to accommodate its inherent high volatility. This refined configuration ensures the appropriateness of the early warning system.

[0135] The purpose of step S51 is to change the traditional fixed threshold early warning method, and to set a dynamic and reasonable early warning line based on the historical fluctuation pattern of the indicator itself, so as to reduce false alarms and missed alarms and make the early warning more accurate and targeted.

[0136] S52 enables the scenario-based construction of a multi-dimensional operation and control system.

[0137] Develop a benchmarking analysis engine that supports flexible configuration of benchmarking dimensions (national, provincial, and peer units). It automatically calculates the gaps in key indicators such as sales volume, structure, and market share within the region, and visualizes the magnitude and direction of these gaps (leading / lagging) in a dashboard format.

[0138] We construct a collaborative data view between industry and commerce to analyze the performance of each industrial company in the local market from three dimensions: scale, structure, and efficiency. For example, through a "structural contribution matrix" chart, we can intuitively display the distribution and growth contribution of each industrial company's products at different price points in the local market, providing data support for sourcing negotiations.

[0139] Establish a holiday analysis template. Around holidays such as Spring Festival and Mid-Autumn Festival, automatically focus on relevant data (population inflow and outflow, sales of high-priced gift-box cigarettes, changes in market share of key brands), conduct year-on-year and month-on-month analyses, and generate special reports on holiday markets to summarize holiday consumption patterns.

[0140] The "barometer-spotlight" tool for intelligent management of terminal operations: Terminal Barometer (Confidence Index Tiered Monitoring): The calculated "Terminal Confidence Index" is broken down and sorted according to retailer tiers (1-30 tiers). A "Bi-period Comparison Bar Chart" displays the change in confidence values ​​for each tier between the current and previous periods; a "Trend Line Chart" displays the long-term trend of confidence for key tiers. Clicking on any tier will bring up a details page, displaying the detailed indicators (order fulfillment rate, gross profit margin, etc.) that constitute the confidence index for that tier.

[0141] Operational Spotlight (Root Cause Analysis for Problem Tiers): When the "Barometer" shows low confidence for a certain tier (e.g., consistently yellow or red), the "Spotlight" analysis can be activated with a single click. This function automatically pulls aggregated data from customers in that tier across four dimensions: "Purchase, Sales, Inventory, and Price," compares it with data from tiers with high confidence, quickly pinpoints the root cause of the problem (e.g., is it a significantly high "Inventory Turnover Days" or a low "Average Gross Profit Margin"), and provides targeted improvement suggestions (e.g., "It is recommended to provide inventory optimization guidance for customers in this tier").

[0142] The purpose of step S52 is to integrate market control from fragmented and isolated analysis into a systematic and scenario-based intelligent control platform covering the three core dimensions of market, brand, and terminal, thereby achieving comprehensive empowerment.

[0143] S53 enables the automatic distribution and tracking of intelligent early warnings.

[0144] Establish an early warning information routing rule engine. Based on the importance of the early warning indicators (e.g., involving high-priced cigarettes vs. regular cigarettes), severity (e.g., exceeding 3σ vs. exceeding 2σ), and the area of ​​occurrence, early warning information is categorized into "major," "general," and "advice" levels. Through a configured responsibility matrix, early warning information of different levels and types is automatically pushed to the corresponding provincial, municipal, and county-level managers or specific personnel (e.g., brand managers, marketing managers) via WeChat, DingTalk, SMS, or the system's message center.

[0145] The warning message includes a "Quick Response" link. Clicking it redirects to a response page pre-filled with warning details and providing standardized response options (such as "Adjust Supply," "Conduct Market Inspections," and "Activate Promotional Plans") as well as a free-filling box. The respondent selects or fills in the response measures and submits them. The system automatically records the respondent, time, and plan, and can assign tasks to collaborators, enabling online collaborative response.

[0146] The system automatically tracks changes in market indicators after an alert is triggered. For example, after an alert is issued regarding the "inventory-to-sales ratio" in a certain region, the response measure is to "tighten supply." The system will monitor whether the inventory-to-sales ratio in that region falls back to the normal range in subsequent periods. By analyzing a large amount of data on alerts, responses, and outcomes, the effectiveness of various response measures can be evaluated, and in turn, the alert threshold (k-value) and the response suggestion library can be optimized, forming a reinforced "alert-response-learning" cycle.

[0147] The purpose of step S53 is to ensure that the early warning information can be delivered to the relevant responsible persons in a timely and accurate manner, and to form a closed management loop from early warning to handling and feedback, so as to avoid early warning becoming a mere formality.

[0148] Step S5 aims to establish a dynamic monitoring and control system that moves from "delayed response" to "proactive early warning" and from "one-size-fits-all" to "personalized" approaches. Through intelligent rules and algorithms, it automatically identifies anomalies and opportunities in market operations and guides the precise and agile allocation of resources.

[0149] Step S6 involves quantitatively evaluating the business value and technical effectiveness of the market perception center system itself, and then continuously iterating and optimizing based on the evaluation data and feedback.

[0150] In this embodiment, step S6, which involves quantitatively evaluating the business value and technical effectiveness of the market perception center system itself, and conducting continuous iterative optimization based on the evaluation data and feedback, may specifically include the following steps: S61, Establish and track a multi-dimensional effectiveness evaluation indicator system.

[0151] The average time taken to generate various analytical reports (such as market weekly reports and brand evaluation reports) after using the system was compared with the time spent on manual production, and the percentage of time saved was calculated.

[0152] Market Response Agility: Tracks the average time interval (MTTA) from when the system issues an alert to when business personnel complete an initial response.

[0153] Establish causal analysis models (such as the difference-in-differences method, DID) to compare the differences in key performance indicators (such as the growth rate of high-end cigarette sales and social inventory turnover rate) between regions with different levels of system usage, and isolate the incremental effect contributed by the system.

[0154] Continuously monitor the amount of abnormal data and the success rate of processing detected by the automatic data inspection module; track the source consistency rate of the core "one data source" indicator.

[0155] Periodically (e.g., monthly), calculate the explanatory power (R) of the five major index models on market conditions. 2 The model also includes the prediction error (MAPE) of the sales and structural prediction models, and plots an accuracy trend chart.

[0156] Monitor the timeliness of data processing (such as T+1 data on-time rate), page response time, and concurrent user support capabilities.

[0157] By analyzing system logs, we track the number of active users, access frequency, and usage time for each functional module to identify popular and idle functions. Quarterly online questionnaires collect user (decision-making, management, and execution levels) ratings and qualitative feedback on system data accuracy, user-friendliness, and functional usability. Annually, we select typical users and successful application cases for in-depth interviews to uncover the specific work changes and value stories brought about by the system.

[0158] The purpose of step S61 is to comprehensively and quantitatively measure the changes brought about by the system after its launch from multiple dimensions such as business value, technical efficiency, and user satisfaction, and to use data to prove the return on investment.

[0159] S62, feedback-based model and algorithm iterative optimization process.

[0160] Establish a centralized "System Optimization Dashboard". Record and categorize all issues collected in step S61 (such as "Recent error of XX prediction model has increased," "Users report that a certain indicator of the brand health index is unreasonable"), feedback, and suggestions. The dashboard is linked to specific model versions, functional modules, and data sources.

[0161] For significant algorithm or rule optimizations (such as adjusting the weight of an attribute in the "10A evaluation method"), A / B testing is employed. A small subset of users or regions within the system are designated as the experimental group, applying the new algorithm; the remainder serves as the control group, maintaining the old algorithm. After running for a period, the core performance indicators of the two groups are compared (e.g., whether the success rate of introducing new product specifications in the experimental group has improved), and statistical tests are used to determine the effectiveness of the optimization.

[0162] Strict version control (e.g., using Git) is implemented for all core models (five-dimensional exponential model, predictive model). Any optimization or modification must create a new version. Before going live, new model versions must be backtested offline using historical data. A canary release strategy is adopted during deployment, first enabling the new model with limited traffic to monitor its output stability and business impact. Only after confirming everything is correct is a full release gradually implemented. Each version update records change logs, optimization reasons, and verification results in the knowledge base.

[0163] The purpose of step S62 is to establish a standardized process that transforms the problems and user feedback discovered during the evaluation into specific optimization tasks for the data model, evaluation algorithm, and prediction model, thereby ensuring the continuous evolution of the system's core intelligence.

[0164] S63 involves planning the continuous evolution of technical architecture and data assets.

[0165] Conduct regular system architecture reviews to identify "technical debt" (such as outdated dependency libraries, performance bottleneck code, and unreasonable database design). Develop a technical debt repayment plan and incorporate it into regular development iterations. For example, plan to migrate some batch processing tasks to a real-time stream computing framework (such as Flink) to improve the real-time performance of alerts; consider introducing a microservice architecture to improve the system's scalability and maintainability.

[0166] Graph Neural Networks (GNNs) are used to more accurately construct "consumer-product" transfer networks, providing deeper insights into consumer migration paths. Reinforcement learning is being explored to optimize automated product delivery strategies, allowing the system to autonomously learn optimal control strategies through interaction with the market.

[0167] Generative AI (AIGC) is used to enhance intelligent report generation capabilities, upgrading from "data listing" to "natural language analysis with insights and reasoning." For promising technologies, small pilot projects are established for proof-of-concept (PoC) testing, and large-scale integration is planned only after success.

[0168] The purpose of step S63 is to plan for future upgrades to the system's technical architecture and the introduction of new data and capabilities, ensuring that the system can adapt to business development and technological progress and maintain a forward-looking approach.

[0169] Step S6 aims to establish a scientific method to quantitatively evaluate the business value and technical effectiveness of the market perception center system itself, and based on the evaluation data and feedback, drive the continuous iterative optimization of each component of the system (data, models, applications) to ensure the longevity of the system's vitality and competitiveness.

[0170] The beneficial effects of implementing this embodiment are as follows: by streamlining the data resource collection path, it enables large-scale sorting, integration, and management of data resources; by constructing a market potential prediction model, it enables thorough, accurate, and far-sighted exploration of potential space; by creating market status evaluation standards, it enables three-dimensional, quantitative, and visual status evaluation; by establishing a market monitoring and early warning mechanism, it enables proactive, sensitive, and automated operational monitoring; by building an intelligent economic operation control system, it enables clear, stable, and effective operational regulation; and it successfully integrates big data and artificial intelligence technologies with the specific business scenarios of tobacco industry marketing, providing a feasible, assessable, and evolving methodology for the industry's digital transformation, and overall promoting a fundamental shift in market perception from experience-based and fragmented to scientific and systematic.

[0171] This invention can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This invention can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This invention can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0172] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing related hardware through computer-readable instructions. These computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, or read-only memory (ROM), or random access memory (RAM).

[0173] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0174] Example 2 Further reference Figure 2 As a response to the above Figure 1 The present invention provides an embodiment of a market perception and decision support device, which is similar to the method shown. Figure 1 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0175] like Figure 2As shown, the multidimensional dynamic evaluation device 70 based on multi-source heterogeneous data described in this embodiment includes: a data fusion module 71, an index construction module 72, a model construction module 73, a closed-loop fusion module 74, a control module 75, and an optimization module 76. Wherein: Data fusion module 71 is used to fuse multi-source heterogeneous market data; The index construction module 72 is used to construct a multi-dimensional dynamic evaluation index for assessing market status based on the fused multi-source heterogeneous market data. Model building module 73 is used to build a long-term market forecasting model based on the multidimensional dynamic evaluation index; The closed-loop fusion module 74 is used to perform closed-loop fusion of intelligent algorithms and business scenarios based on the long-cycle market prediction model. The control module 75 is used to construct an intelligent early warning and dynamic control system. Optimization module 76 is used to quantitatively evaluate the business value and technical effectiveness of the market perception center system itself, and to continuously iterate and optimize it based on the evaluation data and feedback.

[0176] The beneficial effects of implementing this embodiment are as follows: by streamlining the data resource collection path, it enables large-scale sorting, integration, and management of data resources; by constructing a market potential prediction model, it enables thorough, accurate, and far-sighted exploration of potential space; by creating market status evaluation standards, it enables three-dimensional, quantitative, and visual status evaluation; by establishing a market monitoring and early warning mechanism, it enables proactive, sensitive, and automated operational monitoring; by building an intelligent economic operation control system, it enables clear, stable, and effective operational regulation; and it successfully integrates big data and artificial intelligence technologies with the specific business scenarios of tobacco industry marketing, providing a feasible, assessable, and evolving methodology for the industry's digital transformation, and overall promoting a fundamental shift in market perception from experience-based and fragmented to scientific and systematic.

[0177] Example 3 To address the aforementioned technical problems, embodiments of the present invention also provide an electronic device. Please refer to [link / reference needed]. Figure 3 , Figure 3 This is a basic structural block diagram of the electronic device in this embodiment.

[0178] The aforementioned electronic device 8 includes a memory 81, a processor 82, and a network interface 83 that are interconnected via a system bus. It should be noted that only the electronic device 8 with components 81, 82, and 83 is shown in the figure; however, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the electronic device described here is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0179] The aforementioned electronic devices can be computing devices such as desktop computers, laptops, handheld computers, and cloud servers. These electronic devices can interact with users via keyboards, mice, remote controls, touchpads, or voice-activated devices.

[0180] The aforementioned memory 81 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the aforementioned memory 81 may be an internal storage unit of the aforementioned electronic device 8, such as the hard disk or memory of the electronic device 8. In other embodiments, the aforementioned memory 81 may also be an external storage device of the aforementioned electronic device 8, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the electronic device 8. Of course, the aforementioned memory 81 may also include both internal storage units and external storage devices of the aforementioned electronic device 8. In this embodiment, the aforementioned memory 81 is typically used to store the operating system and various application software installed on the aforementioned electronic device 8, such as computer-readable instructions based on a multi-dimensional dynamic evaluation method for multi-source heterogeneous data. In addition, the aforementioned memory 81 can also be used to temporarily store various types of data that have been output or will be output.

[0181] In some embodiments, the processor 82 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip. The processor 82 is typically used to control the overall operation of the electronic device 8. In this embodiment, the processor 82 is used to execute computer-readable instructions stored in the memory 81 or to process data, such as executing computer-readable instructions for the multidimensional dynamic evaluation method based on multi-source heterogeneous data.

[0182] The aforementioned network interface 83 may include a wireless network interface or a wired network interface, which is typically used to establish communication connections between the aforementioned electronic device 8 and other electronic devices.

[0183] The beneficial effect of implementing this embodiment is that it promotes a fundamental shift in market perception from experience-based and fragmented to scientific and systematic.

[0184] Example 4 The present invention also provides another embodiment, namely, providing a computer-readable storage medium storing computer-readable instructions that can be executed by at least one processor to cause the at least one processor to perform the steps of the multidimensional dynamic evaluation method based on multi-source heterogeneous data as described above.

[0185] The beneficial effect of implementing this embodiment is that it promotes a fundamental shift in market perception from experience-based and fragmented to scientific and systematic.

[0186] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0187] Obviously, the embodiments described above are merely some embodiments of the present invention, not all embodiments. The accompanying drawings show preferred embodiments of the present invention, but do not limit the patent scope of the present invention. The present invention can be implemented in many different forms; rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the patent protection scope of this invention.

Claims

1. A multidimensional dynamic evaluation method based on multi-source heterogeneous data, characterized in that, Includes the following steps: Integrate heterogeneous market data from multiple sources; Based on the fused multi-source heterogeneous market data, a multi-dimensional dynamic evaluation index for assessing market status is constructed. Based on the aforementioned multidimensional dynamic evaluation index, a long-term market forecasting model is constructed. Based on the aforementioned long-cycle market forecasting model, a closed-loop fusion of intelligent algorithms and business scenarios is performed. Construct an intelligent early warning and dynamic control system; The business value and technical effectiveness of the market perception center system itself are quantitatively evaluated, and continuous iterative optimization is carried out based on the evaluation data and feedback.

2. The multidimensional dynamic evaluation method based on multi-source heterogeneous data according to claim 1, characterized in that, The steps for integrating multi-source heterogeneous market data specifically include: Perform targeted crawling and structured mapping of macroeconomic data; Compliant access and anonymization processing of mobile operator population data; Align and integrate industry data with data from the middle platform; Specifically, the compliance access and desensitization processing of mobile operator population data includes: developing data access microservices based on operator APIs, and using hash algorithms for irreversible encryption or generalization techniques for desensitization processing of the accessed data. The alignment and integration of industry and middle platform data specifically includes: constructing a product specification label mapping dictionary knowledge base; using natural language processing technology to calculate the similarity of product specification names in industry downstream data and middle platform data; establishing same-item-different-name mapping pairs in combination with manual verification; and designing a data cleaning pipeline that includes multi-level verification rules for format verification, logical verification, and outlier handling.

3. The multidimensional dynamic evaluation method based on multi-source heterogeneous data according to claim 1, characterized in that, The step of constructing a multidimensional dynamic assessment index for evaluating market status based on the fused multi-source heterogeneous market data specifically includes: Scientific selection and construction of a multidimensional evaluation index system; Dynamic weight assignment is performed based on the tiered approach; The standardized index values ​​are combined with dynamic weights to calculate the specific value of the multidimensional dynamic evaluation index. The multidimensional dynamic evaluation index includes five dimensions: external environment index, market prosperity index, brand health index, terminal confidence index, and consumer activity index. The steps for dynamic weight assignment based on the grade separation method specifically include: Constructing an evaluation matrix ,in For the number of evaluation objects, For the number of indicators, For the first The object in the first Standardized values ​​for each indicator; calculation matrix ,in for The transpose of the matrix; Solving the matrix Maximum eigenvalue and its corresponding eigenvectors ; to feature vector The optimal weight vector is obtained after normalization. ,in For the first The weights of each indicator make the comprehensive evaluation function... To maximize the variance, For the first The overall evaluation value of each evaluation object.

4. The multidimensional dynamic evaluation method based on multi-source heterogeneous data according to claim 1, characterized in that, The steps for constructing a long-term market forecasting model based on the multidimensional dynamic evaluation index specifically include: A sales potential prediction model that integrates multi-source big data; Construct a predictive model for structural upgrading driven by the macroeconomy; The sales potential prediction model and the structural improvement prediction model are subjected to self-learning and dynamic optimization. The sales potential prediction model integrates a time series model based on operator population data, a smoking rate prediction model based on historical consumer survey data, and a smoothing prediction model based on the trend of per capita smoking volume to calculate and predict market capacity. The structural improvement prediction model is based on macroeconomic indicators and the sales proportion of high-end cigarettes, constructing a multiple linear regression model. ,in This represents the sales proportion of high-end cigarettes. For the selected Key macroeconomic indicators For the regression intercept term, These are the regression coefficients corresponding to each independent variable. The random error term is used in this model to quantify the impact of various macroeconomic indicators on the improvement of cigarette structure and to predict the future proportion of high-end cigarettes.

5. The multidimensional dynamic evaluation method based on multi-source heterogeneous data according to claim 1, characterized in that, The steps for closed-loop integration of intelligent algorithms and business scenarios based on the long-term market prediction model specifically include: Provide a multi-dimensional evaluation method for the introduction of new brands or new product specifications; Establish an early warning system based on hard indicators to identify weak product specifications that should be considered for withdrawal and optimize the brand ecosystem; By applying macro-level indices and forecasts to specific business scenarios, personalized strategy recommendations can be automatically generated. The method described above provides a multi-dimensional evaluation method for the introduction of new brands or new product specifications. Specifically, it includes: constructing a 10A evaluation method with two dimensions of product specification attributes and status, quantitatively scoring the attribute and status dimensions of candidate product specifications, generating a visual report and displaying it in order of A number. The establishment of a hard indicator-based early warning system specifically includes: defining and calculating four types of hard indicators: low market share, poor growth, high inventory-to-sales ratio, and price inversion; marking product specifications with red, orange, yellow, and blue cards based on indicator triggering conditions for visual early warning; and providing auxiliary decision-making information such as correlation analysis, historical cases, and simulations.

6. The multidimensional dynamic evaluation method based on multi-source heterogeneous data according to claim 1, characterized in that, The specific steps for constructing the intelligent early warning and dynamic control system include: Dynamic threshold early warning is based on the 3σ rule; Develop a scenario-based multi-dimensional operation and control system; Automatic distribution and tracking of intelligent early warnings; Specifically, the dynamic threshold early warning based on the 3σ rule includes: calculating the mean of historical data for the indicator. and standard deviation Set the warning threshold to ,in The average of the historical data of the indicator. Standard deviation, Adjustable parameters are set based on business risk tolerance and are periodically recalculated using the latest rolling historical data. and Update the threshold; The scenario-based construction of the multi-dimensional operation and control system includes: developing a benchmarking analysis engine to support flexible configuration of benchmarking dimensions; constructing a collaborative data view between industry and commerce to conduct penetrating analysis from three dimensions: scale, structure, and efficiency; establishing a holiday analysis template; and developing a barometer-spotlight tool for intelligent management of terminal operations, used for tiered monitoring of terminal confidence index and root cause analysis of problem tiers.

7. The multidimensional dynamic evaluation method based on multi-source heterogeneous data according to any one of claims 1 to 6, characterized in that, The steps of quantitatively evaluating the business value and technical effectiveness of the market perception center system itself, and conducting continuous iterative optimization based on the evaluation data and feedback, specifically include: Establish and track a multi-dimensional performance evaluation indicator system; Feedback-based iterative optimization process for models and algorithms; Plan for the continuous evolution of technical architecture and data assets; The multi-dimensional effectiveness evaluation index system includes efficiency improvement indicators and market response agility indicators in the business value dimension, data quality indicators and model accuracy indicators in the technical effectiveness dimension, and function usage analysis indicators and qualitative feedback indicators in the user satisfaction dimension. The feedback-based model and algorithm iterative optimization process includes: establishing a system optimization dashboard to centrally manage issues and feedback, using A / B testing to verify the effects of important algorithm optimizations, and implementing version control and canary release strategies for core models; The continuous evolution plan for the technical architecture and data assets includes: regularly reviewing and identifying the system architecture and repaying technical debt, and planning to introduce new technologies such as graph neural networks, reinforcement learning, and generative AI for pilot verification.

8. A multidimensional dynamic evaluation device based on multi-source heterogeneous data, characterized in that, include: The data fusion module (71) is used to fuse multi-source heterogeneous market data; The index construction module (72) is used to construct a multi-dimensional dynamic evaluation index for evaluating market status based on the fused multi-source heterogeneous market data; The model building module (73) is used to build a long-term market forecasting model based on the multidimensional dynamic evaluation index. The closed-loop fusion module (74) is used to perform closed-loop fusion of intelligent algorithms and business scenarios based on the long-cycle market prediction model. The control module (75) is used to construct an intelligent early warning and dynamic control system; The optimization module (76) is used to quantitatively evaluate the business value and technical effectiveness of the market perception center system itself, and to continuously iterate and optimize based on the evaluation data and feedback.

9. An electronic device, characterized in that, The method includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of the multidimensional dynamic evaluation method based on multi-source heterogeneous data as described in any one of claims 1 to 7.

Images