Cigarette product demand feature extraction method and device, electronic device, and storage medium

Abstract

The application relates to a cigarette product demand feature extraction method and device, an electronic device and a storage medium. The cigarette product demand feature extraction method comprises the following steps: acquiring multi-source data of a cigarette product market state; dividing the multi-source data into multiple data blocks, identifying and removing abnormal values of the data blocks through a three-sigma principle to obtain an abnormal value identification result; performing cross-validation and merging on the abnormal value identification result, determining abnormal data in the data blocks, removing the abnormal data from the multi-source data, and obtaining a unified data set; calculating characteristic values of cigarette product demand according to the unified data set, and performing dimension unification on the characteristic values of the cigarette product demand through an extreme value standardization method to obtain characteristic parameters of the cigarette product demand. Through the division and abnormal value identification of the multi-source data, and the parallel verification of the abnormal value identification result, the abnormal identification efficiency of large-scale market state data is improved.

Description

Technical Field

[0001] This application relates to the field of tobacco market status data processing technology, and in particular to methods, apparatus, electronic devices, and storage media for extracting demand characteristics of cigarette products. Background Technology

[0002] By collecting, processing, and analyzing market data on cigarette product price performance, social inventory levels, and terminal demand, we can quantitatively assess the supply and demand situation and operational quality of cigarette products in different regional markets, thereby providing data support for the formulation and adjustment of cigarette distribution strategies.

[0003] Currently, traditional methods often rely on the experience and judgment of account managers or use a single indicator (such as retail price) for simple threshold classification. While related technologies can achieve weighted scoring based on multiple indicators, data access and processing are still primarily batch-based, with limited capabilities for efficient verification and real-time calculation of large-scale, multi-source data. Furthermore, market status data, sales data, and marketing activity data are scattered across different systems, with inconsistent data quality and a lack of effective cleaning, correlation, and fusion mechanisms, forming "data silos" and resulting in low efficiency in analyzing and calculating cigarette product market status data.

[0004] There is currently no effective solution to the problem of low efficiency in analyzing and calculating market status data for cigarette products in related technologies. Summary of the Invention

[0005] This embodiment provides a method, apparatus, electronic device, and storage medium for extracting demand characteristics of cigarette products, in order to solve the problem of low efficiency in analyzing and calculating market status data of cigarette products in related technologies.

[0006] Firstly, this embodiment provides a method for extracting demand features of cigarette products, including:

[0007] Obtain multi-source data on the market status of cigarette products;

[0008] The multi-source data is divided into multiple data blocks, and outlier identification and removal are performed on the data blocks using the three Sigma principle to obtain outlier identification results.

[0009] The outlier identification results are cross-validated and merged to determine the outlier data in the data block, and the outlier data is removed from the multi-source data to obtain a unified dataset;

[0010] The characteristic values ​​of cigarette product demand are calculated based on the unified dataset, and the dimensions of the characteristic values ​​of cigarette product demand are unified by the extreme value standardization method to obtain the characteristic parameters of cigarette product demand.

[0011] In some embodiments, the step of calculating the feature values ​​of cigarette product demand based on the unified dataset and unifying the dimensions of the feature values ​​of cigarette product demand using the extreme value standardization method to obtain the feature parameters of cigarette product demand includes: calculating the feature values ​​of time-sensitive cigarette product demand using incremental updates; wherein:

[0012] Store the cached values, timestamps, and version numbers of the characteristic values ​​required for each cigarette product;

[0013] When new data arrives in the unified dataset, the incremental indicator value is calculated from the new data.

[0014] The incremental indicator value and the cached value are weighted and merged according to time weight to obtain the updated cached value;

[0015] The characteristic parameters of the cigarette product requirements are calculated and updated based on the updated cache value.

[0016] In some embodiments, the method for extracting demand features of cigarette products further includes:

[0017] A decision-making model is constructed to conduct quantitative analysis of cigarette products from different dimensions, calculate the contribution rate of historical sales growth of cigarette products from different dimensions, and obtain decomposed targets for different dimensions based on the contribution rate and the preset overall cultivation target. The contribution rate and the decomposed targets are used as the analysis results of the decision-making model. The decision-making model is used to conduct quantitative analysis of the contribution rate of cigarette products from different dimensions and generate quantitative analysis models of decomposed targets for each dimension.

[0018] Based on the characteristic parameters of the demand for the cigarette products and the analysis results of the decision model, a collaborative strategy for the cigarette products is formulated.

[0019] The collaborative strategy is transformed into specific instructions through a preset strategy rule base, a combined cultivation task package is generated according to the specific instructions, and the combined cultivation task package is executed according to the corresponding cigarette product.

[0020] In some embodiments, the construction of the decision model involves quantitative analysis of cigarette products across different dimensions, calculating the contribution rate of historical sales growth of cigarette products in different dimensions, and obtaining decomposed targets in different dimensions based on the contribution rate and a preset overall development target. The contribution rate and the decomposed targets are then used as the analysis results of the decision model, including:

[0021] Based on two-factor analysis, the contribution rate of product growth indicators to the historical sales growth of the cigarette products was quantified.

[0022] Based on the contribution rate and the preset overall cultivation target, the preset overall cultivation target is decomposed into quantitative sub-targets of different dimensions to obtain the decomposed target.

[0023] In some embodiments, formulating a collaborative strategy for the cigarette product based on the characteristic parameters of the cigarette product demand and the analysis results of the decision model includes:

[0024] When it is identified that the market status of a cigarette product belongs to a preset category that requires a reduction in supply, the supply of cigarette products that have a complementary relationship with the cigarette product is increased according to a preset complementarity coefficient.

[0025] When the cigarette product is in the discontinuation or version switching stage, the average household distribution amount adjustment value of related cigarette products that have a substitution relationship with the cigarette product is calculated according to the preset transfer coefficient, and the average household distribution amount of the related cigarette products is modified according to the average household distribution amount adjustment value.

[0026] Generate and manage combined display schemes, joint promotional activities, and collaborative launch plans for multiple cigarette products.

[0027] In some embodiments, the step of converting the collaborative strategy into specific instructions through a preset strategy rule base, generating a combined cultivation task package based on the specific instructions, and distributing the combined cultivation task package to the cigarette product business system for execution includes:

[0028] Based on the characteristic parameters of the cigarette product demand and the analysis results of the decision model, the system matches and sorts the data in the preset strategy rule base to generate a strategy package that includes specific measures, meets resource needs, and has expected effects.

[0029] The strategy package is converted into specific instructions, a combined cultivation task package is generated based on the specific instructions, and the combined cultivation task package is sent to the cigarette product business system for execution.

[0030] In some embodiments, the method for extracting demand features of cigarette products further includes:

[0031] Monitor the execution effect of the task package and obtain feedback data on the execution effect;

[0032] The policy rule base is optimized based on the feedback data.

[0033] Secondly, this embodiment provides a device for extracting the demand characteristics of cigarette products, including: a data unification module and a scoring calculation module; wherein:

[0034] The data unification module is used to acquire multi-source data on the market status of cigarette products; divide the multi-source data into multiple data blocks, and perform outlier identification and removal on the data blocks using the three Sigma principle to obtain outlier identification results; perform cross-validation and merging on the outlier identification results to determine the abnormal data in the data blocks, and remove the abnormal data from the original data to obtain a unified dataset;

[0035] The scoring calculation module is used to calculate the feature values ​​of cigarette product demand based on the unified dataset, and to unify the dimensions of the feature values ​​of cigarette product demand through the extreme value standardization method to obtain the feature parameters of cigarette product demand.

[0036] Thirdly, this embodiment provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the cigarette product demand feature extraction method described in the first aspect above.

[0037] Fourthly, this embodiment provides a storage medium storing a computer program that, when executed by a processor, implements the steps of the cigarette product demand feature extraction method described in the first aspect above.

[0038] Compared with related technologies, the method, apparatus, electronic device, and storage medium for extracting cigarette product demand features provided in this embodiment are as follows: First, the method acquires multi-source data on the market status of cigarette products. Second, it divides the multi-source data into multiple data blocks and identifies and removes outliers from these blocks using the three Sigma principle, obtaining outlier identification results. Next, it cross-validates and merges the outlier identification results to determine the abnormal data within the data blocks and removes these outliers from the multi-source data, obtaining a unified dataset. Finally, it calculates the feature values ​​of cigarette product demand based on the unified dataset and unifies the dimensions of these feature values ​​using extreme value standardization, obtaining the feature parameters of cigarette product demand. By dividing the multi-source data and identifying outliers, and by performing parallel verification of the outlier identification results, the efficiency of anomaly identification for large-scale market status data and the efficiency of analyzing and calculating cigarette product market status data are improved.

[0039] Details of one or more embodiments of this application are set forth in the following drawings and description to make other features, objects and advantages of this application more readily apparent. Attached Figure Description

[0040] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0041] Figure 1 This is a hardware structure block diagram of a terminal for a cigarette product demand feature extraction method according to an embodiment of this application;

[0042] Figure 2 This is a flowchart of a method for extracting demand features of cigarette products according to an embodiment of this application;

[0043] Figure 3 This is a schematic diagram of the data segmentation and parallel verification architecture of a cigarette product demand feature extraction method according to an embodiment of this application;

[0044] Figure 4 This is a schematic diagram of the cigarette product linkage cultivation process of a cigarette product demand feature extraction method according to an embodiment of this application;

[0045] Figure 5 This is a flowchart illustrating the incremental update and sliding window mechanism of a cigarette product demand feature extraction method according to an embodiment of this application.

[0046] Figure 6 This is a flowchart of a method for extracting demand features of cigarette products according to another embodiment of this application;

[0047] Figure 7 This is a structural block diagram of a cigarette product demand feature extraction device according to an embodiment of this application. Detailed Implementation

[0048] To better understand the purpose, technical solution, and advantages of this application, the application is described and illustrated below in conjunction with the accompanying drawings and embodiments.

[0049] Unless otherwise defined, the technical or scientific terms used in this application shall have the general meaning understood by a person skilled in the art to which this application pertains. Words such as “a,” “an,” “an,” “the,” “the,” and “these” used in this application do not indicate quantitative limitation and may be singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that comprises a series of steps or modules (units) is not limited to the listed steps or modules (units) but may include steps or modules (units) not listed, or may include other steps or modules (units) inherent to these processes, methods, products, or devices. Words such as “connected,” “linked,” and “coupled” used in this application are not limited to physical or mechanical connections but may include electrical connections, whether direct or indirect. “Multiple” used in this application refers to two or more. “And / or” describes the relationship between related objects, indicating that three relationships may exist; for example, “A and / or B” can represent: A alone, A and B simultaneously, or B alone. Normally, the character " / " indicates that the objects before and after it are in an "or" relationship. The terms "first," "second," "third," etc., used in this application are merely to distinguish similar objects and do not represent a specific order of objects.

[0050] The method embodiments provided in this example can be executed on a terminal, computer, or similar electronic device with a certain computing power. For example, it can run on a terminal. Figure 1 This is a hardware structure block diagram of a terminal for a cigarette product demand feature extraction method according to an embodiment of this application. For example... Figure 1 As shown, a terminal may include one or more ( Figure 1 Only one is shown in the diagram. A processor 102 and a memory 104 for storing data are also included. The processor 102 may be, but is not limited to, a microprocessor (MCU) or a programmable logic device (FPGA). The terminal may also include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that… Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the terminal described above. For example, the terminal may also include components that are larger than... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown are illustrated.

[0051] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the cigarette product demand feature extraction method in this embodiment. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, thereby implementing the above-described method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0052] The transmission device 106 is used to receive or send data via a network. This network includes a wireless network provided by the terminal's communication provider. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 can be a Radio Frequency (RF) module used for wireless communication with the Internet.

[0053] This embodiment provides a method for extracting demand features of cigarette products. Figure 2 This is a flowchart of a method for extracting demand features of cigarette products according to an embodiment of this application, such as... Figure 2 As shown, the process includes the following steps:

[0054] Step S210: Obtain multi-source data on the market status of cigarette products.

[0055] Specifically, the system connects to the price reporting system, inventory reporting system, and order system, collecting real-time data on the market status of cigarette products from each system. Price reporting refers to retailers directly reporting the actual retail price of cigarettes they sell through the price reporting system. Inventory reporting refers to retailers reporting the quantity of cigarette products they have in stock through the inventory reporting system. The order system records retailers' ordering behavior data, including indicators such as purchase volume, order fulfillment rate, and order fulfillment area. The data obtained from the price reporting system, inventory reporting system, and order system constitute multi-source data on the market status of cigarette products.

[0056] Step S220: Divide the multi-source data into multiple data blocks, and identify and remove outliers in the data blocks using the three Sigma principle to obtain outlier identification results.

[0057] During the collection and transmission of multi-source data, outliers are inevitable due to factors such as manual data entry errors by retailers, system terminal malfunctions, network transmission anomalies, or differences in data formats between different data sources. Examples include retailers mistakenly entering 1000 yuan as 100 yuan in price data, and sudden numerical changes caused by false alarms from terminal sensors in inventory data. If these outliers are not effectively identified and removed, they will directly affect the accuracy of subsequent price index calculations, the standardization of indicator units, and market status assessments. Therefore, it is necessary to perform outlier identification and removal processing on the collected multi-source data.

[0058] Specifically, the multi-source data is first divided into multiple data blocks according to preset partitioning dimensions. These preset partitioning dimensions include time and / or geographic dimensions. Partitioning by time means grouping data generated from the same data source within a continuous time window into one data block; for example, data collected within the same week, day, or hour is grouped into one data block. Partitioning by geographic dimension means grouping data collected within the same geographic region (such as at the prefecture-level city, district / county level, or based on a grid) into one data block.

[0059] Then, the divided data blocks are distributed to parallel computing nodes. Each computing node independently performs outlier identification and removal using the Three Sigma principle for its assigned data block. The Three Sigma principle (3σ principle) is a statistical outlier identification method based on the assumption of normal distribution. Specifically, for data that follows a normal or approximately normal distribution, the probability of its values ​​falling within the range of mean (μ) ± 3 standard deviations (σ) is approximately 99.73%, while the probability of data points falling outside this range is only 0.27%, which is a low-probability event and can therefore be identified as outliers. Therefore, for each indicator data in each data block, the mean μ and standard deviation σ of that indicator are calculated for that data block. Data values ​​greater than μ + 3σ or less than μ - 3σ are identified as outliers and removed from that data block. After each computing node completes the outlier identification and removal, the outlier identification results for each data block are output. The outlier identification results include the records of outliers that have been removed from each data block and the remaining data after removal.

[0060] Step S230: Cross-validate and merge the outlier identification results to identify outlier data in the data block and remove the outlier data from the multi-source data to obtain a unified dataset.

[0061] Since outlier identification is performed independently for each data block, the same data record may be identified as outlier by some nodes and normal by others depending on the data block it belongs to. Therefore, it is necessary to cross-validate the outlier identification results of each data block in order to comprehensively determine whether the data is a real outlier.

[0062] Specifically, for suspected outliers located near data block boundaries (such as records within the first and last 5% of a data block), a secondary cross-data block verification is performed. The mean μ1 and standard deviation σ1 of the data block containing the suspected outlier are extracted. Simultaneously, the mean μ2 and standard deviation σ2 of the subsets of data points adjacent to the data point in neighboring data blocks are extracted. The standard scores Z1=(x-μ1) / σ1 and Z2=(x-μ2) / σ2 for the data point in both data block distributions are calculated. If the absolute values ​​of Z1 and Z2 are both greater than 3, the data point is confirmed as an outlier. If the absolute value of the standard score is greater than 3 only in one data block but less than or equal to 3 in adjacent data blocks, it indicates that the data point falls within the normal range in the distribution of adjacent data blocks, and the outlier determination is a misjudgment caused by the current data block boundary effect. Therefore, the outlier determination is revoked, and the data point is restored to normal data. After cross-validation, based on the confirmed final outlier determination results, all data records ultimately determined to be outliers are removed from the original multi-source data, retaining only those ultimately determined to be normal. The retained normal data records from each data block are then concatenated according to a preset chronological order or geographic region numbering to form a unified dataset.

[0063] After the aforementioned cross-validation and boundary secondary verification, based on the confirmed final outlier determination results, all data records ultimately determined to be outliers are removed from the original multi-source data, retaining only those ultimately determined to be normal. The normal data records retained from each data block are then concatenated according to a preset merging order (e.g., chronological order or geographic region number order) to form a unified dataset for subsequent indicator calculations and market status evaluation. By distributing the overall computational task across multiple nodes for parallel execution, the computational time complexity is reduced from O(n) to O(n / m) (where m is the number of parallel computing nodes), shortening the processing latency for data verification in large-scale data scenarios.

[0064] Step S240: Calculate the characteristic values ​​of cigarette product demand based on the unified dataset, and unify the dimensions of the characteristic values ​​of cigarette product demand through the extreme value standardization method to obtain the characteristic parameters of cigarette product demand.

[0065] Specifically, the characteristic values ​​of cigarette product demand can include the market price index, retail price index, number of days of available inventory, procurement availability, order fulfillment rate, and order fulfillment rate. Among these, higher market price and retail price indices indicate more stable market prices, shorter number of days of available inventory indicates faster market sales, and higher procurement availability, order fulfillment rate, and order fulfillment rate indicate stronger demand from end consumers. These characteristic parameters can reflect the market demand for cigarette products from different perspectives.

[0066] Because the characteristic values ​​of the aforementioned cigarette product demands have different dimensions—for example, the market price index and retail price index are in yuan, the number of days of social inventory available for sale is in days, while the purchase volume, order fulfillment rate, and order fulfillment volume are all percentage values—the numerical ranges of these characteristic values ​​differ significantly, making direct comprehensive comparison and weighted fusion impossible. Therefore, the extreme value standardization method can be used to unify the dimensions of the characteristic values ​​of the cigarette product demands. Extreme value standardization is a dimensionless processing method that linearly transforms the original data to a unified numerical range. Its calculation formula is: Standardized value = (Original value - Minimum value) / (Maximum value - Minimum value), where the maximum value is the maximum value of the characteristic value within a preset historical period, and the minimum value is the minimum value of the characteristic value within the preset historical period. After extreme value standardization, all characteristic values ​​are mapped to the interval between 0 and 1, thus preserving the relative magnitude of the original data while eliminating the impact of differences in dimensions and orders of magnitude on subsequent comprehensive evaluation. The standardized values ​​obtained after extreme value standardization constitute the characteristic parameters of cigarette product demand, used for subsequent comprehensive evaluation of the cigarette product market status.

[0067] Compared to related technologies that use multi-indicator weighted scoring to comprehensively evaluate the market status of cigarette products, this approach still relies heavily on batch processing for data access and processing, resulting in weak capabilities for efficient verification and real-time calculation of large-scale, multi-source data. Furthermore, different business systems (such as retailer reporting systems, order systems, and marketing platforms) employ different data formats and transmission protocols, leading to inconsistent data standards and a lack of unified data verification rules. This results in frequent occurrences of outliers, missing values, and format inconsistencies, making it difficult to directly perform correlation analysis and fusion calculations on multi-source data. Calculating core indicators such as price indices and order fulfillment rates typically requires traversing and calculating the entire historical data set. As the scale of terminal data grows, computational complexity and storage costs increase, leading to increased data processing latency. This makes it difficult to meet the real-time or near-real-time processing requirements of large-scale, multi-source data scenarios, resulting in low efficiency in analyzing and calculating the market status of cigarette products.

[0068] In steps S210 to S240 above, this method for extracting the demand features of cigarette products first acquires multi-source data on the market status of cigarette products; second, it divides the multi-source data into multiple data blocks and identifies and removes outliers from these blocks using the three Sigma principle, obtaining outlier identification results; next, it cross-validates and merges the outlier identification results to determine the outliers in the data blocks and removes them from the multi-source data, obtaining a unified dataset; finally, it calculates the feature values ​​of cigarette product demand based on the unified dataset and unifies the dimensions of these feature values ​​using extreme value standardization, obtaining the feature parameters of cigarette product demand. By dividing the multi-source data and identifying outliers, and then performing parallel verification of the outlier identification results, the efficiency of anomaly identification for large-scale market status data and the efficiency of analyzing and calculating cigarette product market status data are improved.

[0069] Optionally, in one embodiment, feature values ​​of cigarette product demand are calculated based on a unified dataset, and the feature values ​​of cigarette product demand are standardized in terms of dimensions using an extreme value standardization method to obtain feature parameters of cigarette product demand. This includes: calculating feature values ​​of time-sensitive cigarette product demand using incremental updates; wherein, the cached value, timestamp, and version number of each cigarette product demand feature value are stored; when new data arrives in the unified dataset, incremental index values ​​are calculated on the new data; the incremental index values ​​and cached values ​​are weighted and fused according to time weights to obtain updated cached values; and the feature parameters of cigarette product demand are calculated and updated based on the updated cached values.

[0070] Specifically, retailers can update price and inventory data daily through the price reporting system and inventory reporting system. The order system continuously generates new order fulfillment rates and order fulfillment figures during the delivery cycle, and all characteristic values ​​change dynamically over time. Therefore, the characteristic values ​​calculated at the current moment become historical values ​​after a certain period of time, requiring continuous updates based on the latest data to accurately reflect the real-time market demand for cigarette products. However, as the scale of terminal data grows, recalculating all historical data with each update would accumulate computational load over time, leading to a continuous decline in data processing efficiency. Therefore, this embodiment employs an incremental update mechanism to calculate the characteristic values ​​of time-sensitive cigarette product demand.

[0071] For each cigarette product's required feature value, a cached value, timestamp, and version number are maintained for each feature value. The cached value stores the currently calculated feature value result, serving as the basis for subsequent incremental updates; the timestamp records the data time corresponding to the cached value, indicating its timeliness; and the version number indicates the number of times the cached value has been updated, preventing duplicate processing of the same batch of new data.

[0072] When new data arrives in the unified dataset, the cached values ​​of the feature values ​​for each cigarette product's current demand, along with their corresponding timestamps and version numbers, are first read. Then, only the newly added data is used for indicator calculation, rather than recalculating all historical data, to obtain incremental indicator values. Next, according to a preset time weighting strategy, the newly calculated incremental indicator values ​​are weighted and fused with the cached original feature values. The fusion formula is as follows: Among them, V cache ΔV represents the cached historical indicator values, ΔV represents the indicator change calculated from the newly added data, and λ represents the time decay factor, where 0 < λ < 1. It can be configured according to the timeliness requirements of the indicator. For example, a larger λ value can be used for indicators with higher timeliness requirements, so that the newly added data has a higher weight.

[0073] Time weights are determined based on the relative duration of the time intervals corresponding to newly arrived data and those corresponding to cached values. Newly arrived data is assigned higher weights to reflect changes in the current market state, while historical cached values ​​are weighted according to a preset decay coefficient. This preserves historical trend information while highlighting the timeliness contribution of recent data. After weighted fusion, the fusion result is stored as updated feature values, with the corresponding timestamp updated to the current data time and the version number incremented. Finally, based on the updated cached values ​​of each feature value, the characteristic parameters of cigarette product demand are recalculated for subsequent market state evaluation.

[0074] The incremental update mechanism described above allows for calculation of only the newly added data and weighted fusion with the cached value during each data update, avoiding redundant calculation of the entire dataset. This effectively reduces computational resource consumption and storage overhead, and improves the real-time performance and computational efficiency of feature parameter extraction in large-scale, multi-source data scenarios.

[0075] In one embodiment, the method for extracting cigarette product demand characteristics further includes: for indicators that need to be recalculated periodically, a sliding window mechanism is used for indicator calculation and data management. Periodically recalculated indicators refer to indicators that need to be recalculated according to a fixed time period (such as daily, weekly, or monthly). Taking the weekly average price index as an example, this indicator needs to be recalculated based on the price reporting data within the most recent week to reflect the price level of cigarette products within the current time window and provide updated indicator data for cigarette product demand.

[0076] As data collection continues, the scale of the original data continues to grow. If all historical data is traversed and calculated every time a recalculation is performed, the amount of data will accumulate over time, and the calculation time will increase linearly. Therefore, in this embodiment, a sliding window mechanism is adopted, which only retains the data from the most recent N periods for indicator calculation. Historical data outside this time window is automatically transferred to offline archive storage and no longer participates in real-time or near-real-time indicator calculation.

[0077] For example, taking the calculation of the weekly average price index as an example, the sliding window size is set to 4 weeks, using the natural week as the period unit. When new week's data arrives, the newly added price reports for that week are included in the calculation window, while the earliest week's data that exceeds the window range is removed from the calculation window. The removed data is automatically transferred to offline archive storage. The data within the window always remains the complete dataset of the most recent 4 weeks, and the weekly average price index is recalculated only based on the data within this 4-week window.

[0078] By employing the sliding window mechanism described above, only a limited amount of data within the window needs to be processed each time an indicator is recalculated. The computational load does not increase indefinitely with the data collection time, effectively controlling the scale of online data storage and avoiding the storage bloat problem caused by the infinite accumulation of historical data. The time and space complexity of indicator calculation is reduced from O(T×D) (where T is the number of time periods and D is the amount of data) to O(ΔD) (where ΔD is the incremental data amount). Simultaneously, the sliding window archiving mechanism effectively controls the growth of online data storage, ensuring stable computational performance and storage efficiency during long-term operation.

[0079] Furthermore, in one embodiment, the above-mentioned method for extracting the demand characteristics of cigarette products may further include: constructing a decision model to perform quantitative analysis of cigarette products from different dimensions, calculating the contribution rate of historical sales growth of cigarette products from different dimensions, and obtaining decomposed targets for different dimensions based on the contribution rate and a preset overall cultivation target, using the contribution rate and decomposed targets as the analysis results of the decision model; formulating a collaborative strategy for cigarette products based on the characteristic parameters of cigarette product demand and the analysis results of the decision model; converting the collaborative strategy into specific instructions through a preset strategy rule base, generating a combined cultivation task package based on the specific instructions, and executing the combined cultivation task package according to the corresponding cigarette products; the decision model is used to perform quantitative analysis of the contribution rate of cigarette products from different dimensions and generate a quantitative analysis model for the decomposed targets of each dimension.

[0080] The decision-making model is a quantitative analysis model used to quantify the contribution rate of different dimensions of cigarette product growth drivers, and to generate decomposed targets for each dimension based on a preset overall development goal. These different dimensions include four aspects: penetration (G), price power (O), repurchase rate (R), and expansion power (W). Penetration aims to scientifically measure the breadth of product market coverage and the quality of covered terminals. The specific calculation formula is: G = Terminal Coverage Rate × Coverage Quality Coefficient × K1. Where, Terminal Coverage Rate = Na / Nt; where Na is the number of terminals that actually purchased the product within the statistical period, and terminals refer to retail locations in the tobacco industry that directly sell cigarettes to consumers; Nt is the total number of target terminals. The determination of Nt is dynamically generated based on a multi-factor objective rule model based on geographical grids, business district types, and historical sales scale levels. Coverage Quality Coefficient = Among them, S i T represents the current period's purchase volume of this product for terminal i; j This represents the total purchase volume of terminal j in the same price range or product category during the current period. The coverage quality coefficient reflects the purchase share of the covered terminals for this product. K1 is the normalization coefficient, used to convert the result into an easily understandable percentage or exponential form. Penetration power can eliminate the influence of subjectively defining the terminal scope, and through quality weighting, the penetration power indicator can truly reflect the effectiveness and health of product channel construction.

[0081] Price power can comprehensively assess a product's price compliance and market stability, thus avoiding the bias of comparing a single mean. The formula for price power is: O = w1 × Price Compliance + w2 × Price Stability. Where, Price Compliance = Pa / Pt; Pa is the actual weighted average retail price of the product within the statistical period; Pt is the benchmark value for the product's target price range. Price Stability = σ represents the standard deviation of terminal prices for each sample within the statistical period, reflecting the degree of price dispersion. The smaller the dispersion, the higher the stability. w1 and w2 are configurable weighting coefficients (w1+w2=1), which can be adjusted according to the product life cycle. Price power can simultaneously consider both "whether the price is at the ideal level" and "whether prices are consistent across stores," enabling a more comprehensive and robust assessment of the health of the pricing system.

[0082] Repeat purchase power integrates actual repeat purchase behavior data and order intention data from terminals, providing a more reliable assessment of customer loyalty. The formula for repeat purchase power is: R = α × Behavioral Repeat Purchase Rate + β × Intended Repeat Purchase Intensity. Where, Behavioral Repeat Purchase Rate = Mr / Ma; Mr is the number of terminals that made two or more purchases within the statistical period; Ma is the total number of terminals that made purchases within the statistical period (i.e., Na); the behavioral repeat purchase rate directly measures the actual repeat purchase rate of terminals. Intended Repeat Purchase Intensity = I iThe order fulfillment rate or order fulfillment coverage data for terminal i reflects the strength of its willingness to order. The order fulfillment rate is the ratio of the order quantity of a specific customer to the supply limit within a certain period; the order fulfillment coverage is the ratio of the number of customers who have fulfilled the supply limit to the total number of customers within a certain period. α and β are harmonic coefficients used to balance the reliability and availability of objective behavioral data and subjective intention data. Furthermore, by combining the behavioral signals of "whether to buy repeatedly" and the intention signals of "whether to buy more," the repurchase power assessment is made closer to market reality, avoiding the bias of relying solely on order intention data.

[0083] Expansion capability assesses the relative level of sales depth at a single point within the already covered terminals, revealing the potential for further expansion of existing terminals. The formula for expansion capability is: W = (Average sales volume of this product at a single point / Average sales volume of the reference category at a single point) × K². Where, average sales volume of this product at a single point = Qa / Na. Qa is the total sales volume of this product within the covered terminals (Na); average sales volume of the reference category at a single point = Q... total / Nt. Q total The total sales volume of the selected reference product category across all target terminals (Nt); K2 is the scaling factor. The extensibility index is a relative value. When W is greater than 100, it indicates that the product's single-point sales capability in covered terminals is higher than the overall market average, indicating deeper market penetration. When W is less than 100, it indicates that although coverage is achieved, the single-point output efficiency is lower than the overall market, suggesting significant room for improvement. Extensibility allows for standardization by incorporating a market benchmark, eliminating absolute sales differences caused by varying terminal sizes and locations, making extensibility a purely comparable indicator reflecting "single-point penetration efficiency."

[0084] Specifically, the decision-making model conducts quantitative analysis of cigarette products across different dimensions: First, it obtains the quantitative performance values ​​of the cigarette product's penetration, pricing power, repurchase rate, and sales extension power within the current analysis period. Then, based on a two-factor analysis, it calculates the contribution ratio of the changes in each dimension relative to the previous comparison period to the historical sales growth of the cigarette product, obtaining the contribution rate of each dimension. Next, based on the contribution rates of each dimension and the preset overall development target, the preset overall development target is decomposed into the quantitative improvement targets required for each dimension, serving as the decomposed targets for each dimension. Finally, the contribution rates and decomposed targets are output as the analysis results of the decision-making model.

[0085] Then, based on the characteristic parameters of cigarette product demand and the analysis results of the decision model, a collaborative strategy for cigarette products is formulated. The characteristic parameters of cigarette product demand are standardized values ​​of various demand characteristics obtained after extreme value standardization, including the standardized results of indicators such as market price index, retail price index, social inventory days available for sale, procurement coverage, order fulfillment rate, and order fulfillment coverage. These characteristic parameters reflect the performance level of various demand characteristics of cigarette products under the current market conditions. The analysis results of the decision model provide the contribution rate of each dimension to historical sales growth and the decomposed targets to be achieved in the future. The contribution rate and decomposed targets together provide a complete data foundation and quantitative objectives for the formulation of collaborative strategies. These collaborative strategies include complementary placement strategies, cluster promotion strategies, and quantitative linkage strategies. The complementary placement strategy automatically recommends similar price-range or series products for replenishment when the placement of a certain product needs to be reduced due to status maintenance, ensuring overall sales stability. The cluster promotion strategy supports multi-product combination displays, joint promotion, and collaborative placement of derivative products, enhancing the brand cluster effect. The quantitative linkage strategy automatically adjusts the average quantity per household for related products during product switching or discontinuation phases, smoothing sales fluctuations.

[0086] Finally, the collaborative strategy is transformed into specific instructions through a pre-defined strategy rule base. Based on these instructions, a combined cultivation task package is generated and executed according to the corresponding cigarette product. The strategy rule base stores placement strategy rules and marketing activity rules defined in the form of "IF-THEN-WITH". Once the collaborative strategy is determined, the strategy rule base matches the corresponding strategy rules with the market status level calculated based on the characteristic parameters of the current cigarette product demand and the analysis results of the decision model. This transforms the collaborative strategy into executable specific instructions, which are then further encapsulated into combined cultivation task packages. These packages are then distributed to the corresponding business systems for execution according to the cigarette product, region, and execution parameters specified in the task package.

[0087] By constructing a decision-making model, the characteristic values ​​of cigarette product demand can be transformed into quantifiable, traceable, and actionable decomposition targets in each growth driver dimension, thereby providing a clear quantitative basis for the formulation of subsequent collaborative strategies.

[0088] In one embodiment, a decision-making model is constructed to perform quantitative analysis of cigarette products from different dimensions, calculate the contribution rate of historical sales growth of cigarette products from different dimensions, and obtain decomposed targets for different dimensions based on the contribution rate and the preset overall development target. The contribution rate and decomposed targets are used as the analysis results of the decision-making model, including:

[0089] Based on the two-factor analysis method, the contribution rate of product growth indicators to the historical sales growth of cigarette products is quantified; according to the contribution rate and the preset overall development target, the preset overall development target is decomposed into quantitative sub-targets of different dimensions to obtain the decomposed target.

[0090] After calculating the specific values ​​of penetration, pricing power, repurchase rate, and shelf appeal, two-factor analysis, regression models, or machine learning models can be used to quantify the contribution of each dimension to sales growth. In one embodiment, two-factor analysis is used for quantitative analysis. Two-factor analysis is a statistical analysis method used to quantify the contribution of multiple independent variables to the changes in the dependent variable. Its core idea is to decompose the total change in the dependent variable into the contribution of each factor's change by calculating the changes in each factor at different times while keeping other factors constant. Specifically, in this embodiment, the dependent variable is the sales growth of cigarette products between two periods, and the independent variables are the four product growth indicators: penetration, pricing power, repurchase rate, and shelf appeal. The specific process of quantifying the contribution rate of each product growth indicator based on the two-factor analysis method is as follows: First, obtain the sales data and indicator values ​​of penetration, price competitiveness, repurchase competitiveness, and extensibility corresponding to the current analysis period and the previous comparison period, and calculate the total sales growth and the change of each indicator between the two periods; then, for each product growth indicator, calculate the change in sales contributed by the indicator when the indicator changes alone (while the other three indicators remain unchanged), that is, the contribution value of the indicator; finally, divide the contribution value of the indicator by the total sales growth to obtain the contribution rate of the indicator to the historical sales growth.

[0091] For example, taking market penetration as an example, its contribution rate is calculated as follows: Market Penetration Contribution Rate = (((Gcurrent period - Gsame period) × Comprehensive Benchmark Value of Price Strength, Repeat Purchase Strength, and Expansion Strength in the Same Period) / ΔQ) × 100%; where ΔQ is the total sales growth between the two periods, Gcurrent period is the market penetration value in the current analysis period, Gsame period is the market penetration value in the previous comparison period, and the comprehensive benchmark value of price strength, repeat purchase strength, and expansion strength in the same period is the product or combined benchmark value of the three indicators in the previous comparison period. This formula shows that, assuming the other three indicators (price strength, repeat purchase strength, and expansion strength) remain unchanged in the same period, the contribution of market penetration to sales growth is calculated separately, and this contribution value is divided by the total sales growth to obtain the contribution rate of market penetration to sales growth. Through the above calculation method, the contribution rates of market penetration, price strength, repeat purchase strength, and expansion strength can be obtained respectively, thereby identifying the core and weak indicators driving sales growth over a period of time, providing a quantitative basis for subsequent target decomposition.

[0092] Then, based on the calculated contribution rates and the preset overall development target, the preset overall development target is decomposed into quantitative sub-targets of different dimensions, resulting in decomposed targets. The preset overall development target is the total sales growth or growth rate that cigarette products need to achieve within a certain period of time. Since the overall development target is a total indicator and cannot directly guide specific business operations, it needs to be decomposed into the dimensions of each product growth indicator to form executable and traceable quantitative sub-targets. First, based on the contribution rate of each product growth indicator to historical sales growth, the core indicators and weak indicators that drove sales growth in the past period are identified. Indicators with high contribution rates are the past growth engines, and indicators with low contribution rates are the past growth weaknesses. Then, combining the current actual performance of each indicator with the preset potential assessment, the expected contribution ratio of each indicator to the overall development target is determined, and indicators with high contribution rates are assigned a certain maintenance weight, while indicators with low contribution rates but large room for improvement are assigned a higher improvement weight. Finally, the preset overall development target is allocated to each product growth indicator according to the above expected contribution ratios, resulting in quantitative sub-targets for each indicator.

[0093] In one embodiment, the method for extracting the demand characteristics of cigarette products further includes: calculating a comprehensive evaluation score based on the characteristic parameters of cigarette product demand using a weighted summation or weighted geometric average method, and mapping the market status of cigarette products to five levels: "hot," "tight," "neutral," "loose," and "soft." Specifically, a comprehensive evaluation score range corresponding to each level is pre-defined. For example, a comprehensive evaluation score greater than or equal to 85 points is mapped to "hot," indicating that the market is in short supply and sales are booming; a score greater than or equal to 70 points and less than 85 points is mapped to "tight," indicating that the market is tight and demand exceeds supply; a score greater than or equal to 55 points and less than 70 points is mapped to "neutral," indicating that the market supply and demand are basically balanced; a score greater than or equal to 40 points and less than 55 points is mapped to "loose," indicating that the market is loose and supply slightly exceeds demand; and a score less than 40 points is mapped to "soft," indicating that the market is weak and sales are sluggish. The threshold values ​​for the above score ranges can be dynamically configured and adjusted according to the actual market characteristics of different regions and products to adapt to the status evaluation needs under different market environments.

[0094] At the same time, based on the comparison between the market status level of this cycle and the market status level of the previous cycle, four types of operational suggestions are generated: "improvement, early warning, maintenance, and opportunity". Specifically, if the market condition level in this cycle improves compared to the previous cycle (e.g., from "soft" to "loose," or from "loose" to "neutral"), an "improvement" suggestion is generated, indicating that the market condition is improving and the current strategy can continue. If the market condition level in this cycle declines compared to the previous cycle (e.g., from "neutral" to "loose," or from "tight" to "neutral"), an "early warning" suggestion is generated, indicating that the market condition is trending downward and attention should be paid to adjusting the strategy accordingly. If the market condition level in this cycle is the same as the previous cycle and is in a good state (e.g., "stable" or "tight"), a "maintain" suggestion is generated, indicating that the market condition is stable and the existing strategy can be maintained. If the market condition level in this cycle is the same as the previous cycle or slightly improves but is still at a low level (e.g., "loose" or "soft"), an "opportunity" suggestion is generated, indicating that there is room for improvement in the market and increased investment or marketing efforts can be considered to further improve the condition.

[0095] Through the aforementioned market status level mapping and operational suggestion generation mechanism, multi-dimensional market status evaluation indicators are quantified into intuitive market status labels and actionable suggestion levels, which avoids the problems of strong subjectivity and poor consistency caused by relying on human experience.

[0096] In one embodiment, a collaborative strategy for cigarette products is formulated based on the characteristic parameters of cigarette product demand and the analysis results of the decision model, including:

[0097] When the market status of a cigarette product is identified as belonging to a preset category requiring reduced supply, the supply of cigarette products that complement the cigarette product is increased based on a preset complementarity coefficient. When a cigarette product is in a phase of discontinuation or version switching, the average supply quantity adjustment value per household for related cigarette products that are substitutes for the cigarette product is calculated based on a preset transfer coefficient, and the average supply quantity per household for related cigarette products is modified according to the average supply quantity adjustment value per household. Combined display schemes, joint promotional activities, and collaborative supply plans for derivative products for multiple cigarette products are generated and managed.

[0098] A multi-product collaborative cultivation strategy can achieve synergistic cultivation among products. In this embodiment, the multi-product collaborative cultivation strategy includes a complementary placement strategy, a cluster promotion strategy, and a quantitative linkage strategy. Specifically, the complementary placement strategy is used to supplement the placement of other products when the placement of cigarette products needs to be reduced due to market conditions, in order to maintain the stability of the overall market structure and market share. Specifically, when it is identified that the market status of a certain cigarette product (product A) belongs to a preset category that requires a reduction in placement (such as "hot" or "tight"), the recommended reduction in placement of product A is automatically calculated as ΔA, and complementary products (product B) with the same price range, consumer group, or brand series are retrieved. Based on the complementary effect coefficient k obtained by training with historical sales data, the recommended supplementary placement of product B is calculated as ΔB = ΔA × k. The complementary effect coefficient can be dynamically learned and updated. The initial value is determined based on the transfer ratio in historical data and continuously optimized with feedback from actual implementation effects.

[0099] The cluster promotion strategy is used to achieve joint marketing and brand cluster promotion of multiple cigarette products, enhancing brand synergy. Based on a built-in multi-product combination marketing digital template library, it stores standardized marketing plans with different themes and product combinations. When it identifies a region requiring concentrated development of multiple products, it automatically matches and generates a corresponding cluster promotion plan from the multi-product combination marketing digital template library. For example, when it identifies a region requiring concentrated development of multiple high-end products, it can automatically generate a "Pure Enjoy High-End Tasting Event" plan. This plan includes recommended product combinations (such as cigarette specification 1, cigarette specification 2, and Sunshine Liqun), standardized display illustrations, promotional scripts, and a list of supporting materials, and is then distributed to the account manager's workbench for execution, achieving a synergistic amplification of brand image and a cluster effect.

[0100] Quantitative linkage strategies are used to coordinate the adjustment of the distribution quantity of related products during product switching or version iteration, thus smoothing sales fluctuations. Specifically, when a cigarette product is discontinued or undergoing a version switch (e.g., old version discontinued, new version launched), the system can predict the potential sales gap caused by the discontinuation of the product based on historical sales transfer models. Then, according to preset smoothing rules and transfer ratios, it intelligently calculates and adjusts the average distribution quantity per household for related products that are substitutes for the old product. For example, during product updates, after the old version is discontinued, some of its original market demand will shift to the new version or substitute product in the same series. Based on transfer coefficients trained from historical sales data, the system calculates the adjustment value for the average distribution quantity per household of related products and modifies the average distribution quantity per household accordingly, thereby ensuring a smooth market transition and avoiding significant sales fluctuations caused by product switching.

[0101] By adopting a multi-product collaborative development strategy, complementary allocation of distribution volume, brand cluster promotion, and coordinated adjustment of distribution volume are achieved among multiple cigarette products. This effectively avoids the internal competition and resource consumption problems that are prone to occur in the development of a single product, and fully leverages the cluster effect and synergistic value of the product portfolio.

[0102] In one embodiment, a collaborative strategy is transformed into specific instructions through a pre-defined strategy rule base, a combined cultivation task package is generated based on the specific instructions, and the combined cultivation task package is distributed to the cigarette product business system for execution, including:

[0103] Based on the characteristic parameters of cigarette product demand and the analysis results of the decision-making model, the system matches and sorts the data in a pre-set strategy rule base to generate a strategy package that includes specific measures, meets resource needs, and has expected effects. The strategy package is then converted into specific instructions, and a combined cultivation task package is generated based on these instructions. Finally, the combined cultivation task package is sent to the cigarette product business system for execution.

[0104] The pre-defined strategy rule base stores several placement strategy rules and marketing activity rules defined in a "condition-action-attribute" structure. The condition section can combine various factors such as market status level, product demand characteristic parameter range, decision model analysis results, product type, and market level; the action section defines specific strategy types (such as expanded placement, price stabilization and volume control, special promotional events, in-store product testing, and fixed pricing) and execution parameters; the attribute section includes strategy priority, expected results, and resource consumption estimates, used for subsequent conflict resolution and effect prediction.

[0105] After the strategy rule base is constructed, rule matching and output are performed. First, the system receives analysis results including characteristic parameters of cigarette product demand and the decision model. Then, using the input characteristic parameters and analysis results as matching conditions, multi-condition matching is performed in the strategy rule base to retrieve all matching strategy rules. Next, the matched strategy rules are ranked based on a comprehensive evaluation score. Weighted scores are applied based on attributes such as priority, expected effect, and resource consumption estimation. The rules are then ranked from highest to lowest based on the weighted comprehensive evaluation score, and conflict detection and resolution are performed on the ranking results (i.e., when multiple rules conflict, the rule with the highest comprehensive evaluation score is retained, and other conflicting rules are eliminated). One or more top-ranked strategy rules are combined into a strategy package. Finally, the task decomposition unit transforms the strategy package into a specific, executable sequence of instructions. This sequence includes instructions for adjusting distribution volume, terminal screening, and activity execution.

[0106] After the task decomposition unit completes the generation of the instruction sequence, the above specific instructions are packaged into a combined cultivation task package and distributed according to the task type and target system: for example, the delivery volume adjustment instruction is synchronously delivered and executed through the application programming interface; the terminal screening instruction and the activity execution instruction are pushed to the task list of the account manager's mobile workbench, along with the execution template and resource support list, for the account manager to execute and provide feedback on-site.

[0107] In addition, in one embodiment, the method for extracting the demand features of cigarette products further includes: monitoring the execution effect of the task package and obtaining feedback data on the execution effect; and optimizing the strategy rule base based on the feedback data.

[0108] The execution progress and effectiveness data of each task package are tracked in real time through the execution logs of a visualization platform or business system. This includes the actual execution rate of distribution volume adjustments, the achievement of terminal coverage, the participation rate and completion rate of marketing activities, as well as changes in market status indicators such as the price index, order fulfillment rate, and procurement area of ​​cigarette products after task execution, forming a feedback dataset of execution effectiveness. Furthermore, the feedback dataset is compared with the baseline data before the task package was issued to calculate the deviation between the actual and expected effects of each strategy. When the deviation exceeds a preset threshold, the optimization mechanism of the strategy rule base is automatically triggered, including adjusting the condition thresholds of rules, updating the priority ranking of rules, correcting the execution parameters of actions, and disabling rules with consistently poor matching effects.

[0109] Through the aforementioned monitoring and optimization mechanisms, the strategy rule base can be continuously optimized based on actual execution results, thereby improving the accuracy of strategy matching as operational data accumulates and enabling the evolutionary capability from static rule configuration to dynamic adaptive optimization.

[0110] In one embodiment, the cigarette product portfolio cultivation method further includes: unified display and scheduling management of end-to-end data, strategies, and execution results through a visualization platform. The visualization platform includes a global situation map, a product health dashboard, a strategy simulation sandbox, and a collaborative task operation dashboard. Specifically, the global situation map, based on a geographic information system, displays the distribution of market status levels at the provincial, municipal, and grid levels in the form of a heat map. Different colors represent different market statuses; for example, green represents "flat," yellow represents "tight" or "loose," and red represents "hot" or "soft." Users can click on the map to view specific areas and detailed indicator data for those areas.

[0111] The Product Health Cockpit focuses on a single cigarette product. It displays the current comprehensive evaluation score on a dashboard, shows the historical trends and changes in penetration, pricing power, repurchase rate, and market reach using trend charts, and compares the current values ​​of penetration, pricing power, repurchase rate, and market reach with the target values ​​using radar charts. It also displays the progress of each dimension's objectives, the results of competitor analysis, and Gantt charts and effect tracking curves of implemented strategies, which are used to track and manage the entire lifecycle of a single product's development.

[0112] The strategy simulation sandbox allows users to manually adjust hypothetical parameters through an interactive interface, such as "increasing the distribution of product A in region X by 10%" or "conducting a joint promotion of product B and product C in market Y." Based on a built-in market response model, it simulates and extrapolates changes in market conditions and sales forecasts over a future period after the strategy adjustment, and presents the simulation results in the form of visual charts to assist users in making important strategic decisions.

[0113] It also includes a collaborative task operation dashboard, which centrally displays a list of all automatically generated or manually initiated strategy tasks in a dashboard format. Each task clearly identifies the responsible person, deadline, current execution progress, and phased effect feedback, supporting task assignment, progress tracking, effect evaluation, and closed-loop management, and achieving transparent collaboration across departments and levels.

[0114] Through the aforementioned visualization platform, users can intuitively perceive the overall market situation, accurately identify product development shortcomings, scientifically predict the effects of strategy adjustments, and efficiently coordinate task execution, thereby improving the efficiency of end-to-end decision-making and the transparency of execution.

[0115] Figure 3 This is a schematic diagram of the data segmentation and parallel verification architecture of a cigarette product demand feature extraction method according to an embodiment of this application. Figure 3 As shown, this application interfaces with multiple data source systems at the data input layer, including a retailer direct reporting system, an order system, and a marketing platform, to collect real-time price reporting data, inventory reporting data, order data, and marketing activity data for cigarette products.

[0116] The collected multi-source data first enters the data partitioning and distribution engine. This engine can divide a large-scale dataset into multiple independent data blocks of similar size according to a preset data partitioning strategy. Preset data partitioning strategies include partitioning by time dimension (e.g., by day / week) or by geographical dimension (e.g., by city / grid). After the data blocks are partitioned, each data block is scheduled to multiple parallel computing nodes in the parallel computing cluster.

[0117] Each parallel computing node operates independently, performing outlier identification and removal using the Three Sigma principle on its assigned data block. For each metric data point within a data block, the computing node calculates the mean μ and standard deviation σ for that metric within that block. Data points greater than μ+3σ or less than μ-3σ are identified as outliers and removed from the data block. After processing their respective data blocks, each computing node outputs the outlier identification results for each block, including the data records identified as outliers and their positions within the original data block.

[0118] The outlier identification results from each computing node are integrated and processed by a merging and secondary verification layer. This layer first collects and merges the outlier identification results from different data blocks to remove duplicates. For each data record identified as an outlier, a secondary verification is performed across data block boundaries. Outlier records near the boundaries require re-evaluation based on statistics from adjacent blocks; while outlier records not near boundaries are directly identified as outliers. Finally, after removing the outlier data, a verified unified dataset is output for standardized data storage and subsequent indicator calculations.

[0119] Through the above data segmentation and multi-node parallel computing architecture, the overall data verification task is distributed across multiple computing nodes for simultaneous execution, thereby shortening the processing latency of outlier identification in large-scale datasets and providing an efficient and reliable data foundation.

[0120] Figure 4 This is a schematic diagram of the cigarette product linkage cultivation process of a cigarette product demand feature extraction method according to an embodiment of this application. Figure 4 As shown, the example is the coordinated development of the mature product "cigarette specification 1" and the high-end new product "cigarette specification 2".

[0121] In Phase One: Diagnostic Analysis - Problem Identification, the process begins by acquiring and validating multi-source data for both products, calculating standardized values ​​for various demand characteristics, and then using weighted fusion to obtain a comprehensive evaluation score. Based on this score, the market status of each product is mapped to five levels. For example... Figure 4 As shown, the market status of "Cigarette Specification 1" is "flat". After analyzing the contribution rates of its penetration, pricing power, repurchase rate, and expansion power, the decision-making model diagnoses its weakness as penetration, meaning that the breadth of terminal coverage is not yet saturated and there is significant room for improvement. The market status of "Cigarette Specification 2" is "loose". The decision-making model diagnoses its weaknesses as pricing power and repurchase rate, meaning that price stability and terminal repeat purchase willingness are the main factors restricting its growth.

[0122] In Phase Two: Strategy Generation and Execution, the intelligent strategy engine automatically matches linkage rules and generates linkage strategy packages based on the analysis results of the aforementioned state levels and decision-making models. These strategy packages include complementary placement strategies, cluster promotion strategies, and quantitative linkage strategies. Regarding Strategy 1: Complementary Placement Strategy, given the relatively weak market performance of "Cigarette Specification 2," its placement pace is tightened, while the coverage and placement volume of "Cigarette Specification 1" are expanded, creating a market gradient support between the two products. Regarding Strategy 2: Cluster Promotion Strategy, "Dual-Product Tasting Events" are uniformly conducted at core high-end terminals, with themed displays to strengthen the high-end brand cluster image. Regarding Strategy 3: Quantitative Linkage Strategy, during the transition from the old version to the new version of "Cigarette Specification 2," some of the old version's quantitative plans are smoothly transferred to mature products in the same series based on historical transfer coefficients, ensuring a smooth transition in overall sales.

[0123] In Phase Three: Effect Monitoring and Optimization, the execution effect is monitored and tracked in real time through a visual decision-making platform. After the strategy cycle ends, feedback data on the execution effect is collected. The data shows that the terminal penetration rate of "Cigarette Specification 1" has increased significantly, and its weakest dimension of penetration has been effectively improved. The market price of "Cigarette Specification 2" has stabilized, and the repurchase rate at terminals has increased significantly. Overall, the target effect indicates that the overall market share of high-end products in the target region has steadily increased. The above feedback data is input into the data collection engine. Based on the deviation between the actual and expected effects, the complementary coefficient in the complementary delivery strategy is automatically optimized, and the weight of the strategy matching rules in the intelligent strategy engine is automatically adjusted. This enables the model to learn itself and iterate its parameters, providing optimized rules for the next round of strategy generation. This verifies the effectiveness and scalability of the proposed method in the actual market environment.

[0124] Figure 5 This is a flowchart illustrating the incremental update and sliding window mechanism of a cigarette product demand feature extraction method according to an embodiment of this application. Figure 5 As shown, this illustrates the incremental update calculation process of cigarette product demand characteristic values ​​in the context of new data, as well as the sliding window management mechanism when recalculating periodic indicators.

[0125] Specifically, during the startup phase, the system first calculates metrics on all historical data and initializes cached records for each metric. Each cached record contains three elements: cached value, timestamp, and version number. After startup, the system enters the incremental calculation loop phase. It continuously monitors for changes in multi-source data. When a data change (including addition, update, or deletion) is detected, the adaptation layer identifies the changed record and determines the change type and scope of impact. Metric calculations are performed only on the changed data to obtain the marginal change ΔV. Subsequently, the marginal change is weighted and merged with the historical metric values ​​in the cache according to time weights. After weighted merging, the merged result is stored as the updated metric value, the version number of the cached record is incremented synchronously, and the current timestamp is recorded as the latest timestamp. If no data change is detected, the current cache state remains unchanged.

[0126] For metrics requiring periodic recalculation, a sliding window mechanism is implemented. While in cached state, it checks whether the sliding window has been triggered, i.e., whether a new period boundary has been reached. When the sliding window moves forward, data from the most recent period is discarded. Subsequently, the metric baseline value is recalculated based on the most recent N periods of data retained within the window, the cache baseline value is updated, and historical data exceeding the window range is automatically transferred to offline archive storage. This sliding window mechanism ensures a constant online storage volume (retaining only the most recent N periods of data), effectively controlling the unlimited growth of online data storage. If no periodic recalculation is detected, the current cache state remains unchanged.

[0127] It also provides a real-time indicator query application programming interface (API). Whenever the cached value, timestamp, and version number are updated, the latest indicator value and corresponding timestamp and version number in the current cache can be obtained through the API, which can be used for subsequent market status evaluation and collaborative strategy formulation.

[0128] Figure 6 This is a flowchart of a method for extracting the demand features of cigarette products according to another embodiment of this application. Figure 6 As shown, the method for extracting the demand characteristics of this cigarette product includes the following steps:

[0129] Step S601: Obtain multi-source data on the market status of cigarette products.

[0130] Step S602: Divide the multi-source data into multiple data blocks, and identify and remove outliers in each data block using the three Sigma principle.

[0131] Step S603: Cross-validate and merge the outlier identification results of each data block to obtain a unified dataset.

[0132] Step S604: Calculate the characteristic values ​​of cigarette product demand based on the unified dataset, and unify the dimensions of the characteristic values ​​of cigarette product demand through the extreme value standardization method to obtain the characteristic parameters of cigarette product demand.

[0133] Step S605: The indicator caching calculation method based on the incremental update mechanism is used to calculate the characteristic values ​​of the demand for time-sensitive cigarette products.

[0134] Step S606: For indicators that need to be recalculated periodically, a sliding window mechanism is used for indicator calculation and data management.

[0135] Step S607: Based on the feature parameters, a comprehensive evaluation score is obtained through weighted fusion. The market status of cigarette products is mapped to different market status levels based on the comprehensive evaluation score, and different operational suggestions are generated.

[0136] Step S608: Construct a decision-making model to conduct quantitative analysis of cigarette products across four dimensions: market penetration, price competitiveness, repurchase rate, and market reach.

[0137] Step S609: Based on the two-factor analysis method, quantify the contribution of penetration power, price power, repurchase power and extension power to the historical sales growth of cigarette products.

[0138] Step S610: Based on the contribution rate and the preset overall cultivation goal, the preset overall cultivation goal is decomposed into quantitative sub-goals of each dimension to obtain the decomposed goals. The contribution rate and the decomposed goals are used as the analysis results of the decision model.

[0139] Step S611: Based on the analysis results of the feature parameters and the decision model, match and sort them in the preset strategy rule base to generate a strategy package.

[0140] Step S612: Convert the strategy package into specific instructions, generate a combined cultivation task package based on the specific instructions, and send the combined cultivation task package to the cigarette product business system for execution.

[0141] Step S613: Monitor the execution effect of the task package, obtain feedback data on the execution effect, and optimize the strategy rule base based on the feedback data.

[0142] Step S614: The entire data, strategies, and execution results are uniformly displayed and managed through a visualization platform.

[0143] Steps S601 to S614 above, through data block parallel verification, incremental update caching and sliding window archiving mechanism, combined with decision model and multi-product linkage strategy, realize efficient processing of large-scale multi-source data and collaborative cultivation of cigarette product portfolio, effectively improving data processing efficiency and cross-regional collaborative cultivation effect.

[0144] Figure 7 This is a structural block diagram of a cigarette product demand feature extraction device 70 according to an embodiment of this application, as shown below. Figure 7 As shown, the cigarette product demand feature extraction device 70 includes: a data unification module 72 and a scoring calculation module 74; wherein: the data unification module 72 is used to acquire multi-source data on the market status of cigarette products; divide the multi-source data into multiple data blocks, and perform outlier identification and removal on the data blocks using the three sigma principle to obtain outlier identification results; perform cross-validation and merging on the outlier identification results to determine the abnormal data in the data blocks, and remove the abnormal data from the original data to obtain a unified dataset; the scoring calculation module 74 is used to calculate the feature values ​​of cigarette product demand based on the unified dataset, and perform dimension unification on the feature values ​​of cigarette product demand using the extreme value standardization method to obtain the feature parameters of cigarette product demand.

[0145] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can reside in the same processor; or the above modules can be located in different processors in any combination.

[0146] This embodiment also provides an electronic device including a memory and a processor, the memory storing a computer program and the processor being configured to run the computer program to perform the steps in any of the above method embodiments.

[0147] Optionally, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor and the input / output device is connected to the processor.

[0148] Optionally, in this embodiment, the processor can be configured to perform the following steps via a computer program:

[0149] S1, acquire multi-source data on the market status of cigarette products;

[0150] S2 divides the multi-source data into multiple data blocks and uses the three sigma principle to identify and remove outliers from the data blocks, thus obtaining the outlier identification results.

[0151] S3, cross-validate and merge the outlier identification results to identify outlier data in the data block and remove the outlier data from the multi-source data to obtain a unified dataset;

[0152] S4. Calculate the characteristic values ​​of cigarette product demand based on the unified dataset, and unify the dimensions of the characteristic values ​​of cigarette product demand through the extreme value standardization method to obtain the characteristic parameters of cigarette product demand.

[0153] It should be noted that the specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementations, and will not be repeated in this embodiment.

[0154] Furthermore, in conjunction with the cigarette product demand feature extraction method provided in the above embodiments, this embodiment can also provide a storage medium for implementation. This storage medium stores a computer program; when executed by a processor, the computer program implements any of the cigarette product demand feature extraction methods described in the above embodiments.

[0155] It should be understood that the specific embodiments described herein are merely illustrative of the application and not intended to limit it. All other embodiments derived by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application.

[0156] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0157] Obviously, the accompanying drawings are merely some examples or embodiments of this application. Those skilled in the art can apply this application to other similar situations based on these drawings without any creative effort. Furthermore, it is understood that although the work done in this development process may be complex and lengthy, for those skilled in the art, certain design, manufacturing, or production modifications made based on the technical content disclosed in this application are merely conventional technical means and should not be considered as insufficient disclosure of this application.

[0158] The term "embodiment" in this application refers to a specific feature, structure, or characteristic described in connection with an embodiment that may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily imply the same embodiment, nor does it imply that it is mutually exclusive with or independent of other embodiments. It will be clearly or implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0159] The above embodiments merely illustrate several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of patent protection. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the appended claims.

Claims

1. A method for extracting demand characteristics of cigarette products, characterized in that, include: Obtain multi-source data on the market status of cigarette products; The multi-source data is divided into multiple data blocks, and outlier identification and removal are performed on the data blocks using the three Sigma principle to obtain outlier identification results. The outlier identification results are cross-validated and merged to determine the outlier data in the data block, and the outlier data is removed from the multi-source data to obtain a unified dataset; The characteristic values ​​of cigarette product demand are calculated based on the unified dataset, and the dimensions of the characteristic values ​​of cigarette product demand are unified by the extreme value standardization method to obtain the characteristic parameters of cigarette product demand.

2. The method for extracting demand characteristics of cigarette products according to claim 1, characterized in that, The step of calculating the feature values ​​of cigarette product demand based on the unified dataset, and unifying the dimensions of the feature values ​​of cigarette product demand through extreme value standardization to obtain the feature parameters of cigarette product demand, includes: calculating the feature values ​​of time-sensitive cigarette product demand using incremental updates; wherein: Store the cached values, timestamps, and version numbers of the characteristic values ​​required for each cigarette product; When new data arrives in the unified dataset, the incremental indicator value is calculated from the new data. The incremental indicator value and the cached value are weighted and merged according to time weight to obtain the updated cached value; The characteristic parameters of the cigarette product requirements are calculated and updated based on the updated cache value.

3. The method for extracting demand characteristics of cigarette products according to claim 1, characterized in that, The method for extracting demand characteristics of cigarette products also includes: A decision-making model is constructed to conduct quantitative analysis of cigarette products from different dimensions, calculate the contribution rate of historical sales growth of cigarette products from different dimensions, and obtain decomposed targets for different dimensions based on the contribution rate and the preset overall cultivation target. The contribution rate and the decomposed targets are used as the analysis results of the decision-making model. The decision-making model is used to conduct quantitative analysis of the contribution rate of cigarette products from different dimensions and generate quantitative analysis models of decomposed targets for each dimension. Based on the characteristic parameters of the demand for the cigarette products and the analysis results of the decision model, a collaborative strategy for the cigarette products is formulated. The collaborative strategy is transformed into specific instructions through a preset strategy rule base, a combined cultivation task package is generated according to the specific instructions, and the combined cultivation task package is executed according to the corresponding cigarette product.

4. The method for extracting demand characteristics of cigarette products according to claim 3, characterized in that, The aforementioned decision-making model involves conducting quantitative analysis of cigarette products across different dimensions, calculating the contribution rate of historical sales growth for cigarette products in different dimensions, and obtaining decomposed targets for different dimensions based on the contribution rate and a preset overall development target. The contribution rate and the decomposed targets are then used as the analysis results of the decision-making model, including: Based on two-factor analysis, the contribution rate of product growth indicators to the historical sales growth of the cigarette products was quantified. Based on the contribution rate and the preset overall cultivation target, the preset overall cultivation target is decomposed into quantitative sub-targets of different dimensions to obtain the decomposed target.

5. The method for extracting demand characteristics of cigarette products according to claim 3, characterized in that, The step of formulating a collaborative strategy for the cigarette product based on the characteristic parameters of the cigarette product demand and the analysis results of the decision model includes: When it is identified that the market status of a cigarette product belongs to a preset category that requires a reduction in supply, the supply of cigarette products that have a complementary relationship with the cigarette product is increased according to a preset complementarity coefficient. When the cigarette product is in the discontinuation or version switching stage, the average household distribution amount adjustment value of related cigarette products that have a substitution relationship with the cigarette product is calculated according to the preset transfer coefficient, and the average household distribution amount of the related cigarette products is modified according to the average household distribution amount adjustment value. Generate and manage combined display schemes, joint promotional activities, and collaborative launch plans for multiple cigarette products.

6. The method for extracting demand characteristics of cigarette products according to claim 3, characterized in that, The process of converting the collaborative strategy into specific instructions through a preset strategy rule base, generating a combined cultivation task package based on the specific instructions, and then distributing the combined cultivation task package to the cigarette product business system for execution includes: Based on the characteristic parameters of the cigarette product demand and the analysis results of the decision model, the system matches and sorts the data in the preset strategy rule base to generate a strategy package that includes specific measures, meets resource needs, and has expected effects. The strategy package is converted into specific instructions, a combined cultivation task package is generated based on the specific instructions, and the combined cultivation task package is sent to the cigarette product business system for execution.

7. The method for extracting demand characteristics of cigarette products according to claim 6, characterized in that, The method for extracting demand characteristics of cigarette products also includes: Monitor the execution effect of the task package and obtain feedback data on the execution effect; The policy rule base is optimized based on the feedback data.

8. A device for extracting demand characteristics of cigarette products, characterized in that, include: The data unification module and the scoring calculation module; among which: The data unification module is used to acquire multi-source data on the market status of cigarette products; divide the multi-source data into multiple data blocks, and perform outlier identification and removal on the data blocks using the three Sigma principle to obtain outlier identification results; perform cross-validation and merging on the outlier identification results to determine the abnormal data in the data blocks, and remove the abnormal data from the original data to obtain a unified dataset; The scoring calculation module is used to calculate the feature values ​​of cigarette product demand based on the unified dataset, and to unify the dimensions of the feature values ​​of cigarette product demand through the extreme value standardization method to obtain the feature parameters of cigarette product demand.

9. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to run the computer program to perform the cigarette product demand feature extraction method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the cigarette product demand feature extraction method according to any one of claims 1 to 7.

Images