Cigarette putting area fitness evaluation method, device and system and computer readable storage medium

Abstract

The invention discloses a cigarette putting area fitness evaluation method, device and system and a computer readable storage medium, and relates to the technical field of deep learning. The method comprises the following steps: acquiring cigarette sales data, social and economic indicators and retailer inventory data of a target area; constructing a multi-dimensional spatial-temporal feature sequence based on the historical data; the sequence is input to a trained regional fitness evaluation model, the model is a multilayer LSTM network based on attention mechanism enhancement, long-period dependence features are extracted through a gating loop unit, and regional consumption preference features and macropolicy impact factors are optimized in a combined mode for training; generating a fitness score and a recommended delivery amount in a future time period according to the model output; and finally, outputting a regionalized delivery strategy scheme. According to the method, the problems that the market change cannot be dynamically predicted and the regional difference is neglected in the existing method are solved, and the prediction precision and the refinement level of decision support are improved.

Description

Technical Field

[0001] This invention relates to the field of deep learning technology, and in particular to a method, apparatus, system, and computer-readable storage medium for evaluating the fitness of cigarette placement areas. Background Technology

[0002] In a mature and increasingly competitive cigarette market, a company's ability to accurately grasp the needs and characteristics of consumers in different regions has become a key factor in enhancing its market competitiveness. Regional research on cigarette distribution not only helps companies achieve more refined market segmentation and design targeted marketing strategies, but also plays a crucial role in optimizing the allocation of production, inventory, and logistics resources and improving overall operational efficiency. Through in-depth insights into regional markets, companies can respond more flexibly to external market fluctuations and adjust their product distribution layout in a timely manner, thereby continuously enhancing the overall competitiveness of their products and services.

[0003] However, existing methods for assessing the regional adaptability of cigarette distribution still primarily rely on historical static data, using limited dimensions such as sales revenue and the HHI index for market classification and trend projection. They lack in-depth analysis of multi-source dynamic indicators and spatiotemporal dependencies. These methods struggle to effectively capture the complex market dynamics formed by the interplay of cultural customs, consumer behavior, and economic fluctuations in different regions. This results in limited predictive power and an inability to provide enterprises with timely and accurate regional adaptability assessments and decision support.

[0004] Therefore, this application proposes a method, device, system, and computer-readable storage medium for assessing the adaptability of cigarette placement areas, which can dynamically and accurately predict the adaptability of cigarette placement areas, thereby improving enterprises' responsiveness to market changes and decision support levels. Summary of the Invention

[0005] The main objective of this invention is to provide a method, apparatus, system, and computer-readable storage medium for assessing the suitability of cigarette distribution areas, aiming to solve the technical problem that existing methods rely on historical static data and cannot effectively predict future dynamic changes in the market.

[0006] To achieve the above objectives, the present invention provides a method for evaluating the adaptability of cigarette distribution areas, comprising the following steps:

[0007] Acquire historical data for the target area, including cigarette sales data, socioeconomic indicators, and retailer inventory data;

[0008] Based on the historical data, a multi-dimensional spatiotemporal feature sequence is constructed and used to characterize the dynamic adaptability of the cigarette market as time and space change.

[0009] The spatiotemporal feature sequence is input into a trained regional fitness evaluation model, which extracts long-period dependency features from the spatiotemporal feature sequence through a gated recurrent unit; wherein, the regional fitness evaluation model is a multi-layer LSTM network enhanced by an attention mechanism, which is trained by jointly optimizing regional consumption preference features and macroeconomic policy influencing factors;

[0010] Based on the long-term dependency characteristics, a cigarette placement fitness score and recommended placement amount for the target area in the future time period are generated.

[0011] Based on the cigarette distribution adaptability score and recommended distribution amount, a regionalized cigarette distribution strategy is output.

[0012] As a further improvement to this application, the construction of the multi-dimensional spatiotemporal feature sequence includes the following steps:

[0013] The raw historical data is cleaned, denoised, and standardized preprocessed.

[0014] Key features are extracted from the preprocessed data, including: sales revenue of each brand of cigarettes, Herfindahl-Hirschman Index (HHI), regional annual GDP, social inventory, and time-series data on retail order demand.

[0015] As a further improvement to this application, the single-layer computation process of the attention-enhanced multilayer LSTM network includes:

[0016] Calculate the input gate i t Forgotten Gate t Output gate o t and candidate memory g t The calculation expression is:

[0017] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0018] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0019] o t =σ(W o ·[h t-1 ,x t ]+b o )

[0020]

[0021] Update the current state of the memory cell c t and hidden state h t The calculation expression is:

[0022] c t =f t ·c t-1 +i t ·g t

[0023] h t =o t ·tanh(c t )

[0024] Where, x t h is the input feature vector at time t; t-1 It is the hidden state at time t-1; c t-1 It is the state of memory cells at time t-1; W i W f W o W g b is the weight matrix; i b f b o b g σ is the bias vector; σ is the sigmoid activation function.

[0025] As a further improvement to this application, the operation of the attention mechanism includes:

[0026] Calculate the attention weight distribution of the hidden states at all time steps of the top layer of the multilayer LSTM network;

[0027] The hidden states at all time steps are weighted and summed according to the attention weights to obtain the context vector;

[0028] The context vector is concatenated with the hidden state at the final time step to form a comprehensive feature representation for generating the fitness score and recommendation delivery volume.

[0029] As a further improvement to this application, the joint optimization of regional consumption preference characteristics and macroeconomic policy influencing factors for training refers to:

[0030] The regional consumption preference characteristics are quantified by the historical brand sales share and volatility.

[0031] The macroeconomic policy impact factors are used as an additional input channel in model training or as a regularization term in a custom loss function to guide model training.

[0032] As a further improvement to this application, the method employs gradient clipping techniques during model training to prevent gradient explosion.

[0033] As a further improvement to this application, the regionalized cigarette delivery strategy includes the following steps:

[0034] The fitness score is compared with a preset threshold to classify the target area into priority delivery, regular delivery, observation delivery or restricted delivery categories.

[0035] Differentiated cigarette distribution strategies are generated for different types of regions, taking into account their recommended distribution volumes.

[0036] Furthermore, to achieve the above objectives, the present invention also provides a cigarette placement area adaptability assessment device, comprising:

[0037] The data acquisition module is used to acquire historical data for the target area;

[0038] The preprocessing module is used to clean, denoise, and standardize the historical data.

[0039] The feature extraction module is used to extract multi-dimensional spatiotemporal features from the preprocessed data to construct a feature sequence. The features include at least the sales revenue of each brand of cigarettes, the HHI index, and the regional annual GDP.

[0040] The model building module is used to build a multi-layer LSTM network based on an attention mechanism enhancement. The network includes gated recurrent units for extracting long-period dependencies in the input feature sequence.

[0041] The training module is used to train the LSTM network using the feature sequences. During the training process, it jointly optimizes regional consumption preference features and macroeconomic policy influencing factors, and uses gradient pruning to prevent gradient explosion.

[0042] The prediction module is used to predict the fitness of cigarette distribution in the target area in the future time period using the trained model, and generate fitness scores and recommended distribution amounts.

[0043] The decision output module is used to output a regionalized cigarette distribution strategy based on the score and distribution volume.

[0044] The storage module is used to store model parameters, intermediate data, and prediction results.

[0045] In addition, to achieve the above objectives, the present invention also provides a cigarette placement area fitness assessment system, comprising: a memory, a processor, and a cigarette placement area fitness assessment program stored in the memory and executable on the processor, wherein the cigarette placement area fitness assessment program, when executed by the processor, implements the steps of the cigarette placement area fitness assessment method as described above.

[0046] Furthermore, to achieve the above objectives, the present invention also provides a computer-readable storage medium, characterized in that the computer-readable storage medium stores a cigarette placement area fitness evaluation program, which, when executed by a processor, implements the steps of the cigarette placement area fitness evaluation method as described above.

[0047] The technical solution provided by this invention may include the following beneficial effects:

[0048] 1. In the process of use, this invention introduces a multi-layer LSTM network based on attention mechanism enhancement, which can effectively capture long-term dependencies and complex dynamic patterns in multi-dimensional spatiotemporal feature sequences such as sales data, socioeconomic indicators and retail inventory. It overcomes the limitation of traditional methods that rely only on historical static data and cannot effectively predict future dynamic changes in the market, and improves the accuracy of predicting the adaptability of cigarette distribution areas and the timeliness of responding to market fluctuations.

[0049] 2. In the process of using this invention, by jointly optimizing regional consumption preference characteristics and macroeconomic policy influencing factors during model training, the evaluation model can deeply integrate the complex market characteristics formed by the intertwining of multiple factors such as cultural customs, consumption behavior and economic fluctuations in different regions. This enhances the model's ability to quantitatively characterize regional differences and market complexity and to judge its adaptability, thus solving the problem that traditional methods do not adequately consider regional differences and are unable to fully reflect market complexity.

[0050] 3. During use, this invention ultimately outputs a quantifiable fitness score and recommended distribution volume, and further generates differentiated regional distribution strategy plans based on this, providing enterprises with clear and operable decision-making basis. It overcomes the shortcomings of existing methods that mostly focus on macro trend analysis and lack precise quantitative support, and improves the scientificity, refinement and operability of cigarette distribution strategy formulation. Attached Figure Description

[0051] The above and other objects, features and advantages of the present invention will become more apparent from the more detailed description of exemplary embodiments of the invention in conjunction with the accompanying drawings, wherein the same reference numerals generally represent the same components in the exemplary embodiments of the invention.

[0052] Figure 1This is a schematic diagram of the hardware operating environment of the cigarette placement area adaptability assessment system according to an embodiment of the present invention.

[0053] Figure 2 This is a flowchart illustrating the first embodiment of the cigarette placement area adaptability assessment method of the present invention.

[0054] Figure 3 This is a network structure diagram of the second embodiment of the cigarette placement area adaptability assessment method of the present invention;

[0055] Figure 4 This is a model training and validation loss curve diagram of the second embodiment of the cigarette placement area fitness assessment method of the present invention;

[0056] Figure 5 This is a map showing the adaptability scores of various prefectures in the second embodiment of the cigarette distribution area adaptability assessment method of the present invention;

[0057] Figure 6 This is a comparison chart of recommended cigarette placement amounts in the second embodiment of the cigarette placement area adaptability assessment method of the present invention;

[0058] Figure 7 This is an attention weighting diagram of the second embodiment of the cigarette placement area adaptability assessment method of the present invention;

[0059] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0060] To better understand the technical solutions of the application, exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0061] As one implementation scheme, Figure 1 This is a schematic diagram of the hardware operating environment of the cigarette placement area adaptability assessment system involved in the embodiments of the present invention.

[0062] like Figure 1As shown, the cigarette distribution area suitability assessment system may include: a processor 1001, such as a CPU; a memory 1005; a user interface 1003; a network interface 1004; and a communication bus 1002. The communication bus 1002 is used to establish communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be a high-speed RAM or a stable, non-volatile memory, such as a disk drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0063] Those skilled in the art will understand that Figure 1 The cigarette placement area fitness assessment system architecture shown in the figure does not constitute a limitation on the cigarette placement area fitness assessment system. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0064] like Figure 1 As shown, the memory 1005, which serves as a storage medium, may include an operating system, a network communication module, a user interface module, and a cigarette placement area fitness assessment program. The operating system is a program that manages and controls the hardware and software resources of the cigarette placement area fitness assessment system, as well as the operation of the cigarette placement area fitness assessment program and other software or programs.

[0065] exist Figure 1 In the cigarette placement area fitness assessment system shown, the user interface 1003 is mainly used to connect to the terminal and communicate with the terminal; the network interface 1004 is mainly used to communicate with the backend server; and the processor 1001 can be used to call the cigarette placement area fitness assessment program stored in the memory 1005.

[0066] In this embodiment, the cigarette placement area adaptability assessment system includes: a memory 1005, a processor 1001, and a cigarette placement area adaptability assessment program stored in the memory and executable on the processor, wherein:

[0067] When processor 1001 calls the cigarette placement area fitness evaluation program stored in memory 1005, it performs the following operations:

[0068] Step S20: Obtain historical data for the target area, including cigarette sales data, socioeconomic indicators, and retailer inventory data;

[0069] Step S20: Construct a multi-dimensional spatiotemporal feature sequence based on the historical data, and use it to characterize the dynamic adaptability of the cigarette market as time and space change;

[0070] Step S30: Input the spatiotemporal feature sequence into the trained regional fitness evaluation model. The regional fitness evaluation model extracts long-period dependency features from the spatiotemporal feature sequence through gated recurrent units. The regional fitness evaluation model is a multi-layer LSTM network enhanced by attention mechanism, which is trained by jointly optimizing regional consumption preference features and macroeconomic policy influencing factors.

[0071] Step S40: Based on the long-term dependency characteristics, generate a cigarette placement fitness score and recommended placement amount for the target area in the future time period;

[0072] Step S50: Based on the cigarette distribution adaptability score and recommended distribution amount, output a regionalized cigarette distribution strategy.

[0073] Based on the hardware architecture of the cigarette placement area fitness assessment system using the aforementioned cigarette placement area fitness assessment technology, an embodiment of the cigarette placement area fitness assessment method of the present invention is proposed.

[0074] Reference Figure 2 In the first embodiment, the method for assessing the adaptability of cigarette distribution areas includes the following steps:

[0075] Step S10: Obtain historical data for the target area, including cigarette sales data, socioeconomic indicators, and retailer inventory data;

[0076] In this embodiment, after receiving a query command from a user for a target area, the cigarette delivery area adaptability assessment system collects historical data of the target area for subsequent analysis.

[0077] Historical data refers to multi-dimensional time-series information related to the cigarette market in the target region over a period of time, which is the basis for the model to extract spatiotemporal features and predict fitness.

[0078] Optionally, historical data includes, but is not limited to: sales volume, sales revenue, and market share of each brand of cigarettes; socioeconomic indicators such as the region's annual GDP, population, and per capita disposable income; and real-time dynamic supply chain data such as retailer inventory levels, turnover rate, and order fulfillment rate.

[0079] Step S20: Construct a multi-dimensional spatiotemporal feature sequence based on the historical data, and use it to characterize the dynamic adaptability of the cigarette market as time and space change;

[0080] In this embodiment, the system fuses and structures the collected multi-source historical data to construct a feature sequence that combines temporal continuity with spatial differences. This sequence aims to quantitatively characterize the responsiveness and adaptability of different regional markets to cigarette distribution strategies from a dynamic evolution perspective.

[0081] For example, constructing a multi-dimensional spatiotemporal feature sequence includes the following steps:

[0082] The raw historical data is cleaned, denoised, and standardized preprocessed.

[0083] Extract key features from the preprocessed data.

[0084] Optionally, key features include, but are not limited to: sales revenue of each brand of cigarettes, Herfindahl-Hirschman Index (HHI), regional annual GDP, social inventory, and time-series data on retail order demand.

[0085] Step S30: Input the spatiotemporal feature sequence into the trained regional fitness evaluation model. The regional fitness evaluation model extracts long-period dependency features from the spatiotemporal feature sequence through gated recurrent units. The regional fitness evaluation model is a multi-layer LSTM network enhanced by attention mechanism, which is trained by jointly optimizing regional consumption preference features and macroeconomic policy influencing factors.

[0086] In this embodiment, the system inputs the constructed multi-dimensional spatiotemporal feature sequence into a pre-trained regional fitness evaluation model. The core of this model is a multi-layer long short-term memory network (LSTM) enhanced with attention mechanism, which extracts long-period dependency features from the input sequence through a gated recurrent unit structure to capture the complex temporal patterns in the dynamic changes of the cigarette market.

[0087] For example, the training process of the region fitness evaluation model includes the following steps:

[0088] First, the constructed multi-dimensional spatiotemporal feature sequence dataset is divided into training set, validation set and test set in chronological order. The training set is used for model parameter learning, the validation set is used for hyperparameter tuning and early stopping, and the test set is used to finally evaluate the model's generalization performance.

[0089] Model training is based on the aforementioned LSTM forward computation process, with network parameters optimized through backpropagation. During training, the system enhances the model's business adaptability by jointly optimizing regional consumption preference features and macroeconomic policy impact factors. Regional consumption preferences, quantified using historical brand sales share and volatility, are used as part of the feature input; macroeconomic policy impact is used as a regularization term in a custom loss function, guiding the model's learning direction to align with policy guidance. The loss function can be expressed as:

[0090] Total loss = Prediction error loss + λ × Policy compliance regularization term

[0091] Here, λ is a hyperparameter used to balance forecast accuracy with policy compliance requirements.

[0092] Training employs stochastic gradient descent or its variants (such as the Adam optimizer) for parameter updates, and gradient pruning is used to prevent gradient explosion. The system monitors performance on the validation set and triggers an early stopping mechanism when performance no longer improves to avoid overfitting. Ultimately, the system saves the optimal model parameters from the validation set.

[0093] After the model is trained, its predictive performance is evaluated using a test set to ensure that the model can maintain high predictive accuracy and stability on unknown data before it can be put into practical application.

[0094] The following is a detailed description of the model's forward computation process:

[0095] The forward computation process of each LSTM unit in the region fitness evaluation model includes:

[0096] Calculate the input gate i t Forgotten Gate t Output gate o t and candidate memory g t The calculation expression is:

[0097] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0098] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0099] o t =v(W o ·[h t-1 ,x t ]+b o )

[0100] g t =tanh(W g ·[h t-1 ,x t ]+b g )

[0101] Update the current state of the memory cell c t and hidden state ht The calculation expression is:

[0102] c t =f t ·c t-1 +i t ·g t

[0103] h t =o t ·tanh(c t )

[0104] Where, x t h is the input feature vector at time t; t-1 It is the hidden state at time t-1; c t-1 It is the state of memory cells at time t-1; W i W f W o W g b is the weight matrix; i b f b o b g σ is the bias vector; σ is the sigmoid activation function.

[0105] Furthermore, the model enhances its ability to focus on features at key time steps through an attention mechanism. This mechanism calculates the attention weights of all hidden states in the top-level LSTM network, thereby generating a context vector. This context vector is then compared with the final state h. t The combined features are then used for final prediction.

[0106] Step S40: Based on the long-term dependency characteristics, generate a cigarette placement fitness score and recommended placement amount for the target area in the future time period;

[0107] In this embodiment, the system generates a cigarette distribution fitness score and recommended distribution amount for the target area in the future time period based on the long-term dependency features extracted by the regional fitness assessment model. This process utilizes a trained model to decode and map the comprehensive feature representation enhanced by an attention mechanism, outputting quantitative indicators that can be used for decision support.

[0108] For example, the final layer of the model maps high-level features to predicted values ​​through a fully connected network and a linear output unit. The fitness score S is normalized to the [0,1] interval using the Sigmoid function, and the calculation formula is as follows:

[0109] S=σ(W s ·h fused +b s )

[0110] Among them, hfused is the comprehensive feature vector obtained after being weighted by attention and concatenated with the final hidden state, where W and b s are the weight and bias parameters of the output layer. The higher the score, the better the market fitness of the area and the more suitable for cigarette placement.

[0111] The recommended placement volume V is obtained through linear activation function regression, and its calculation is expressed as:

[0112] V = W v ·h fused + b v

[0113] To improve the rationality of the output, the system can post-process V by combining the historical sales volume range, such as truncating or scaling, to ensure that the recommended volume is within the actual executable range.

[0114] Furthermore, the system performs regional classification and strategy generation based on the fitness score SS. Set the threshold intervals T1 < T2 < T3, for example, T1 = 0.3, T2 = 0.6, T3 = 0.8, and divide the regions into the following categories:

[0115] When S ≥ T3, it is marked as "priority placement area", and it is recommended to significantly increase the placement volume and allocate resources preferentially;

[0116] When T2 ≤ S < T3, it is marked as "regular placement area", and the placement is maintained or slightly increased according to the historical level;

[0117] When T1 ≤ S < T2, it is marked as "observation placement area", and it is recommended to carefully control the placement and strengthen market monitoring;

[0118] When S < T1, it is marked as "restricted placement area", and it is recommended to reduce the placement volume or suspend the placement and conduct market research.

[0119] Finally, the system outputs a structured strategy plan, including the classification results of each region, the adapted recommended placement volume, and the corresponding strategy suggestions, providing clear and operable quantitative basis for cigarette placement decisions.

[0120] Step S50: Based on the cigarette placement fitness score and the recommended placement volume, output a regionalized cigarette placement strategy plan.

[0121] In this embodiment, the system automatically generates a regionalized cigarette placement strategy plan based on the cigarette placement fitness score and the recommended placement volume generated in step S40 through the built-in strategy mapping and decision logic, providing clear and executable decision support for managers.

[0122] For example, the system first compares the fitness score of the target area with a preset threshold range and automatically classifies the area. The classification results include four categories: priority delivery area, regular delivery area, observation delivery area, and restricted delivery area, with each category corresponding to different resource allocation and market operation strategies.

[0123] Subsequently, the system matches and verifies the recommended delivery volume with the region's historical sales, inventory levels, and fluctuations in end-user demand, and makes minor adjustments to the recommended value to form the final suggested delivery volume, ensuring that the output plan meets actual business constraints and operational continuity requirements.

[0124] Furthermore, the system automatically generates structured strategy reports, which clearly list each regional category and its corresponding deployment volume recommendations, core rationale, and risk warnings. For example, for "priority deployment areas," the system will suggest increasing supply allocation, strengthening terminal displays, and providing supporting promotional resources; for "restricted deployment areas," it will suggest suspending new deployments and strengthening sales monitoring and market research.

[0125] Ultimately, the system supports the visualization and multi-terminal publishing of strategy solutions, and can generate decision-making materials in various forms, including data dashboards, regional heat maps, and detailed tables. It also provides application programming interfaces for downstream business systems to call and execute, thereby achieving closed-loop management from prediction to strategy generation to business implementation.

[0126] In this embodiment, a technical solution combining a multi-layer long short-term memory (LSTM) network with an attention mechanism is used to achieve accurate prediction and strategy generation for regional adaptability in the cigarette market. The method first acquires multi-source historical data and constructs a spatiotemporal feature sequence. Then, an LSTM network enhanced with an attention mechanism is used to extract long-term dependent features, fully capturing the temporal patterns and spatial differences in market dynamics. During model training, by jointly optimizing regional consumption preference features and macroeconomic policy influencing factors, the model's adaptability to business scenarios and predictive accuracy are effectively improved. Finally, based on the model's output fitness score and recommended distribution volume, regional classification results and differentiated distribution strategy schemes are automatically generated, providing enterprises with quantitative and actionable decision support. This invention not only significantly improves the scientific rigor and refinement of cigarette distribution strategies but also has strong practicality and promotional value.

[0127] Based on the cigarette distribution area adaptability assessment method proposed in the above embodiments, this embodiment takes the Yunnan Province 2024 cigarette precision distribution decision support project as an example to specifically demonstrate the practical application of the method.

[0128] This embodiment selects historical commercial data from January 2020 to December 2023 from the Dianzhong urban agglomeration (Kunming, Qujing, Yuxi, and Chuxiong) in Yunnan Province, covering all 16 prefectures and cities in the province. The data dimensions used include, but are not limited to: monthly sales volume and sales revenue of each brand of cigarettes, market concentration measured by the Herfindahl-Hirschman Index (HHI), annual GDP and population data of each prefecture and city, and time-series data such as retailer social inventory and order fulfillment rates. In the data preprocessing stage, the raw data was cleaned and standardized; continuous numerical features were normalized using the Z-score method to eliminate the influence of dimensions.

[0129] In terms of model construction, this embodiment adopts the following... Figure 3 The network structure is shown. This region fitness evaluation model consists of a three-layer LSTM network enhanced by an attention mechanism, with 128 hidden units. During model training, we used the Adam optimizer with a learning rate of 0.001, a batch size of 32, and employed gradient pruning (with the gradient norm limited to 1.0) to prevent instability during training. Figure 4 As shown, the training loss and validation loss of the model converge well, the optimal validation point appears in the 68th round, and finally the model achieves a mean absolute error (MAE) of 0.042 and a root mean square error (RMSE) of 0.067 on the test set, indicating that the model has high accuracy.

[0130] After applying the trained model to predict the first quarter of 2024, the system outputs fitness scores and recommendation delivery volumes for each region. For example... Figure 5 , 6 As shown in the heatmap in Figure 7, for example, economically developed areas such as Kunming, Qujing, and Yuxi generally have adaptability scores higher than 0.8 and are classified as "priority deployment areas" by the system, with a recommended deployment volume increase of 10%-15% compared to the same period last year. Meanwhile, some areas in northwestern Yunnan have scores lower than 0.4 and are marked as "restricted deployment areas," with a recommended reduction of 20% in deployment volume and the need for market research. The remaining areas are categorized into "regular deployment" and "observation deployment" based on their score ranges, and corresponding differentiated strategies are generated.

[0131] After being piloted in some cities within the province, this strategy has achieved significant results through business verification. Compared with traditional methods, the turnover rate of cigarette social inventory in the pilot areas increased by approximately 12.6%, and the order fulfillment rate of retailers remained at a high level of over 95%, effectively optimizing resource allocation efficiency and demonstrating the effectiveness and superiority of the proposed method in actual business scenarios.

[0132] Furthermore, this embodiment also proposes a cigarette placement area adaptability assessment device, which includes:

[0133] The data acquisition module is used to acquire historical data for the target area;

[0134] The preprocessing module is used to clean, denoise, and standardize the historical data.

[0135] The feature extraction module is used to extract multi-dimensional spatiotemporal features from the preprocessed data to construct a feature sequence. The features include at least the sales revenue of each brand of cigarettes, the HHI index, and the regional annual GDP.

[0136] The model building module is used to build a multi-layer LSTM network based on an attention mechanism enhancement. The network includes gated recurrent units for extracting long-period dependencies in the input feature sequence.

[0137] The training module is used to train the LSTM network using the feature sequences. During the training process, it jointly optimizes regional consumption preference features and macroeconomic policy influencing factors, and uses gradient pruning to prevent gradient explosion.

[0138] The prediction module is used to predict the fitness of cigarette distribution in the target area in the future time period using the trained model, and generate fitness scores and recommended distribution amounts.

[0139] The decision output module is used to output a regionalized cigarette distribution strategy based on the score and distribution volume.

[0140] The storage module is used to store model parameters, intermediate data, and prediction results.

[0141] Furthermore, those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program includes program instructions and can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the cigarette placement area fitness assessment system to implement the process steps of the embodiments of the above methods.

[0142] Therefore, the present invention also provides a computer-readable storage medium storing a cigarette placement area fitness evaluation program, which, when executed by a processor, implements the various steps of the cigarette placement area fitness evaluation method as described in the above embodiments.

[0143] The computer-readable storage medium can be any computer-readable storage medium capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), magnetic disk, or optical disk.

[0144] It should be noted that, since the storage medium provided in the embodiments of this application is the storage medium used to implement the methods of the embodiments of this application, those skilled in the art can understand the specific structure and variations of the storage medium based on the methods described in the embodiments of this application, and therefore will not be repeated here. All storage media used in the methods of the embodiments of this application fall within the scope of protection of this application.

[0145] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0146] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0147] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0148] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0149] It should be noted that any reference signs placed between parentheses in the claims should not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

[0150] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0151] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for evaluating fitness of a cigarette dispensing area, characterized by, The method comprises the following steps: obtaining historical data of a target region, wherein the historical data comprises cigarette sales data, social economic indicators and retail inventory data; constructing a multi-dimensional spatio-temporal feature sequence based on the historical data, and using the multi-dimensional spatio-temporal feature sequence to represent the dynamic adaptability of the cigarette market with respect to time and space; inputting the spatio-temporal feature sequence into a trained regional adaptability evaluation model, wherein the regional adaptability evaluation model extracts long-period dependence features in the spatio-temporal feature sequence through a gate control loop unit; the regional adaptability evaluation model is a multi-layer LSTM network enhanced based on an attention mechanism, and is trained by jointly optimizing regional consumption preference features and macro policy influence factors; generating a cigarette launch adaptability score and a recommended launch amount of the target region in a future period according to the long-period dependence features; outputting a regionalized cigarette launch strategy scheme based on the cigarette launch adaptability score and the recommended launch amount.

2. The method of claim 1, wherein, The method of constructing the multi-dimensional spatio-temporal feature sequence comprises the following steps: performing cleaning, denoising and standardization preprocessing on the historical data raw data; extracting key features from the preprocessed data, wherein the key features comprise time series data of sales of each brand of cigarette, Herfindahl-Hirschman Index (HHI), annual GDP of the region, social inventory and order demand of retailers.

3. The method of claim 1, wherein, The single-layer calculation process of the multi-layer LSTM network enhanced based on the attention mechanism comprises: Compute input gate i t , forget gate f t , output gate o t , and candidate memory g t The computation expression is: i t = σ(W i · [h t-1 , x t ]+ b i ) f t = σ(W f · [h t-1 , x t ]+ b f ) o t = σ(W o · [h t-1 , x t ]+ b o ) g t = tanh(W g · [h t-1 , x t ] + b g ) updating the memory cell state c of the current time point t and the hidden state h t The calculation expression is: c t = f t · c t-1 + i t · g t h t = o t tanh(c t ) wherein x t is the input feature vector at time t; h t-1 is the hidden state at time t-1; c t-1 is the memory cell state at time t-1; W i , W f , W o , W g are weight matrices; b i , b f , b o , b g are bias vectors; and σ is a sigmoid activation function.

4. The method of claim 3, wherein, The operation of the attention mechanism comprises: calculating the attention weight distribution of all time step hidden states of the top layer of the multi-layer LSTM network; performing weighted summation on the hidden states of all time steps according to the attention weights to obtain a context vector; concatenating the context vector and the hidden state of the final time step to serve as a comprehensive feature representation for generating the adaptability score and the recommended launch amount.

5. The method of claim 1, wherein, The joint optimization of the regional consumption preference features and the macro policy influence factors during the training process comprises: The regional consumption preference features are quantified by historical brand sales proportion and volatility; The macro policy influence factor is used as an additional input channel in the model training or as a regularization term in a self-defined loss function to guide the model training.

6. The method of claim 1, wherein, Gradient clipping technology is used in the training process of the regional adaptability evaluation model to prevent gradient explosion.

7. The method of claim 1, wherein, The output of the regionalized cigarette launch strategy scheme comprises the following steps: comparing the adaptability score with a preset threshold to divide the target region into a priority launch category, a regular launch category, an observation launch category or a restricted launch category; generating a differentiated cigarette launch strategy scheme for different categories of regions in combination with the recommended launch amount.

8. A cigarette dispensing area fitness evaluation device, characterized by comprising: a cigarette dispensing area fitness evaluation device according to any one of claims 1 to 7. The method comprises the following steps: a data acquisition module for obtaining historical data of a target region; a preprocessing module for cleaning, denoising and standardizing the historical data; a feature extraction module for extracting multi-dimensional spatio-temporal features from the preprocessed data to construct a feature sequence, wherein the features at least comprise sales of each brand of cigarette, HHI index and annual GDP of the region; A model construction module is configured to construct a multi-layer LSTM network enhanced based on an attention mechanism, which comprises a gated recurrent unit for extracting long-period dependencies in an input feature sequence; A training module is configured to train the LSTM network using the feature sequence, and jointly optimize regional consumption preference features and macro policy influence factors during the training process, and adopt gradient clipping to prevent gradient explosion; A prediction module is configured to use the trained model to predict the cigarette delivery fitness of a target region in a future period, and generate a fitness score and a recommended delivery amount; A decision output module is configured to output a regional cigarette delivery strategy based on the score and the delivery amount; A storage module is configured to store model parameters, intermediate data, and prediction results.

9. A system for evaluating the adaptability of cigarette distribution areas, characterized in that, The method comprises the following steps: A memory, a processor, and a cigarette delivery region fitness evaluation program stored on the memory and executable on the processor, wherein the cigarette delivery region fitness evaluation program, when executed by the processor, implements the steps of the cigarette delivery region fitness evaluation method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a cigarette delivery region fitness evaluation program, and the cigarette delivery region fitness evaluation program, when executed by the processor, implements the steps of the cigarette delivery region fitness evaluation method according to any one of claims 1 to 7.

Images