Cigarette supply and demand target balanced cigarette delivery plan method, device, equipment and system

Abstract

The present application relates to a kind of cigarette supply and demand target balanced cigarette delivery plan method, device, equipment and system, belong to on-demand delivery technical field, wherein the cigarette supply and demand target balanced cigarette delivery plan method includes obtaining cigarette sales plan, the cigarette sales plan includes annual sales plan, half-year sales plan, monthly sales plan and week sales plan;Based on the plan decomposition function corresponding to the cigarette sales plan is solved, and the optimal week sales plan is obtained;Based on the plan delivery benchmark quantity and the individual demand prediction information determines the individualized source delivery scheme of progressive recommended delivery;Cigarette delivery is carried out through the individualized source delivery scheme. By fusing week delivery plan and individual demand, the coupling fitting of guest supply two-way target is realized, the interests of both sides are balanced, and the rationality of goods delivery is improved.

Description

Technical Field

[0001] This invention relates to the field of on-demand delivery technology, and in particular to a method, apparatus, equipment and system for cigarette delivery planning that balances supply and demand. Background Technology

[0002] Against the backdrop of a new round of technological revolution and industrial transformation, new productive forces in the field of artificial intelligence, represented by deep learning and big data analytics, are profoundly reshaping the value chain system of traditional industries. As an important pillar industry of the national economy, the tobacco industry is also undergoing profound changes in the continuous deepening of market-oriented reforms in cigarette marketing, which places higher demands on the accuracy of sales plan breakdown and the scientific nature of supply allocation.

[0003] In the sales plan breakdown stage, the traditional model relies heavily on the experience and judgment of employees, supplemented by manual calculations of a few core indicators. Limited by the limits of human energy and cognitive ability, the manual decision-making model struggles to achieve full coverage and accurate calculations when faced with a high-dimensional decision-making space interwoven with hundreds of cigarette brands and specifications (hereinafter referred to as specifications), multiple dimensions (time, region, customer level), and multiple indicators (sales volume, structure, efficiency). In the supply allocation stage, although the current allocation method based on customer level ensures fair distribution, it is difficult to effectively respond to dynamic market changes and differentiated needs at the individual level. Summary of the Invention

[0004] In view of this, it is necessary to provide a method, apparatus, equipment and system for cigarette distribution planning that balances supply and demand objectives, in order to solve the problem of difficulty in effectively responding to dynamic changes in the market and differentiated needs at the individual level.

[0005] To address the aforementioned problems, in a first aspect, the present invention provides a method for cigarette distribution planning that balances supply and demand, comprising: Obtain cigarette sales plans, including annual sales plans, semi-annual sales plans, monthly sales plans, and weekly sales plans; The optimal weekly sales plan is obtained by solving the plan decomposition function corresponding to the cigarette sales plan. The plan decomposition function includes an annual plan specification refinement function with pre-set optimization objectives, a semi-annual plan decomposition function to monthly functions, and a monthly plan decomposition function to weekly functions. The plan decomposition function is initially predicted by a deep learning algorithm, and the optimal weekly sales plan is obtained by iterative optimization using a reference point-oriented multi-objective genetic optimization method. A personalized supply distribution plan is determined based on the optimal weekly sales plan and individual demand forecast information, using a progressive recommendation approach.

[0006] In one possible implementation, the step of decomposing the sales plan based on the cigarette sales plan to obtain the optimal weekly sales plan includes: Construct a plan decomposition function based on the cigarette sales plan; The plan decomposition function is initially predicted using a deep learning algorithm, and the optimal weekly sales plan is obtained by iteratively optimizing each plan decomposition function in descending order of plan cycle length using a reference point-guided multi-objective genetic optimization method.

[0007] In one possible implementation, the step of constructing the plan decomposition function based on the cigarette sales plan includes: Obtain annual core indicator information, which includes sales data, sales growth rate data, and sales proportion data; Based on the aforementioned core annual indicator information, the following information is determined: weighting coefficient information, model calculation information, nonlinear sensitivity adjustment coefficient, specification allocation smoothing penalty coefficient, annual predicted sales volume of specifications, historical average sales volume, and the total number of cigarette specifications decomposed in the annual plan. Based on the weighting coefficient information, the model calculation information, the nonlinear sensitivity adjustment coefficient, the specification allocation smoothing penalty coefficient, the annual predicted sales volume of the specification, the historical average sales volume, and the total number of cigarette specifications, an annual plan specification refinement function is constructed. Based on the annual plan specification refinement function, the plan is decomposed to obtain the semi-annual sales plan, monthly sales plan, and weekly sales plan; The plan decomposition function is determined based on the annual plan specification refinement function, the semi-annual sales plan, the monthly sales plan, and the weekly sales plan.

[0008] In one possible implementation, the step of decomposing the plan based on the annual plan specification refinement function to obtain the semi-annual sales plan, monthly sales plan, and weekly sales plan includes: Based on the annual plan specification refinement function, the annual plan is decomposed into a six-month sales plan; Based on the product specification refinement function of the semi-annual plan, the annual plan is decomposed into monthly sales plans; The annual plan is decomposed into weekly sales plans based on the monthly plan specification refinement function.

[0009] In one possible implementation, determining the personalized supply allocation scheme based on the planned allocation baseline and the individual demand forecast information, including: Obtain individual demand information, and perform order volume prediction based on deep time series decomposition based on the individual demand information to determine individual demand prediction information; The baseline quantity for planned sales is determined based on the optimal weekly sales plan. A personalized supply allocation plan is determined based on the planned baseline quantity and the individual demand forecast information, using a progressive recommendation approach.

[0010] In one possible implementation, determining the personalized supply allocation scheme based on the planned allocation baseline and the individual demand forecast information, including: Based on the planned deployment baseline quantity and the individual demand forecast information, a dual-path baseline forecast solution is performed to obtain the dual-path baseline value, which includes the single-product-specification-single-customer planned baseline deployment value and the individual demand forecast value. Based on the single-product-specification-single-customer plan baseline deployment value and individual demand forecast value, the optimal dynamic fusion weight for the current period is obtained through multi-dimensional feature quantification and PSO adaptive weight optimization. Substitute the dual-path baseline value and the current optimal dynamic fusion weight into the multi-constraint fusion formula, and iteratively solve to obtain the optimal delivery amount. A personalized supply distribution plan is determined based on the optimal supply volume.

[0011] In one possible implementation, the step of solving the dual-path baseline prediction based on the planned deployment baseline quantity and the individual demand forecast information to obtain the dual-path baseline value includes: Based on the planned deployment baseline quantity and the multi-level plan decomposition model, the planned baseline deployment value for a single product specification and a single customer is calculated. Based on the individual demand prediction information and the inverse residual deep time series decomposition model, the individual demand is solved.

[0012] Secondly, this application also provides a cigarette distribution planning device for balancing supply and demand, comprising: a planning module, an optimization module, a recommendation module, and a distribution module; the planning module is used to obtain a cigarette sales plan, which includes an annual sales plan, a semi-annual sales plan, a monthly sales plan, and a weekly sales plan; the optimization module is used to decompose the cigarette sales plan to obtain an optimal weekly sales plan; the recommendation module is used to determine a personalized supply distribution scheme with progressive recommendations based on the planned distribution baseline quantity and the individual demand forecast information; the distribution module distributes cigarettes according to the personalized supply distribution scheme.

[0013] Thirdly, the present invention also provides an electronic device, including a memory and a processor, wherein, The memory is used to store programs; The processor, coupled to the memory, is used to execute the program stored in the memory to implement the steps in the cigarette distribution plan method for balancing supply and demand targets as described in any of the above implementations.

[0014] Fourthly, the present invention also provides a cigarette distribution planning system for balancing supply and demand targets. The cigarette distribution planning system for balancing supply and demand targets includes electronic equipment, data acquisition equipment, and output equipment. The electronic equipment is connected to the data acquisition equipment and the output equipment respectively. The data acquisition equipment is used to acquire cigarette sales plans, and the output equipment is used to generate and output personalized supply distribution plans.

[0015] Fifthly, the cigarette distribution planning system for balancing supply and demand targets provided by the present invention includes the aforementioned electronic device and a computer-readable storage medium. The storage medium is used to store computer-readable programs or instructions, which, when executed by the processor of the electronic device, can implement the steps in the cigarette distribution planning method for balancing supply and demand targets described in any of the above implementations.

[0016] The beneficial effects of this invention are as follows: The cigarette distribution planning method for balancing supply and demand provided by this invention first obtains a pre-defined cigarette sales plan, then decomposes the cigarette sales plan step by step to obtain an optimal weekly sales plan. Based on the optimal weekly sales plan, a personalized supply distribution scheme is progressively recommended from both the planning side and the individual demand side. Finally, cigarettes are distributed and sold according to the formulated personalized supply distribution scheme. By integrating the weekly distribution plan with individual demand, a coupled fit between customer and supply objectives is achieved, balancing the interests of both parties and improving the rationality of product distribution. Attached Figure Description

[0017] Figure 1 A flowchart illustrating the steps of a cigarette distribution planning method for balancing supply and demand targets, as provided in an embodiment of this application; Figure 2 A schematic diagram of the plan decomposition of a cigarette distribution planning method for balancing supply and demand targets provided in an embodiment of this application; Figure 3 A flowchart illustrating another step of the cigarette distribution planning method for balancing supply and demand targets provided in an embodiment of this application; Figure 4 A functional block diagram of a cigarette distribution planning device for balancing supply and demand targets provided in an embodiment of this application; Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0019] In the description of the embodiments of the present invention, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0020] The terms "first," "second," etc., used in the embodiments of this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a technical feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature.

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

[0022] This invention provides a method, apparatus, electronic device, and system for cigarette distribution planning that balances supply and demand, which will be described below.

[0023] Figure 1 This is a flowchart illustrating an embodiment of the cigarette distribution planning method for balancing supply and demand targets provided by the present invention. The execution entity of the cigarette distribution planning method for balancing supply and demand targets can be a computer or a server, or it can be a portable smart terminal. This embodiment does not limit it to this.

[0024] like Figure 1 As shown, the methods for cigarette distribution planning that balance supply and demand include: S101. Obtain the cigarette sales plan, which includes an annual sales plan, a semi-annual sales plan, a monthly sales plan, and a weekly sales plan.

[0025] It should be noted that, in response to technical issues, this study is based on the intersection of artificial intelligence and cigarette marketing, and proposes the following research objectives: (1) To construct a hierarchical, multi-objective collaborative intelligent sales plan decomposition model, integrating deep learning and time-series optimization technology to achieve fully automatic and high-precision decomposition from annual plan to weekly plan; (2) To design a set of on-demand delivery strategies that are compatible with the current system and empowered by technology, and through the four mechanisms of "fair guarantee, personalized response, regional adaptation, and new product cultivation", to promote the upgrading of the allocation of goods from "passive allocation" to "active matching" under the premise of ensuring fairness and compliance.

[0026] It should be understood that in the field of sales plan decomposition, early practices were mainly based on administrative directives, relying on the experience and judgment of local marketing personnel and the calibration of macro indicators. A review of domestic and international research shows that significant breakthroughs have been made in methodology and technical tools, particularly in the integration and application of artificial intelligence. However, current cigarette marketing still faces bottlenecks that urgently need to be overcome in the two key areas of intelligent sales plan decomposition and precise allocation of supplies. 1. Experience-dependent decomposition of plans: Although existing research has introduced regression and time series models, the decomposition process lacks systematic modeling of multi-objective constraints (total quantity control, structural optimization). The traditional model is mainly based on "total quantity control and hierarchical allocation," and the decomposition is highly dependent on the experience of marketing personnel and limited macro indicators, making it difficult to adapt to the complexity of high-dimensional decision-making spaces.

[0027] 2. Limitations of the rules for product allocation: The existing allocation mechanism based on product tiers does not adequately consider the actual needs of retail customers. Order recommendations rely heavily on the subjective experience of account managers, which can easily lead to problems such as insufficient accuracy in allocation and low customer satisfaction.

[0028] In view of this, the solution in this embodiment focuses on the collaborative optimization mechanism of intelligent decomposition and precise delivery driven by artificial intelligence. It breaks through the boundaries of human experience through deep learning and multi-objective optimization technology, and explores the transformation from "extensive allocation" to "precise delivery". The specific research contents include: 1. Constructing a sales plan decomposition model with multi-objective constraints to realize the systematic modeling of total volume control and structural optimization; 2. Designing a four-dimensional on-demand delivery strategy of "fair bottom line, personalized response, regional adaptation, and new product cultivation".

[0029] In practice, the first step is to obtain the cigarette sales plan. The cigarette sales plan can be broken down into annual, semi-annual, quarterly, monthly, and weekly levels (quarterly plans can be combined with semi-annual plans): 1. Annual Plan: It carries the strategic deployment function of the annual business objectives.

[0030] 2. Semi-annual Plan: The annual goals are broken down into two phases: the first half of the year (January-June) and the second half of the year (July-December). The semi-annual plan also involves the signing of business contracts. Its formulation method, key indicators, and decomposition steps are consistent with the annual plan. Usually, the annual plan and the semi-annual plan are formulated simultaneously.

[0031] 3. Quarterly benchmarking: The statistical cycle is the natural quarter. It is mainly used for benchmarking analysis of business performance and process monitoring. It ensures overall progress by identifying and dynamically adjusting target deviations. The quarterly level does not perform direct decomposition functions.

[0032] 4. Monthly Plan: As the core operational execution unit, it is directly decomposed from the semi-annual plan and refined to the product specification level. It is necessary to clarify the sales volume, amount and structural indicators of each product specification.

[0033] 5. Weekly Plan: Generated from the monthly plan, it accurately matches the pace of product delivery and supports the execution of the weekly delivery strategy.

[0034] Based on the above business process analysis, the sales plan decomposition model mainly addresses three core decomposition issues: First, the annual (semi-annual) plan is refined to the product specification level based on key indicators, achieving horizontal decomposition from total volume to product specification details; second, the annual (semi-annual) plan is decomposed to the monthly level, achieving vertical decomposition along the time dimension; and third, the monthly plan is decomposed to the weekly plan, achieving refined execution granularity. Simultaneously, the monthly and weekly plans need to be further refined to specific product specifications, thus forming a complete decomposition matrix of 'time dimension × product specification dimension' (see...). Figure 2 ).

[0035] S102. Solve the plan decomposition function corresponding to the cigarette sales plan to obtain the optimal weekly sales plan. The plan decomposition function includes an annual plan specification refinement function with pre-set optimization objectives, a semi-annual plan decomposition to monthly function, and a monthly plan decomposition to weekly function. The plan decomposition function is initially predicted by a deep learning algorithm, and the optimal weekly sales plan is obtained by iterative optimization using a reference point-oriented multi-objective genetic optimization method.

[0036] It should be noted that after determining the cigarette sales plan, the cigarette sales plan is broken down to obtain the optimal weekly sales plan.

[0037] In some embodiments, step S102 includes: constructing a plan decomposition function based on the cigarette sales plan; The plan decomposition function is initially predicted using a deep learning algorithm, and the optimal weekly sales plan is obtained by iteratively optimizing each plan decomposition function in descending order of plan cycle length using a reference point-guided multi-objective genetic optimization method.

[0038] It should be understood that, firstly, the model relies on artificial intelligence deep learning and multi-objective decomposition algorithms to mine the inherent patterns in historical data, replacing manual experience-based decomposition to achieve fully automated, high-precision decomposition from annual to weekly levels. The core requirements fall into four categories: First, the need for time-series pattern mining: using intelligent algorithms to model and capture the non-linear time-series correlations and product specification interaction characteristics of historical data, adapting to seasonal and trend-based market changes. Second, the need for multi-indicator synergy: achieving joint decomposition of multiple indicators such as sales volume, sales revenue, price range, and brand cultivation, ensuring logical consistency among indicators. Third, the need for full product specification adaptation: covering three categories of product specifications: currently selling, newly added, and planned for withdrawal, with withdrawal product specifications using equivalent price-level substitution. Fourth, the need for hierarchical closed-loop decomposition: achieving a chain-like decomposition from "year (half-year) → month → week," ensuring consistency between the lower-level aggregated value and the upper-level target value.

[0039] In some embodiments, to obtain the annual plan specification refinement function and then determine the optimal weekly sales plan, a plan decomposition function is constructed based on the cigarette sales plan, including: obtaining annual core indicator information, which includes sales volume data, sales growth rate data, and sales proportion data; determining weight coefficient information, model calculation information, nonlinear sensitivity adjustment coefficient, specification allocation smoothing penalty coefficient, annual predicted sales volume of the specification, historical average sales volume, and the total number of cigarette specifications decomposed in the annual plan based on the annual core indicator information; constructing the annual plan specification refinement function based on the weight coefficient information, the model calculation information, the nonlinear sensitivity adjustment coefficient, the specification allocation smoothing penalty coefficient, the annual predicted sales volume of the specification, the historical average sales volume, and the total number of cigarette specifications; performing plan decomposition based on the annual plan specification refinement function to obtain a semi-annual sales plan, a monthly sales plan, and a weekly sales plan; and determining the plan decomposition function based on the annual plan specification refinement function, the semi-annual sales plan, the monthly sales plan, and the weekly sales plan.

[0040] It should be noted that the model is guided by minimizing decomposition bias, conforming to market rules, and ensuring that indicators are met. Objective functions are constructed for the three types of decomposition tasks respectively: Annual (Semi-annual) Plan Product Specification Refinement Function: The core task of this function is to decompose 21 core objectives, such as total annual or semi-annual sales volume and sales revenue, into specific product specifications, forming a product specification-level plan. The objective function is as follows:

[0041] Among them, F1 is the target function value of the annual (semi-annual) plan refined to the product specification level. It is used to quantify the overall deviation between the decomposed plan of each product specification and the core target value. The smaller the F1 value, the higher the fit between the plan and each core target, and the stronger the feasibility of implementation. k is the annual core indicator identifier (k=1~21), covering 21 core indicators such as total sales volume, total sales amount, sales volume of cigarettes within the province, sales volume of cigarettes outside the province, sales volume of imported cigarettes, sales volume of cigarettes from the first to the eighth tier, sales growth rate and sales proportion of key brands, sales growth rate and sales proportion of industry-co-developed product specifications, sales growth rate and sales proportion of low-tar cigarettes, and sales growth rate and sales proportion of key first and second-tier brands; w k The weight coefficient for the k-th core indicator is dynamically assigned based on the operational importance of each indicator. X k pred This is the model-calculated value of the k-th core indicator; X k target Let k be the target value of the core indicator. This is a non-linear sensitivity adjustment coefficient, used to enhance the responsiveness to deviations in core indicators; Assign a smoothing penalty coefficient to each product specification to control the fluctuation range of product specification sales; Forecast annual / semi-annual sales for the i-th product specification; is the average sales volume of the i-th product specification over the past three years; N is the total number of cigarette product specifications participating in the annual / semi-annual plan breakdown.

[0042] In some embodiments, the plan is decomposed based on the annual plan specification refinement function to obtain a semi-annual sales plan, a monthly sales plan, and a weekly sales plan, including: decomposing the annual plan into a semi-annual sales plan based on the annual plan specification refinement function; decomposing the annual plan into a monthly sales plan based on the semi-annual plan specification refinement function; and decomposing the annual plan into a weekly sales plan based on the monthly plan specification refinement function.

[0043] It should be understood that the six-month plan is broken down into monthly functions, which decompose the half-year plan objectives into monthly plans. The objective functions are as follows:

[0044] Here, F2 is the comprehensive objective function value of the semi-annual plan decomposed into monthly values. The smaller the F2 value, the closer the monthly decomposition results are to management objectives, the more consistent they are with historical sales rhythms, and the fewer violations exceeding limits. This function adopts a hierarchical design of "indicator fitting term + structural stability term + constraint penalty term," which, compared to a single error minimization method, can explicitly incorporate business rules and improve the feasibility of the solution. m is the month index (m=1...12), t is the monthly core indicator index (which can be set according to business as sales volume percentage, sales amount percentage, price structure percentage, origin structure percentage, etc.), and i is the product specification index. X m,t pred and X m,t target represents the calculated value and target value of the t-th indicator in month m, respectively; wt is the weight of the t-th indicator, used to characterize the business priority of multiple indicators. q i pred and q i hist These represent the predicted average monthly sales and the historical average monthly sales for product specification i, respectively, used to constrain the consistency between the decomposition results and the historical sales rhythm. q i,m pred For the predicted sales volume of product specification i in month m, L 1，i L 2，i λ1, λ2, and λ3 represent the upper limits of the first and second half-year plans for specification i, respectively; (x) += max(0,x) indicates that only the portion exceeding the limit is penalized. Parameters λ1, λ2, and λ3 control the trade-off strength of the three sub-objectives: "target fitting", "structural stability", and "half-year constraint", respectively, and ε is a numerical stability term introduced to prevent the denominator from being zero.

[0045] In the specific implementation, the monthly plan is broken down into weekly functions, refining the monthly plan to the weekly level. The objective function is as follows:

[0046] Where F3 is the objective function value of the monthly plan broken down to the week, and the optimization objective is to minimize the deviation between the decomposition result and the preset target value. p is the week of the month (usually 4 or 5 weeks); i is the product specification identifier (all product specifications to be launched); x pi target Let be the sales decomposition value of the i-th product specification in week p; x pi pred C represents the predicted sales value for the i-th product specification in week p, calculated based on historical data. p D is the total sales value for week p; pLet δ be the total sales target for week p; δ is a balancing coefficient used to adjust the weight ratio of "product specification weekly deviation" and "weekly total deviation" to ensure balanced control of both deviations. Its value was determined to be δ=0.5 after iterative optimization using a grid search method, based on the experimental samples in this study.

[0047] This function uses a dual deviation constraint to ensure that the weekly breakdown values ​​for each product specification conform to historical consumption patterns, and to guarantee that the total weekly sales target is met.

[0048] In some embodiments, the solution of this embodiment also introduces a dynamic product specification adjustment mechanism: the product specifications on sale are adjusted every six months, and there are situations where product specifications are withdrawn or added; since the new product specifications lack historical data, they cannot be directly predicted and decomposed, so the model needs to establish a product specification adjustment prediction and connection mechanism.

[0049] For new product specifications, an adjustment strategy of "price-same-range trend prediction" is adopted. Based on historical product specification data at the same price range, the sales trend of the new product specifications is predicted to achieve a smooth introduction of the new product specifications. The specific process is as follows: Select historical product specifications that are consistent with the price range, brand type, and target consumer group of the new product specifications, and extract their sales data and decomposition patterns during the introduction period.

[0050] Specifically, the objective function solution and algorithm implementation include: (1) Annual (half-year) product specification decomposition solution: The objective function is solved using a two-step method: First, initial prediction of core indicators is carried out based on deep learning algorithms. Relying on at least five years of historical marketing data, an algorithm model is built to initially fit and predict the target values ​​of 21 core indicators, outputting the initial decomposition values ​​of each indicator and the corresponding product specification sales, and completing the preliminary quantitative calculation at the data level. Subsequently, professional marketing personnel and decision-makers correct and confirm the initial target values. Second, a reference point-oriented high-dimensional multi-objective genetic optimization method is adopted. Through non-dominated sorting and population iterative optimization, the Pareto optimal solution set is solved, providing marketing decision-makers with multiple alternative solutions with different focuses, all of which reach the optimal equilibrium.

[0051] (2) The semi-annual plan is decomposed into a monthly objective function, which is also solved using a two-step method: First, the integrated gradient boosting regression algorithm is applied to predict the initial target values ​​of 12 core indicators, and then the predicted initial target values ​​are manually confirmed. Second, the optimal decomposition scheme is solved by substituting the manually calibrated initial values ​​of the core indicators into the function, and the optimal monthly plan scheme is obtained through multiple rounds of iteration.

[0052] (3) Solving the weekly objective function by decomposing the monthly plan: Based on the integrated gradient boosting regression algorithm, firstly, a product specification-level weekly sales prediction model is constructed to generate the initial value of the weekly sales percentage of each product specification; then, guided by the objective function F3, the weekly total sales target D is constrained.p The predicted values ​​for each product specification are adjusted proportionally and iteratively optimized to ultimately generate a weekly plan.

[0053] S103. Based on the optimal weekly sales plan and individual demand forecast information, determine a personalized supply distribution plan for progressive recommendation.

[0054] It should be understood that after obtaining the optimal weekly sales plan, a gradual recommendation and delivery will be made based on the individual demand forecasts of different merchants and customers.

[0055] In its implementation, the on-demand allocation in this study is not an unrestrained free distribution, but rather an intelligent supply allocation mechanism based on market demand, data, and algorithms, grounded in normative documents such as the regulations for the management of cigarette supply allocation. It can be defined as follows: under rigid constraints such as compliance with current regulations, achievement of predetermined sales targets, profit and tax obligations, and product structure optimization, it comprehensively considers the explicit and implicit needs of retail customers and the historical behavioral characteristics of similar customer groups, using intelligent algorithms to generate differentiated and personalized allocation plans to improve supply allocation efficiency and customer satisfaction.

[0056] The scientific classification and precise definition of product specifications are the foundation for implementing the on-demand distribution strategy. Based on the actual situation of cigarette marketing, four categories of product specifications are identified to provide a clear basis for tiered and differentiated distribution. The specific classifications are as follows: 1. In-demand specifications: Market supply cannot meet demand, total volume is strictly controlled, and distribution quotas are scarce. The core principle is fair and inclusive distribution. 2. Main selling specifications: Specifications with a large sales volume, essential market demand, and stable sales. The core principle is to fully meet demand and ensure supply on demand. 3. Slow-moving specifications: Specifications with high inventory and slow sales. The core principle is to digest inventory through regular on-demand distribution and precise point-of-sale promotion. 4. New product specifications: Newly introduced specifications that have not yet formed a market scale and need to be cultivated and promoted. The core principle is to prioritize high-end and high-quality customers and precise point-of-sale in key areas.

[0057] It should be noted that the on-demand delivery strategy consists of four differentiated delivery methods based on the above definition, forming a complete delivery strategy system of "fair guarantee, personalized response, regional adaptation, and new product cultivation", covering different product attributes, different product types, and different market stages.

[0058] 1. Allocation based on price tiers (core mechanism for ensuring fairness): This method is the basic safety net mechanism for the on-demand allocation strategy. Its core objective is to maintain the fairness of cigarette market allocation and the stability of its operational order. It is mainly applicable to scarce supplies and low-priced cigarettes for daily necessities. The implementation path is as follows: Based on the retail customer's price tier as the core allocation basis, and taking into account factors such as market type, business format, and urban / rural division, standardized allocation rules are formulated.

[0059] 2. On-Demand Selection and Progressive Recommendation Deployment (Personalized Response Mechanism and Intelligent Prediction Empowerment Mechanism): This approach represents the personalized adaptation and intelligent innovation of on-demand deployment. It primarily applies to balanced and fully satisfying demand for key products, balancing customer choice with precise algorithmic empowerment. It is implemented in two phases: On-Demand Selection and Progressive Recommendation. In the first phase, On-Demand Selection adopts a "price segment aggregation + customer self-ordering" model, breaking the limitations of independent deployment for single product specifications. It integrates key product specifications with similar prices and attributes into multiple price segment deployment units, setting a total deployment volume for each price segment. Within this total volume constraint, retailers are allowed to independently select product specifications and order quantities based on their own business conditions. In the second phase, Progressive Recommendation Deployment is the core innovation leveraging AI. Retail customers often have limited understanding of their needs for non-in-demand product specifications, making it difficult to explicitly express implicit potential demands. This phase relies on deep learning algorithms to mine historical ordering patterns and market sales data, integrating them with weekly deployment plans to automatically generate personalized order recommendation schemes. The gradual nature is reflected in two aspects: First, the model parameters are continuously iterated based on customer order feedback and market data, and the recommendation accuracy is gradually improved; second, value-added information such as product specification matching, gross profit calculation, and sales cycle are gradually output to guide customers to accept intelligent recommendations, realize the transformation from passive acceptance to active adoption, and deeply explore potential market demand.

[0060] 3. Regional Allocation (Precise Revitalization Mechanism for Slow-Moving Product Specifications): Regional allocation is a differentiated revitalization mechanism for slow-moving product specifications. It leverages regional consumption differences, customer preferences, and business model characteristics to achieve precise diversion and efficient clearance of slow-moving goods. The implementation process is as follows: Through data clustering analysis, the market is divided into subdivided units such as urban areas, towns, rural areas, business districts, scenic areas, and communities. Differences in consumer price preferences, product specification acceptance, and sales speed in each region are accurately assessed. Slow-moving product specifications are then attribute-labeled and matched to regions with matching consumer preferences and suitable inventory capacity, avoiding areas with high rates of slow-moving products.

[0061] 4. New Product Launch (Precision Cultivation Mechanism): New product launches employ a precise targeting model combining "designated regions + business categories + customer tags" to help new products quickly enter the market and establish a foothold in the market. The implementation path is as follows: First, based on the new product's price point, positioning, and target consumer group, appropriate launch regions are defined; then, suitable business formats such as specialty stores and supermarkets are identified; finally, customer tags are added to filter out the new product's target customer group, achieving precise targeting.

[0062] The four delivery methods described above correspond to different product attributes, market types, and management objectives, constituting a complete on-demand delivery strategy. Among them, delivery based on product tier rules, regional allocation, and new product delivery rely on existing rules and standardized processes. On-demand self-selection and progressive recommendation delivery methods, as core innovative carriers of AI-enabled precision cigarette marketing, are the core direction of this research. Therefore, this embodiment focuses on the progressive recommendation delivery model, constructing an order prediction and product recommendation model tailored to the cigarette retail scenario based on interpretable deep time-series prediction technology.

[0063] Compared with existing technologies, the cigarette distribution planning method for balancing supply and demand targets provided in this embodiment first obtains a pre-defined cigarette sales plan, then decomposes the cigarette sales plan step by step to obtain an optimal weekly sales plan. Based on this optimal weekly sales plan, a personalized supply distribution scheme is progressively recommended from both the planning side and the individual demand side. Finally, cigarettes are distributed and sold according to the formulated personalized supply distribution scheme. By integrating the weekly distribution plan with individual demand, a coupled fit between customer and supply targets is achieved, balancing the interests of both parties and improving the rationality of product distribution.

[0064] Based on the above embodiments, a specific scenario implementation method for the cigarette distribution plan that balances supply and demand objectives is proposed. First, a pre-defined cigarette sales plan is obtained. Then, the cigarette sales plan is decomposed level by level to obtain the optimal weekly sales plan. Next, based on the optimal weekly sales plan, a personalized supply distribution scheme is progressively recommended from both the planning side and the individual demand side. Finally, cigarettes are distributed and sold according to the formulated personalized supply distribution scheme. By integrating the weekly distribution plan with individual demand, a coupled fit between customer and supply objectives is achieved, balancing the interests of both parties and improving the rationality of product distribution.

[0065] The on-demand selection and progressive recommendation delivery models proposed in this paper differ in that the proposed model is supported by artificial intelligence technology. Through the dynamic integration of two modules, "on-demand selection" and "group similarity prediction - individual time series prediction", it achieves forward-looking and accurate recommendation of demand, focusing on "demand guidance empowered by algorithms".

[0066] The solution presented in this embodiment achieves three key benefits: First, it constructs a cigarette sales plan decomposition model based on a multi-objective intelligent optimization framework, covering three levels of decomposition tasks: "annual (semi-annual) → product specification," "semi-annual → monthly," and "monthly → weekly." This model enables high-precision decomposition with multi-indicator synergy, full product specification adaptation, and hierarchical closed-loop, with core indicator deviations controlled within ±2%. Second, it proposes a four-dimensional on-demand delivery strategy system based on "fairness as a safety net, personalized response, regional adaptation, and refined development of new products." The model focuses on constructing a progressive recommendation delivery model, which, under simulation verification, achieves a recommendation adaptation rate of 86.9%, with both group and individual prediction deviations below 8%. Third, empirical verification shows that the model reduces the time consumption for various plan decompositions by more than 95%, effectively solving the problem of indicator imbalance in traditional manual decomposition.

[0067] The solution in this embodiment enriches the research results in the intersection of artificial intelligence and cigarette marketing at the theoretical level, improves the theoretical system of precision marketing, and provides a methodological reference for intelligent marketing research in similar industries. At the practical level, it effectively solves the core pain points in traditional cigarette marketing, significantly reduces manual operation costs, improves the accuracy of sales plan decomposition and supply allocation, and helps industry enterprises achieve the transformation from "experience-driven" to "data-driven".

[0068] Please refer to Figure 3 , Figure 3 This is a flowchart illustrating one embodiment of the cigarette distribution planning method for balancing supply and demand targets according to this application. Depending on different needs, the order of the steps in this flowchart can be changed, and some steps can be omitted. This cigarette distribution planning method for balancing supply and demand targets can be applied to the aforementioned cigarette distribution planning device for balancing supply and demand targets, but is not limited thereto, and the embodiments of this application do not limit it in this regard.

[0069] This embodiment is a further improvement on the aforementioned embodiment. The main improvement lies in the gradual integration of the weekly delivery plan (enterprise planning side) and individual needs (customer demand side). The core objective of this model is to achieve a coupled fit between the goals of both the customer and the supply side by integrating the weekly delivery plan and individual needs, thereby balancing the interests of both parties.

[0070] The specific process of this embodiment is as follows: Figure 3 As shown, it includes the following steps: S301. Obtain individual demand information, and perform order volume prediction based on deep time series decomposition based on the individual demand information to determine individual demand prediction information.

[0071] It should be noted that the model employs three core modules: precise calculation of individual demand, rigid breakdown of weekly delivery plans, and two-way progressive integration recommendation. The core task of the model can be formally expressed as: Let the set of retail customers be C = {c1, c2, ..., c...}. mThe set of cigarette specifications is P = {p1, p2, ..., p}. n}, for any customer-product combination (c i ,p j Using a weekly iteration cycle, the system dynamically integrates the weekly delivery plan baseline value and individual demand forecast value to accurately generate the customer's optimal expected order quantity Q(i,j) for a single week, and then outputs a personalized supply recommendation solution.

[0072] It should be understood that the individual demand forecasting module is based on deep time-series decomposition for order volume forecasting. For time-series forecasting of weekly orders from retail customers, a deep interpretable time-series forecasting technique based on inverse residual links is employed to capture the seasonality, trend, and randomness characteristics of cigarette demand, thus adapting to accurate short-term forecasting of weekly order volumes from retail customers. Let the historical weekly order volume time-series sequence for a customer's single product specification be X=[x1,x2,...,x...]. t The model fits the time series data through multiple residual blocks, and the single-step prediction formula can be expressed as:

[0073] x t+1 pred θ represents the order quantity forecast for the next step; L represents the number of residual module layers; l F represents the weighting coefficients for each module. l (X) represents the fitting output of each residual module for the time series.

[0074] S302. Determine the baseline quantity for planned distribution based on the optimal weekly sales plan.

[0075] In practical implementation, weekly delivery plans are calculated. Based on the constructed multi-level sales plan decomposition model (annual / semi-annual / monthly / weekly), the overall operating goals of the tobacco company (total volume, structure, tax and profit, etc.) are broken down into weekly delivery plans. The planned delivery benchmark quantity Qg(i,j) for a single customer and a single product specification is output, which constitutes the rigid constraint boundary of demand integration, ensuring that personalized recommendations do not exceed the overall business management and control goals of the enterprise.

[0076] S303. Based on the planned deployment baseline and the individual demand forecast information, determine a personalized supply deployment plan for progressive recommendation.

[0077] It should be understood that the process involves first collecting data and performing feature engineering, then performing prediction and solution based on a dual-path baseline, followed by automatic optimization of weights, then iteratively solving the multi-constraint fusion formula to obtain the optimal delivery volume, and finally determining a personalized delivery plan.

[0078] In some embodiments, S303 includes: solving a dual-path benchmark prediction based on the planned deployment benchmark quantity and the individual demand forecast information to obtain dual-path benchmark values, wherein the dual-path benchmark values ​​include a single-item-single-customer planned benchmark deployment value and an individual demand forecast value; further optimizing the single-item-single-customer planned benchmark deployment value and the individual demand forecast value through multi-dimensional feature quantization and PSO adaptive weight optimization to obtain the current optimal dynamic fusion weight; substituting the dual-path benchmark value and the current optimal dynamic fusion weight into a multi-constraint fusion formula and iteratively solving to obtain the optimal deployment quantity; and determining a personalized supply deployment plan based on the optimal deployment quantity.

[0079] In the specific implementation, a dual-path baseline prediction is solved, followed by multi-dimensional feature quantization and PSO adaptive weight optimization. Four-dimensional features—customer order activity, demand stability, prediction reliability, and product specification fit—are quantified, and a time-series decay coefficient is introduced to correct the feature time series. A particle swarm optimization algorithm is used to globally iteratively optimize the four-dimensional adjustment coefficients λ1, λ2, λ3, and λ4, with the goal of minimizing the supply-demand fitting error, to solve for the optimal dynamic fusion weight α for the current period. t This enables intelligent and personalized weight assignment.

[0080] Finally, a multi-constraint nonlinear fusion iterative solution is performed. The dual-path baseline value and the optimal fusion weight are substituted into the multi-constraint fusion formula, and the gradient descent algorithm is used to iteratively solve the optimal delivery quantity Q(i,j) under the constraints of total quantity, single customer threshold, and fluctuation stability. After completing the dual verification of compliance and stability, a personalized cargo delivery plan is output.

[0081] In this embodiment, the progressive fusion recommendation mechanism dynamically balances the rigid constraints of the plan with individual needs. This research originally constructs a multi-constraint time-series adaptive coupled progressive fusion algorithm. Unlike conventional static weighted models, this algorithm integrates time-series decay features, multi-dimensional behavioral feature constraints, a precision feedback iteration mechanism, and business threshold truncation rules to achieve non-linear, progressive, and closed-loop coupled fusion of weekly delivery plans and individual retail needs. The core logic is "multi-feature quantification modeling—time-series decay weight iteration—dual-objective constraint optimization—progressive steady-state convergence".

[0082] First, multi-dimensional feature quantification and time-series decay weight modeling are performed to accurately depict the dynamic operating characteristics and demand credibility of retail customers. A single evaluation dimension is abandoned, and a weight calculation system incorporating four core features—demand activity, demand stability, prediction credibility, and time-series freshness—is constructed. Simultaneously, a time-series decay coefficient is introduced to weaken the influence of historical outdated data and strengthen the weight of recent operating behavior, aligning with the time-series dynamic characteristics of weekly cigarette ordering. An adaptive fusion weight model is constructed based on an improved Sigmoid nonlinear mapping function to achieve nonlinear dynamic updating of the weights. The formula system is as follows:

[0083] α t ∈(0,1) represents the fusion weight of individual needs in week t; σ(x) is an improved Sigmoid activation function, which introduces a slope coefficient k and an offset μ to solve the problems of gradient vanishing and insufficient weight discrimination in the traditional Sigmoid. For customer CI to product specifications p j The frequency of valid orders within the sliding time window represents the real-time business activity level; γ Δ t γ is the time-series decay coefficient (0 < γ < 1), and Δt is the time difference between the historical period and the current period, realizing time-series optimization with "high weight for recent behavior and decay for long-term behavior"; To define the coefficient of variation of order quantity within the sliding window, a minimum value ε is introduced to avoid a denominator of 0, thus accurately representing the stability of demand; Acc p,t-1 This refers to the prediction accuracy of the model in the previous cycle; The customer's product specification fit score is obtained by weighting business format, business district consumption preferences, and inventory carrying capacity; λ1, λ2, λ3, and λ4 are the optimal adjustment coefficients of the four-dimensional features, which are determined by iterative optimization through particle swarm optimization (PSO).

[0084] Then, the dual-objective constraint asymptotic fusion solution model is used: ; Q(i,j) represents the final integrated recommendation volume in week t; Q min Q max The upper and lower limits for weekly deployment of a single customer and a single product specification (industry regulations). This represents the maximum weekly total volume for product specification PJ across all channels; ΔQ limit To control the stability of supply distribution and avoid large fluctuations, a weekly fluctuation threshold is set; Q pre (i,j) represents the actual delivery volume of the previous week. The model solves for the optimal Q(i,j) using a constrained gradient descent iterative algorithm, achieving multiple optimizations in compliance, stability, and accuracy.

[0085] Finally, a multi-cycle progressive convergence collaborative fusion mechanism is proposed, constructing a three-stage closed-loop progressive convergence mechanism. Through weight adaptive iteration, error backpropagation, and dynamic fine-tuning of constraint boundaries, a steady-state coupling between the planning and demand sides is achieved. This mechanism possesses a clear technical iteration logic and innovation barriers. The specific progressive convergence process is as follows: Phase 1 (Cold Start Constraint Adaptation Period): Model initialization phase. Insufficient customer time-series data sample size; prediction reliability metric Acc. p,t 1 is too low, the time-decay weights shrink overall, and the fusion weight α is too low. tConstrained within a low threshold range. The algorithm prioritizes activating the plan's rigid constraint weights, using the weekly deployment plan baseline value Q. g With (i,j) as the core, we strictly adhere to the constraints of total volume and single customer limit to ensure the fairness and stability of market deployment and solve the problem of prediction distortion during the cold start of new models and new customers.

[0086] The second stage (dynamic iterative coupling period): With weekly rolling iterations, the sliding time window data is continuously updated, and the model prediction bias is corrected through the error backpropagation mechanism. p,t 1. Continuous improvement; simultaneously, based on stable customer ordering behavior, demand stability and operational adaptability indicators are optimized in tandem, α t As multidimensional features increase nonlinearly, the algorithm dynamically balances plan control with individual needs. Through a constraint threshold fine-tuning mechanism, it gradually releases personalized delivery space, continuously reduces supply and demand fitting errors, and achieves a gradual transition from "forced plan adaptation" to "dynamic supply and demand equilibrium".

[0087] The third stage (steady-state precise adaptation period): After multiple iterations, the model prediction error converges to the steady-state interval, the multi-dimensional feature weights tend to be optimal, and α... t The algorithm achieves optimal personalized thresholds for each client. While strictly adhering to corporate planning constraints and market stability constraints, it is guided by the real needs of individuals, achieving a stable delivery effect that "does not violate institutional constraints and fully covers market demand." Simultaneously, the model continuously retains its iterative update capabilities, adapting to dynamic scenarios such as market consumption upgrades and client business transformations, and possesses long-term adaptive optimization capabilities.

[0088] Compared with existing technologies, the cigarette distribution planning method for balancing supply and demand targets provided in this embodiment achieves a gradual integration of weekly distribution plans (enterprise planning side) and individual needs (customer demand side). The core objective of this model is to achieve a coupled fit between the two-way targets of supply and demand by integrating weekly distribution plans and individual needs, thereby balancing the interests of both parties.

[0089] Based on the same idea as the cigarette distribution planning method for balancing supply and demand in the above embodiments, this application also provides a cigarette distribution planning device for balancing supply and demand, which can be used to execute the above-described cigarette distribution planning method for balancing supply and demand. For ease of explanation, the structural schematic diagram of the embodiment of the cigarette distribution planning device for balancing supply and demand only shows the parts related to the embodiments of this application. Those skilled in the art will understand that the illustrated structure does not constitute a limitation on the device, and it may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0090] like Figure 4As shown, the cigarette distribution planning device 400 for balancing supply and demand includes a planning module 401, an optimization module 402, a recommendation module 403, and a distribution module 404. In some embodiments, the above modules can be programmable software instructions stored in memory and executable by a processor. It is understood that in other embodiments, the above modules can also be program instructions or firmware embedded in the processor.

[0091] The planning module 401 is used to obtain cigarette sales plans, which include annual sales plans, semi-annual sales plans, monthly sales plans, and weekly sales plans. The optimization module 402 is used to solve the optimal weekly sales plan based on the plan decomposition function corresponding to the cigarette sales plan. The plan decomposition function includes an annual plan specification refinement function with pre-set optimization objectives, a semi-annual plan decomposition to monthly function, and a monthly plan decomposition to weekly function. The plan decomposition function is initially predicted by a deep learning algorithm, and the optimal weekly sales plan is obtained by iterative optimization using a reference point-oriented multi-objective genetic optimization method. The recommendation module 403 is used to determine a personalized supply distribution plan for progressive recommendation based on the optimal weekly sales plan and individual demand forecast information.

[0092] The cigarette distribution planning device 400 with supply and demand target balance provided in the above embodiments can realize the technical solutions described in the above embodiments of the cigarette distribution planning method with supply and demand target balance. The specific implementation principles of each module or unit can be found in the corresponding content in the above embodiments of the cigarette distribution planning method with supply and demand target balance, and will not be repeated here.

[0093] Please refer to Figure 5 , Figure 5 This is a schematic diagram of an embodiment of the electronic device of this application. In this embodiment of the invention, the electronic device 500 includes a processor 501, a memory 502, and a display 503. Figure 5 Only some components of the electronic device 500 are shown, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.

[0094] In some embodiments, processor 501 may be a central processing unit (CPU), microprocessor, or other data processing chip, used to run program code stored in memory 502 or process data, such as the cigarette distribution plan method for balancing supply and demand targets in this invention.

[0095] In some embodiments, processor 501 may be a single server or a group of servers. The server group may be centralized or distributed. In some embodiments, processor 501 may be local or remote. In some embodiments, processor 501 may be implemented on a cloud platform. In one embodiment, the cloud platform may include a private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, intranet, multi-cloud, etc., or any combination thereof.

[0096] In some embodiments, memory 502 may be an internal storage unit of electronic device 500, such as a hard disk or memory of electronic device 500. In other embodiments, memory 502 may also be an external storage device of electronic device 500, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. equipped on electronic device 500.

[0097] Furthermore, the memory 502 may include both internal storage units of the electronic device 500 and external storage devices. The memory 502 is used to store application software and various types of data installed on the electronic device 500.

[0098] In some embodiments, display 503 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 503 is used to display information from electronic device 500 and to display visual user applications. Components 501-503 of electronic device 500 communicate with each other via a system bus.

[0099] In one embodiment, when processor 501 executes the cigarette delivery plan program for supply and demand target balancing stored in memory 502, the following steps can be implemented: Obtain cigarette sales plans, including annual sales plans, semi-annual sales plans, monthly sales plans, and weekly sales plans; The optimal weekly sales plan is obtained by solving the plan decomposition function corresponding to the cigarette sales plan. The plan decomposition function includes an annual plan specification refinement function with pre-set optimization objectives, a semi-annual plan decomposition function to monthly functions, and a monthly plan decomposition function to weekly functions. The plan decomposition function is initially predicted by a deep learning algorithm, and the optimal weekly sales plan is obtained by iterative optimization using a reference point-oriented multi-objective genetic optimization method. A personalized supply distribution plan is determined based on the optimal weekly sales plan and individual demand forecast information, using a progressive recommendation approach.

[0100] It should be understood that when the processor 501 executes the cigarette delivery plan program for supply and demand target balancing in the memory 502, in addition to the functions mentioned above, it can also perform other functions, as can be found in the description of the corresponding method embodiments above.

[0101] Furthermore, the embodiments of the present invention do not specifically limit the type of electronic device 500 mentioned, and the electronic device 500 can be any type of auxiliary electronic device.

[0102] Accordingly, this application also provides a cigarette distribution planning system for balancing supply and demand targets. The cigarette distribution planning system for balancing supply and demand targets includes an electronic device 500 and a computer-readable storage medium. The computer-readable storage medium is used to store computer-readable programs or instructions. When the programs or instructions are executed by a processor, they can realize the steps or functions of the cigarette distribution planning methods for balancing supply and demand targets provided in the above-described method embodiments.

[0103] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.), and the computer program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0104] The above provides a detailed description of the cigarette distribution plan method, apparatus, electronic equipment, and system for balancing supply and demand targets provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for cigarette distribution planning that balances supply and demand, characterized in that, include: Obtain cigarette sales plans, including annual sales plans, semi-annual sales plans, monthly sales plans, and weekly sales plans; The optimal weekly sales plan is obtained by solving the plan decomposition function corresponding to the cigarette sales plan. The plan decomposition function includes an annual plan specification refinement function with pre-set optimization objectives, a semi-annual plan decomposition function to monthly functions, and a monthly plan decomposition function to weekly functions. The plan decomposition function is initially predicted by a deep learning algorithm, and the optimal weekly sales plan is obtained by iterative optimization using a reference point-oriented multi-objective genetic optimization method. A personalized supply distribution plan is determined based on the optimal weekly sales plan and individual demand forecast information, using a progressive recommendation approach.

2. The cigarette distribution planning method for balancing supply and demand targets according to claim 1, characterized in that, The process of decomposing the sales plan based on the cigarette sales plan to obtain the optimal weekly sales plan includes: Construct a plan decomposition function based on the cigarette sales plan; The plan decomposition function is initially predicted using a deep learning algorithm, and the optimal weekly sales plan is obtained by iteratively optimizing each plan decomposition function in descending order of plan cycle length using a reference point-guided multi-objective genetic optimization method.

3. The cigarette distribution planning method for balancing supply and demand targets according to claim 2, characterized in that, The construction of the plan decomposition function based on the cigarette sales plan includes: Obtain annual core indicator information, which includes sales data, sales growth rate data, and sales proportion data; Based on the aforementioned core annual indicator information, the following information is determined: weighting coefficient information, model calculation information, nonlinear sensitivity adjustment coefficient, specification allocation smoothing penalty coefficient, annual predicted sales volume of specifications, historical average sales volume, and the total number of cigarette specifications decomposed in the annual plan. Based on the weighting coefficient information, the model calculation information, the nonlinear sensitivity adjustment coefficient, the specification allocation smoothing penalty coefficient, the annual predicted sales volume of the specification, the historical average sales volume, and the total number of cigarette specifications, an annual plan specification refinement function is constructed. Based on the annual plan specification refinement function, the plan is decomposed to obtain the semi-annual sales plan, monthly sales plan, and weekly sales plan; The plan decomposition function is determined based on the annual plan specification refinement function, the semi-annual sales plan, the monthly sales plan, and the weekly sales plan.

4. The cigarette distribution planning method for balancing supply and demand targets according to claim 3, characterized in that, The plan decomposition based on the annual plan specification refinement function yields a semi-annual sales plan, a monthly sales plan, and a weekly sales plan, including: Based on the annual plan specification refinement function, the annual plan is decomposed into a six-month sales plan; Based on the product specification refinement function of the semi-annual plan, the annual plan is decomposed into monthly sales plans; The annual plan is decomposed into weekly sales plans based on the monthly plan specification refinement function.

5. The cigarette distribution planning method for balancing supply and demand targets according to claim 1, characterized in that, The personalized supply allocation plan, which determines the progressive recommendation allocation based on the planned allocation baseline and the individual demand forecast information, includes: Obtain individual demand information, and perform order volume prediction based on deep time series decomposition based on the individual demand information to determine individual demand prediction information; The baseline quantity for planned sales is determined based on the optimal weekly sales plan. A personalized supply allocation plan is determined based on the planned baseline quantity and the individual demand forecast information, using a progressive recommendation approach.

6. The cigarette distribution planning method for balancing supply and demand targets according to claim 5, characterized in that, The personalized supply allocation plan, which determines the progressive recommendation allocation based on the planned allocation baseline and the individual demand forecast information, includes: Based on the planned deployment baseline quantity and the individual demand forecast information, a dual-path baseline forecast solution is performed to obtain the dual-path baseline value, which includes the single-product-specification-single-customer planned baseline deployment value and the individual demand forecast value. Based on the single-product-specification-single-customer plan baseline deployment value and individual demand forecast value, the optimal dynamic fusion weight for the current period is obtained through multi-dimensional feature quantification and PSO adaptive weight optimization. Substitute the dual-path baseline value and the current optimal dynamic fusion weight into the multi-constraint fusion formula, and iteratively solve to obtain the optimal delivery amount. A personalized supply distribution plan is determined based on the optimal supply volume.

7. The cigarette distribution planning method for balancing supply and demand targets according to claim 6, characterized in that, The process of solving the dual-path baseline prediction based on the planned deployment baseline quantity and the individual demand forecast information to obtain the dual-path baseline value includes: Based on the planned deployment baseline quantity and the multi-level plan decomposition model, the planned baseline deployment value for a single product specification and a single customer is calculated. Based on the individual demand forecast information and the inverse residual deep time series decomposition model, the individual demand forecast value is solved.

8. A cigarette distribution planning device for balancing supply and demand, characterized in that, include: The module includes a planning module, an optimization module, and a recommendation module. The planning module is used to obtain cigarette sales plans, which include annual sales plans, semi-annual sales plans, monthly sales plans, and weekly sales plans. The optimization module is used to solve the optimal weekly sales plan based on the plan decomposition function corresponding to the cigarette sales plan. The plan decomposition function includes an annual plan specification refinement function with pre-set optimization objectives, a semi-annual plan decomposition to monthly function, and a monthly plan decomposition to weekly function. The plan decomposition function is initially predicted by a deep learning algorithm, and the optimal weekly sales plan is obtained by iterative optimization using a reference point-oriented multi-objective genetic optimization method. The recommendation module is used to determine a personalized supply distribution plan based on the optimal weekly sales plan and individual demand forecast information, and to make progressive recommendations.

9. An electronic device, the electronic device comprising a processor and a memory, characterized in that, The memory is used to store instructions, and the processor is used to call the instructions in the memory to cause the electronic device to execute the cigarette distribution plan method for balancing supply and demand as described in any one of claims 1 to 7.

10. A cigarette distribution planning system for balancing supply and demand, characterized in that, The cigarette distribution planning system for balancing supply and demand includes electronic equipment, data acquisition equipment, and output equipment. The electronic equipment is connected to the data acquisition equipment and the output equipment respectively. The data acquisition equipment is used to acquire cigarette sales plans, and the output equipment is used to generate and output personalized supply distribution plans.

Images