CN-121998177-A - Method, device and equipment for regularly allocating cigarettes based on demand prediction deviation

CN121998177ACN 121998177 ACN121998177 ACN 121998177ACN-121998177-A

Abstract

The invention discloses a method, a device and equipment for regularly allocating cigarettes based on demand prediction deviation, which are used for constructing an sales prediction model by fusing holiday disturbance factors and sales hysteresis characteristics in demand construction, solving the limitation of the traditional technology in terms of exogenous variable processing such as holiday effect and the like, secondly, aiming at the problem of risk accumulation possibly caused by prediction errors, constructing a demand interval by adopting a demand construction method of a predicted value and historical sales data, and determining a recommended allocation interval of each cigarette commodity according to the demand interval, and based on the demand interval, compared with the traditional technology, the method for predicting an allocation feasible interval by using a single demand fixed value, the method can reduce prediction errors, thereby comprehensively considering the problem that exogenous variables and prediction errors can be accumulated in risks, providing a set of periodical source allocation decision frames integrating prediction and optimization, and ensuring the accuracy of cigarette allocation.

Inventors

ZHU HONG
HUANG LEI
LUO LI
WU LANG

Assignees

四川省烟草公司宜宾市公司

Dates

Publication Date: 20260508
Application Date: 20260109

Claims (10)

1. A method for regularly allocating cigarettes based on demand prediction deviation is characterized by comprising the following steps: acquiring historical marketing data of cigarette commodities; According to historical marketing data, constructing a sales quantity prediction model taking holiday disturbance characteristics of a sales period and sales hysteresis characteristics of cigarette commodities in the sales period as inputs and predicted sales quantity of the cigarette commodities in the sales period as outputs, wherein the holiday disturbance characteristics are used for representing whether the sales period is in the month of the holiday, and whether the next month of the sales period comprises the holiday and the interval days between the sales period and the holiday; Determining the predicted sales of each cigarette commodity in a plurality of continuous periods in the future from a target period by using a sales prediction model; According to the historical purchase and sale data and the predicted sales of each cigarette commodity, determining a demand interval of each cigarette commodity in a target period, and determining a recommended allocation interval of each cigarette commodity based on each demand interval; And constructing an allocation optimization model which takes the end stock quantity of the minimized target period as an optimization target by utilizing the recommended allocation interval of each cigarette commodity, and solving the allocation optimization model to obtain the optimal allocation of each cigarette commodity in the target period.
2. The method of claim 1, wherein constructing a sales volume prediction model with holiday disturbance characteristics of a sales cycle and sales hysteresis characteristics of a cigarette commodity in the sales cycle as inputs and predicted sales volume of the cigarette commodity in the sales cycle as outputs according to the historical sales data comprises: determining a plurality of sales periods according to the historical sales data, and determining the actual sales of each cigarette commodity in each sales period; constructing holiday disturbance characteristics of each sales period, and constructing sales hysteresis characteristics of each cigarette commodity in each sales period based on the historical marketing data; Constructing feature vectors of all cigarette commodities in all sales periods by utilizing holiday disturbance features of all sales periods and sales hysteresis features of all cigarette commodities in all sales periods; And training a machine learning model by taking the characteristic vector of each cigarette commodity in each sales period as input and the actual sales quantity of each cigarette commodity in each sales period as a label and the predicted sales quantity of each cigarette commodity in each sales period as output so as to obtain the sales quantity prediction model after training is finished.
3. The method of claim 2, wherein the historical stock data comprises actual sales of each of the cigarette goods in each of the historical cycles, wherein constructing a sales hysteresis feature of each of the cigarette goods in each of the sales cycles based on the historical stock data comprises: Acquiring the hysteresis orders of all cigarette commodities; For any sales period, determining the available history period of each cigarette commodity before any sales period according to the hysteresis order, wherein the available history period of any cigarette commodity is as follows And (2) and Representing the arbitrary sales cycle, q representing the hysteresis order; screening out the actual sales corresponding to the available historical periods of the cigarette commodities from the actual sales in the historical periods of the cigarette commodities; and constructing the sales hysteresis characteristic of each cigarette commodity in any sales period by utilizing the actual sales quantity corresponding to the available historical period of each cigarette commodity.
4. The method of claim 3, wherein constructing the feature vector for each of the cigarette products in each of the sales cycles using the holiday disturbance feature for each of the sales cycles and the sales hysteresis feature for each of the cigarette products in each of the sales cycles comprises: For any sales period, determining from each history period in the history stock data according to the hysteresis order to be at the first position History period and the first History periods between history periods; establishing holiday disturbance characteristics corresponding to the determined historical period; according to the determined holiday disturbance characteristics corresponding to the historical period and the holiday disturbance characteristics of any sales period, constructing a hysteresis characteristic of the holiday relative to the any sales period; and constructing a characteristic vector of each cigarette commodity in any sales period by utilizing the hysteresis characteristic of the holiday relative to any sales period and the sales hysteresis characteristic of each cigarette commodity in any sales period.
5. The method of claim 1, wherein the historical stock data comprises actual sales of each of the plurality of cigarette goods in each of the historical periods, wherein determining the demand interval of each of the plurality of cigarette goods in the target period based on the historical stock data and the predicted sales of each of the plurality of cigarette goods comprises: For any cigarette commodity, according to the historical stock data, calculating the safety stock quantity of the any cigarette commodity and the coverage cycle number of the any cigarette commodity relative to the target cycle, wherein each coverage cycle in the coverage cycle number is a cycle after the target cycle; Determining the predicted sales of each coverage week based on the coverage weeks and the predicted sales of any one of the cigarette commodities in a plurality of continuous periods in the future from a target period; Calculating a first predicted total demand using the number of coverage weeks and the predicted sales for each coverage week; calculating historical Zhou Jun sales of any cigarette commodity according to actual sales of the any cigarette commodity in each historical period, and calculating a second predicted total demand according to the historical Zhou Jun sales and the coverage cycle; And constructing a demand interval of any cigarette commodity according to the first predicted total demand and the second predicted total demand.
6. The method of claim 5, wherein the historical stock data further comprises an order advance period for each of the cigarette products at each order; According to the historical stock data, calculating the safety stock quantity of any cigarette commodity and the coverage cycle number of any cigarette commodity relative to the target cycle, wherein the method comprises the following steps: calculating the annual average sales and standard deviation of the sales of any cigarette commodity according to the actual sales of the any cigarette commodity in each history period; determining the demand scale grade and the demand stability grade of any cigarette commodity based on annual average sales and sales standard deviation; calculating the average lead period and the lead period standard deviation by utilizing the lead period of any cigarette commodity when ordering, the demand scale grade and the demand stability grade; and calculating the safety stock quantity of any cigarette commodity and the coverage week number of any cigarette commodity relative to the target period according to the average advance period and the standard deviation of the advance period.
7. The method of claim 1, wherein determining a recommended allocation interval for each cigarette commodity based on each demand interval comprises: For any cigarette commodity, acquiring the current stock quantity of the any cigarette commodity and the planned throwing quantity in a target period; calculating the safe stock quantity of any cigarette commodity according to the historical purchase-sale-stock data; based on the current stock quantity, the planned putting quantity, the safe stock quantity and the demand interval of any cigarette commodity, determining the upper bound and the lower bound of the recommended allocation interval of any cigarette commodity according to the following formula; ; in the formula, The upper boundary of the recommended allocation interval of any cigarette commodity is represented, Represents the upper boundary of the demand interval of any cigarette commodity, Representing the planned delivery amount of any cigarette commodity, Represents the safe stock quantity of any one of the cigarette commodities, Representing the current stock of any of the cigarette goods, Representing the lower boundary of the recommended allocation interval of any cigarette commodity, Representing the lower bound of the demand interval of any cigarette commodity.
8. The method of claim 1, wherein constructing an allocation optimization model targeting an end-of-period inventory of a minimum target period as an optimization objective using recommended allocation intervals for each cigarette commodity comprises: constructing the allocation optimization model according to the following formula; ; in the formula, Representing the said deployment optimization model, Represents the commodity collection of the cigarettes, Indicating the initial inventory of cigarette goods i during the target cycle, The actual allocation quantity of the cigarette commodity i in the target period is represented, The planned delivery amount of the cigarette commodity i in the target period is obtained, Representing the penalty weight coefficient(s), Indicating positive and negative deviations of the amount of modulation of the cigarette provider j from a multiple of 5, Representing a collection of cigarette suppliers; The constraint conditions of the allocation optimization model are as follows: , , , , ; in the formula, , The upper and lower boundaries of the recommended allocation interval of the cigarette commodity i are represented, Representing the maximum inventory of all cigarette suppliers, The allocation price of the cigarette commodity i is represented, Indicating the allocated preset upper limit of all the cigarette commodities, A map representing the correspondence of a cigarette commodity i to a cigarette provider, Indicating the total amount of allocation for the cigarette provider j, Representing that the cigarette provider j outputs the required virtual lot variable in integer multiples of 5.
9. A device is transferred regularly to cigarette based on demand prediction deviation, characterized in that includes: the acquisition unit is used for acquiring historical purchase-sale data of the cigarette commodity; The model construction unit is used for constructing a sales quantity prediction model taking the holiday disturbance characteristic of a sales period and the sales hysteresis characteristic of the cigarette commodity in the sales period as inputs and the predicted sales quantity of the cigarette commodity in the sales period as output according to the historical sales data, wherein the holiday disturbance characteristic is used for representing whether the sales period is in the month of the holiday, and whether the next month of the sales period comprises the holiday and the interval days between the sales period and the holiday; the sales predicting unit is used for determining predicted sales of each cigarette commodity in a plurality of continuous periods in the future from a target period by using a sales predicting model; the allocating unit is used for determining a demand interval of each cigarette commodity in a target period according to the historical purchase and sale data and the predicted sales of each cigarette commodity, and determining a recommended allocation interval of each cigarette commodity based on each demand interval; And the allocation unit is used for constructing an allocation optimization model which takes the end stock quantity of the minimum target period as an optimization target by using the recommended allocation quantity interval of each cigarette commodity, and solving the allocation optimization model to obtain the optimal allocation quantity of each cigarette commodity in the target period.
10. An electronic device comprising a memory, a processor and a transceiver in communication with each other in sequence, wherein the memory is configured to store a computer program, the transceiver is configured to receive and send a message, and the processor is configured to read the computer program and execute the method for regularly allocating cigarettes based on a predicted deviation of demand according to any one of claims 1 to 8.

Description

Method, device and equipment for regularly allocating cigarettes based on demand prediction deviation Technical Field The invention belongs to the technical field of commodity allocation, and particularly relates to a method, a device and equipment for regularly allocating cigarettes based on demand prediction deviation. Background In a cigarette supply chain, goods source allocation, namely a decision that a commercial company makes purchase and arrangement to goods for each cigarette industry enterprise, is a key link for connecting a production end and a terminal market, directly determines the satisfaction of enterprise supply capacity and terminal demand and influences the self inventory level and fund occupation, however, the industrial allocation decision is subjected to the combined action of multiple constraints such as uncertainty of an advance period, fluctuation of demand, heterogeneity of multiple goods regulations (obvious difference of different goods regulations in the advance period of demand fluctuation and replenishment), budget, capacity, whole box play a role in ordering and the like, and complex characteristics of multiple goods regulations and multiple constraints are presented, so how to accurately perform cigarette allocation planning becomes an important ring of inventory management and goods source allocation. In recent years, theoretical researches on inventory management and inventory allocation are deepened continuously, a powerful support is provided for enterprises to improve resource allocation efficiency in a supply chain environment, wherein students propose a demand prediction model based on mixed negative binomial distribution in the face of a highly fluctuating market environment, and an optimization model is built according to the demand prediction model so as to achieve the dual aims of minimizing the cost of goods intake and maximizing the inventory turnover rate, so that cigarette allocation is performed; meanwhile, a learner also proposes a technical scheme for predicting the on-line tobacco inventory by adopting a BP neural network model and constructing an industrial tobacco scene-oriented allocation and demand prediction framework, a verification result shows that the framework has remarkable improvement in the aspects of inventory turnover and supply chain response speed, the learner also proposes a sales prediction model-based and data-driven replenishment decision mechanism for carrying out cigarette allocation, and a historical sales data-based combined optimization model combining demand prediction and multi-objective warehouse allocation in an electronic market scene, but the technology has the following defects that (1) when the demand prediction is carried out, a demand set value is obtained according to the predicted sales, and the demand set value is adopted for allocation prediction, so that prediction errors are large, and (2) when the sales prediction is carried out, the consideration of exogenous variables (such as holiday factors) is insufficient, so that the sales prediction accuracy is poor, and the accuracy of subsequent allocation planning is influenced, and therefore, based on the shortages, how to provide a high-accuracy cigarette allocation based on the demand prediction deviation has become a problem to be solved regularly. Disclosure of Invention The invention aims to provide a method, a device and equipment for regularly allocating cigarettes based on demand prediction deviation, which are used for solving the problems that the exogenous variable is not considered enough and the allocation planning is performed by adopting a demand fixed value in the prior art, and the accuracy of the allocation planning is low. In order to achieve the above purpose, the present invention adopts the following technical scheme: In a first aspect, a method for regularly allocating cigarettes based on demand prediction deviation is provided, including: acquiring historical marketing data of cigarette commodities; According to historical marketing data, constructing a sales quantity prediction model taking holiday disturbance characteristics of a sales period and sales hysteresis characteristics of cigarette commodities in the sales period as inputs and predicted sales quantity of the cigarette commodities in the sales period as outputs, wherein the holiday disturbance characteristics are used for representing whether the sales period is in the month of the holiday, and whether the next month of the sales period comprises the holiday and the interval days between the sales period and the holiday; Determining the predicted sales of each cigarette commodity in a plurality of continuous periods in the future from a target period by using a sales prediction model; According to the historical purchase and sale data and the predicted sales of each cigarette commodity, determining a demand interval of each cigarette commodity in a target period, and determining a r