Supply chain demand prediction method and device based on data fusion, equipment and medium

Abstract

The invention relates to the technical field of supply chains and the technical field of artificial intelligence, and discloses a supply chain demand prediction method and device based on data fusion, equipment and a medium, and the method comprises the steps: carrying out the data fusion of social media text data, user behavior data, industry trend data, a prediction price and a prediction demand, and obtaining a first feature matrix; processing the first feature matrix to obtain a second feature matrix, and selecting the second feature matrix after dimension reduction as a current data feature; based on the current data characteristics, determining a predicted price fluctuation result of the specified product in the next period, when the predicted price fluctuation result is a rising result and the predicted demand quantity is higher than a preset demand quantity, generating a replenishment quantity of the specified product in the next period, and when the predicted price fluctuation result is a falling result and the predicted demand quantity is lower than the preset demand quantity, generating a replenishment quantity of the specified product in the next period. And generating the discount rate of the specified product in the next period. The supply chain demand prediction accuracy and the operation efficiency can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of supply chain and the technical field of artificial intelligence, in particular to a supply chain demand prediction method and device based on data fusion, equipment and medium. BACKGROUND

[0002] Under the background of the current digital transformation of the retail industry accelerating, supply chain demand prediction and inventory management have become a key link for enterprises to build core competitiveness. Because, accurate supply chain demand prediction can anticipate market trends and provide scientific basis for enterprise production planning and procurement strategy.

[0003] However, the existing supply chain demand prediction method is difficult to obtain the replenishment quantity or discount rate of the specified product in the next cycle, which is not conducive to improving the efficiency of obtaining the replenishment quantity or discount rate. The reason is that the existing supply chain demand prediction method mainly adopts an artificial acquisition method to obtain the replenishment quantity or discount rate of the specified product in the next cycle, and the artificial acquisition method consumes a large amount of human resources and time resources, increases the acquisition time of the replenishment quantity or discount rate of the specified product in the next cycle, and is easily affected by human intervention, so it is not conducive to improving the efficiency of obtaining the replenishment quantity or discount rate. SUMMARY

[0004] The embodiments of the present application provide a supply chain demand prediction method and device based on data fusion, equipment and medium to solve the technical problem that the existing supply chain demand prediction method is difficult to obtain the replenishment quantity or discount rate of the specified product in the next cycle, which is not conducive to improving the efficiency of obtaining the replenishment quantity or discount rate.

[0005] In a first aspect, the embodiments of the present application provide a supply chain demand prediction method based on data fusion, applied to an electronic device, the electronic device comprising an ARIMA model, a Transformer model, and a prediction model connected to the ARIMA model and the Transformer model respectively, the supply chain demand prediction method comprising: Accessing a network retail platform of a supply chain, and obtaining a price time series of a specified product from the network retail platform; Processing the price time series using the ARIMA model to generate a predicted price of the specified product in the next cycle, and processing the price time series using the Transformer model to generate a predicted demand quantity of the specified product in the next cycle; Inputting the predicted price and the predicted demand quantity into the prediction model, obtaining social media text data of the specified product using the prediction model, and performing data fusion on the social media text data, user behavior data, industry trend data, the predicted price and the predicted demand quantity to obtain a first feature matrix; The first feature matrix is processed through a self-attention time sequence network in the prediction model to obtain a second feature matrix, and the second feature matrix after dimension reduction is selected as the current data feature. Based on the DTW distance between the current data feature and the category center of each mode category, the predicted price fluctuation result of the specified product in the next period is determined, when the predicted price fluctuation result is an upward result and the predicted demand is higher than the preset demand, the replenishment quantity of the specified product in the next period is generated according to the pre-defined replenishment quantity model, when the predicted price fluctuation result is a downward result and the predicted demand is lower than the preset demand, the discount rate of the specified product in the next period is generated according to the pre-defined discount rate model.

[0006] In a possible implementation manner of the first aspect, the network retail platform of the supply chain is accessed, and a price time sequence of the specified product is obtained from the network retail platform, including: The network retail platform of the supply chain is accessed, and a data file is obtained from the network retail platform. The price of the specified product in the current period is extracted from the data file, and the price time sequence of the specified product is constructed through the price of the specified product in the current period.

[0007] In a possible implementation manner of the first aspect, the price time sequence is processed by using an ARIMA model to generate a predicted price of the specified product in the next period, and the price time sequence is processed by using a Transformer model to generate a predicted demand of the specified product in the next period, including: The price time sequence is processed by using an ARIMA model to generate a predicted price of the specified product in the next period. An estimate of the Transformer model is obtained, an error value between the estimate and an actual demand is obtained, a negative value of the error value is selected as a reward value, in the reinforcement learning training of the Transformer model, a parameter of the Transformer model is adjusted through a proximal policy optimization algorithm with the maximum reward value as the target, when the reward value converges, the adjustment of the parameter of the Transformer model is stopped, a trained Transformer model is obtained, the price time sequence is processed by using the trained Transformer model to generate a predicted demand of the specified product in the next period. In a possible implementation manner of the first aspect, the predicted price and the predicted demand are input into a prediction model, social media text data of the specified product is obtained by using the prediction model, the social media text data, user behavior data, industry trend data, the predicted price and the predicted demand are data fused to obtain a first feature matrix, including: The predicted price and the predicted demand quantity are input into a prediction model, social media text data of the specified product is obtained by using the prediction model, semantic coding is performed on the social media text data to obtain semantic features, behavior feature extraction is performed on user behavior data to obtain behavior features, and trend feature extraction is performed on industry trend data to obtain trend features; The semantic features, the behavior features, the trend features, the predicted price and the predicted demand quantity are spliced to obtain a first feature matrix.

[0008] In a possible implementation manner of the first aspect, the processing of the first feature matrix by the self-attention time sequence network in the prediction model to obtain a second feature matrix, and the selection of the second feature matrix after dimension reduction as the current data feature, include: The first feature matrix is processed by the self-attention time sequence network in the prediction model to obtain a second feature matrix. The second feature matrix is subjected to a dimension reduction operation by using a principal component analysis algorithm to obtain a second feature matrix after dimension reduction, and the second feature matrix after dimension reduction is selected as the current data feature.

[0009] In a possible implementation manner of the first aspect, the determination of the predicted price fluctuation result of the specified product in the next period based on the DTW distance between the current data feature and the category center of each mode category, the generation of the replenishment quantity of the specified product in the next period according to a predefined replenishment quantity model when the predicted price fluctuation result is an upward result and the predicted demand quantity is higher than a preset demand quantity, and the generation of the discount rate of the specified product in the next period according to a predefined discount rate model when the predicted price fluctuation result is a downward result and the predicted demand quantity is lower than the preset demand quantity, include: The historical price data of the specified product is obtained, the historical price data is processed by a dynamic time warping algorithm in the prediction model to obtain a distance matrix, and the historical price data is grouped according to the distance matrix to obtain a plurality of mode categories; The DTW distance between the current data feature and the category center of each mode category is obtained, the DTW distance between the current data feature and the category center of each mode category is processed to obtain the similarity between the current data feature and each mode category, the similarity between the current data feature and each mode category is normalized to obtain a probability value corresponding to each mode category, and the historical fluctuation result of the mode category with the highest probability value is selected as the predicted price fluctuation result of the specified product in the next period; When the predicted price fluctuation result is an upward result and the predicted demand quantity is higher than a preset demand quantity, the replenishment quantity of the specified product in the next period is generated according to the predicted demand quantity, the predicted price fluctuation result and a predefined replenishment quantity model; When the predicted price fluctuation result is a falling result and the predicted demand quantity is lower than the preset demand quantity, a discount rate of the specified product in a next period is generated according to the predicted price fluctuation result and a predefined discount rate model.

[0010] In a possible implementation of the first aspect, the plurality of pattern categories are an upsurging pattern, a falling pattern and a fluctuating pattern. The prediction model comprises a self-attention time sequence network and an algorithm module connected to the self-attention time sequence network, and the algorithm module is a module for performing a dynamic time warping algorithm. The replenishment quantity model is: ; The predicted demand quantity is: The amplitude value in the predicted price fluctuation result is: , The experience coefficient is: The discount rate model is: ; The discount rate is: The adjustment coefficient is: The amplitude value in the predicted price fluctuation result is:

[0011] In a second aspect, the embodiments of the present application provide a supply chain demand prediction device based on data fusion, applied to an electronic device, the electronic device comprising an ARIMA model, a Transformer model, a prediction model connected to the ARIMA model and the Transformer model respectively, and comprising: The acquisition module is configured to access a network retail platform of a supply chain, and acquire a price time sequence of a specified product from the network retail platform. The generation module is configured to process the price time sequence by using the ARIMA model to generate a predicted price of the specified product in a next period, and process the price time sequence by using the Transformer model to generate a predicted demand quantity of the specified product in the next period. The fusion module is configured to input the predicted price and the predicted demand quantity into the prediction model, acquire social media text data of the specified product by using the prediction model, and perform data fusion on the social media text data, user behavior data, industry trend data, the predicted price and the predicted demand quantity to obtain a first feature matrix. The dimension reduction module is configured to process the first feature matrix by using a self-attention time sequence network in the prediction model to obtain a second feature matrix, and select the second feature matrix after dimension reduction as a current data feature. ​​The prediction module is configured to determine a predicted price fluctuation result of the specified product in a next period based on a DTW distance between the current data feature and a category center of each pattern category, and generate a replenishment quantity of the specified product in the next period according to a predefined replenishment quantity model when the predicted price fluctuation result is an upward result and a predicted demand quantity is higher than a preset demand quantity, and generate a discount rate of the specified product in the next period according to a predefined discount rate model when the predicted price fluctuation result is a downward result and the predicted demand quantity is lower than the preset demand quantity.

[0012] In a third aspect, an electronic device is provided, which includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor, and the processor implements the supply chain demand prediction method of any one of the first aspect when executing the computer program.

[0013] In a fourth aspect, a computer readable storage medium is provided, which stores a computer program, and the computer program, when executed by a processor, implements the supply chain demand prediction method of any one of the first aspect.

[0014] In a fifth aspect, a computer program product is provided, which, when running on an electronic device, causes the electronic device to execute the supply chain demand prediction method of any one of the first aspect.

[0015] The embodiments of the present application have two advantages. On the one hand, the predicted price fluctuation result of the specified product in the next period is determined based on the DTW distance between the current data feature and the category center of each pattern category, and the replenishment quantity of the specified product in the next period is generated according to the predefined replenishment quantity model when the predicted price fluctuation result is the upward result and the predicted demand quantity is higher than the preset demand quantity, and the discount rate of the specified product in the next period is generated according to the predefined discount rate model when the predicted price fluctuation result is the downward result and the predicted demand quantity is lower than the preset demand quantity. Since the replenishment quantity or the discount rate does not need to be obtained manually, the time for obtaining the replenishment quantity or the discount rate is reduced, and the efficiency of obtaining the replenishment quantity or the discount rate is improved. On the other hand, since the replenishment quantity or the discount rate is not affected by manual intervention, the reliability of the replenishment quantity or the discount rate is improved. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creating any creative labor.

[0017] Figure 1An application scenario diagram of the supply chain demand prediction method provided by the embodiment of the present application is provided; Figure 2 A flowchart of the supply chain demand prediction method provided by the embodiment of the present application is provided; Figure 3 A flowchart of S205 provided by the embodiment of the present application is provided; Figure 4 A schematic block diagram of the supply chain demand prediction device provided by the embodiment of the present application is provided; Figure 5 A structural schematic diagram of the electronic device provided by the embodiment of the present application is provided. DETAILED DESCRIPTION

[0018] In order to make the objects, technical solutions and advantages of the present application clearer, further detailed description will be made to the present application in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0019] In the following description, specific details are set forth in connection with the specific system structures, techniques and the like to provide a thorough understanding of the embodiments of the present application. However, it should be apparent to those skilled in the art that the present application can be implemented in other embodiments without these specific details. In other cases, detailed description of well-known systems, devices, circuits and methods is omitted so as not to obscure the description of the present application with unnecessary details.

[0020] It should be understood that when used in the specification and the appended claims of the present application, the term "comprising" indicates the presence of the described features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0021] It should also be understood that the term "and / or" used in the specification and the appended claims of the present application means any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0022] As used in the specification and the appended claims of the present application, the term "if" can be interpreted as "when" or "upon" or "in response to a determination" or "in response to detecting" depending on the context. Similarly, the phrase "if it is determined" or "if [a described condition or event] is detected" can be interpreted as meaning "upon determining" or "in response to determining" or "upon detecting [a described condition or event]" or "in response to detecting [a described condition or event]", depending on the context.

[0023] In addition, in the description of the present application and the appended claims, the terms "first", "second", "third", etc. are used only to distinguish descriptions, and cannot be understood as indicating or implying relative importance.

[0024] In the present application, the reference "one embodiment" or "some embodiments" means that the specific features, structures or characteristics described in connection with the embodiment are included in one or more embodiments of the present application. Therefore, the statements "in one embodiment", "in some embodiments", "in other some embodiments", "in further some embodiments" and the like appearing in different places in the specification are not necessarily all referring to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized. The terms "include", "contain", "have" and their variants mean "include but not limited to", unless otherwise specifically emphasized.

[0025] In addition, the technical solutions of various embodiments can be combined with each other, but it must be based on the fact that a person skilled in the art can realize it, and when the combination of technical solutions appears contradictory or cannot be realized, it should be considered that the combination of technical solutions does not exist and is not within the protection scope required by the present application.

[0026] The flowchart shown in the accompanying drawings is only an example and does not necessarily include all contents and operations / steps, nor does it necessarily execute in the order described. For example, some operations / steps can be further divided, combined or partially merged, so the actual execution order may be changed according to the actual situation.

[0027] The supply chain demand prediction method provided by the embodiments of the present application can be applied to mobile phones, tablet computers, wearable devices, vehicle-mounted devices, augmented reality (AR) / virtual reality (VR) devices, notebook computers, ultra-mobile personal computers (UMPC), netbooks, personal digital assistants (PDA) and other electronic devices. The embodiments of the present application do not make any limitation on the specific type of electronic device.

[0028] Please refer to Figure 1 , Figure 1 The application scenario diagram of the supply chain demand prediction method provided by the embodiments of the present application is described in detail as follows: The electronic device accesses the network retail platform of the supply chain, and obtains the price time sequence of the specified product from the network retail platform.

[0029] The network retail platform is a composite e-commerce system based on distributed computing and multi-layer architecture.

[0030] In the embodiment of the present application, the electronic device can access the network retail platform of the supply chain, obtain the price time series of the specified product from the network retail platform, and improve the acquisition rate of the price time series.

[0031] Please refer to Figure 2 , Figure 2 is a flowchart of the supply chain demand prediction method provided by the embodiment of the present application, which can be applied to an electronic device, the electronic device comprising an ARIMA model, a Transformer model, and a prediction model connected to the ARIMA model and the Transformer model respectively. The output layer of the ARIMA model and the output layer of the Transformer model are connected to the input layer of the prediction model respectively.

[0032] As Figure 2 shown, the supply chain demand prediction method provided by the embodiment of the present application comprises the following steps, which are described in detail as follows: S201, accessing the network retail platform of the supply chain, and obtaining the price time series of the specified product from the network retail platform; The step of accessing the network retail platform of the supply chain and obtaining the price time series of the specified product from the network retail platform comprises the following steps: Accessing the network retail platform of the supply chain and obtaining a data file from the network retail platform; Extracting the price of the specified product in the current period from the data file, and constructing the price time series of the specified product by the price of the specified product in the current period.

[0033] The price of the specified product in the current period is , , which represents the price at the first time point, , which represents the price at the second time point, , which represents the price at the first time point.

[0034] , which is arranged in order to form the price time series of the specified product, and the price time series is , .

[0035] For the convenience of description, the following examples are given: Take 5, and the price of the specified product in the past 5 days is T={100, 102, 101, 105, 107}.

[0036] ​​Taking 7, the price of the specified product in the past 7 days: T={100, 102, 101, 105, 107, 106, 108}.

[0037] Among them, the price time series is the expansion of the product price in the time dimension, and the price time series records how the price of the specified product fluctuates over time in the current period. The price fluctuation rule can be captured through the price time series.

[0038] S202, using an ARIMA model to process the price time series to generate a predicted price of the specified product in the next period, using a Transformer model to process the price time series to generate a predicted demand of the specified product in the next period; Among them, the ARIMA model is used to process the price time series to generate a predicted price of the specified product in the next period, and the Transformer model is used to process the price time series to generate a predicted demand of the specified product in the next period, comprising: Using an ARIMA model to process the price time series to generate a predicted price of the specified product in the next period; Obtain the estimate of the Transformer model, obtain the error value between the estimate and the actual demand, select the negative value of the error value as the reward value, and in the reinforcement learning training of the Transformer model, maximize the reward value as the goal, adjust the parameters of the Transformer model through the proximal policy optimization algorithm, when the reward value converges, stop adjusting the parameters of the Transformer model, obtain the trained Transformer model, and use the trained Transformer model to process the price time series to generate a predicted demand of the specified product in the next period. Among them, the reward value of the reward function is inversely related to the error value, that is, the larger the error value, the lower the reward value; the smaller the error value, the higher the reward value. The reward function converts the error value into an intuitive incentive signal, thereby guiding the Transformer model to actively optimize behavior and gradually reduce the error value.

[0039] Among them, the Chinese name of the ARIMA model is: Autoregressive Integrated Moving Average Model, and the English name is: Autoregressive Integrated Moving Average Model.

[0040] S203, input the predicted price and the predicted demand into the prediction model, obtain the social media text data of the specified product using the prediction model, and perform data fusion on the social media text data, user behavior data, industry trend data, predicted price and predicted demand to obtain a first feature matrix; The predicted price and the predicted demand are input into the prediction model, social media text data of the specified product is obtained by using the prediction model, the social media text data, user behavior data, industry trend data, the predicted price and the predicted demand are data fused to obtain a first feature matrix, including: The predicted price and the predicted demand are input into the prediction model, social media text data of the specified product is obtained by using the prediction model, the social media text data is semantically encoded to obtain semantic features, the user behavior data is feature extracted to obtain behavior features, and the industry trend data is feature extracted to obtain trend features. The semantic features, the behavior features, the trend features, the predicted price and the predicted demand are spliced to obtain the first feature matrix.

[0041] The social media text data includes comments, likes, shares and topic discussion contents published by users on a social platform around the specified product.

[0042] The user behavior data includes a browsing duration and a number of times of adding to a shopping cart.

[0043] The industry trend data includes a total sales and a growth rate.

[0044] S204, the first feature matrix is processed by a self-attention time series network in the prediction model to obtain a second feature matrix, and the second feature matrix after dimension reduction is selected as a current data feature. The prediction model includes a self-attention time series network and an algorithm module connected to the self-attention time series network, and the algorithm module is a module for executing a dynamic time warping algorithm.

[0045] The dynamic time warping algorithm is an algorithm for measuring the similarity of two time series, and is especially suitable for processing sequences of different lengths or speed changes. The core idea of the dynamic time warping algorithm is to find the optimal nonlinear alignment path between two sequences by dynamic programming method, so as to minimize the cumulative distance after alignment.

[0046] The first feature matrix is processed by a self-attention time series network in the prediction model to obtain a second feature matrix, and the second feature matrix after dimension reduction is selected as a current data feature, including: The first feature matrix is processed by a self-attention time series network in the prediction model to obtain a second feature matrix. The principal component analysis algorithm is used to perform dimension reduction operation on the second feature matrix to obtain a second feature matrix after dimension reduction, and the second feature matrix after dimension reduction is selected as a current data feature.

[0047] S205, determining a predicted price fluctuation result of the specified product in the next period based on a DTW distance between the current data feature and a category center of each mode category, when the predicted price fluctuation result is an upward result and the predicted demand quantity is higher than a preset demand quantity, generating a replenishment quantity of the specified product in the next period according to a predefined replenishment quantity model, and when the predicted price fluctuation result is a downward result and the predicted demand quantity is lower than the preset demand quantity, generating a discount rate of the specified product in the next period according to a predefined discount rate model.

[0048] wherein the current period and the next period are in weeks or months.

[0049] For the convenience of illustration, the following examples are given: For example, if the current period is the first week, then the next period is the second week.

[0050] For example, if the current period is the second week, then the next period is the third week.

[0051] For example, if the current period is the fifth week, then the next period is the sixth week.

[0052] For example, if the current period is the first month, then the next period is the second month.

[0053] For example, if the current period is the second month, then the next period is the third month.

[0054] For example, if the current period is the fifth month, then the next period is the sixth month.

[0055] wherein the next period refers to the next period of the current period.

[0056] wherein the replenishment quantity model is: ; wherein, is the predicted demand quantity, is the amplitude value in the predicted price fluctuation result, , is an empirical coefficient.

[0057] wherein the discount rate model is: ; wherein, is the discount rate, is an adjustment coefficient, is the amplitude value in the predicted price fluctuation result.

[0058] wherein the plurality of mode categories are an upward mode, a downward mode, and a shock mode.

[0059] The upward mode is characterized by a sustained breakthrough of the resistance level, showing an upward trend. Downward pattern, such as consecutive falling, the price continues to fall; Shocking pattern, the price fluctuates up and down within a certain range, forming a box consolidation pattern.

[0060] Among them, obtaining the replenishment quantity of the specified product in the next period has two beneficial effects. On the one hand, it can avoid overstocking, reduce warehouse costs, inventory loss costs, and capital occupation costs caused by overstocking, and also prevent understocking and avoid emergency replenishment costs caused by lack of goods. On the other hand, based on the replenishment quantity of the specified product in the next period, the enterprise can dynamically adjust the replenishment strategy, flexibly respond to demand fluctuations, achieve accurate matching of inventory and demand, improve the response speed and flexibility of the supply chain, and enhance the competitiveness of the enterprise in the market.

[0061] Among them, obtaining the discount rate of the specified product in the next period is a key strategy to improve inventory turnover efficiency and optimize inventory structure. For slow-selling or overstocked specified products, as the storage time increases, not only will it continue to occupy valuable warehouse space and increase warehouse costs, but also may cause additional loss costs due to the obsolescence, damage, etc. of the specified product. By setting the discount rate of the specified product in the next period, market demand can be effectively stimulated, and the turnover speed of these specified products can be accelerated, so that the specified products can be converted into cash flow as soon as possible. In this way, it not only speeds up the quick clearing of the specified product and reduces the loss risk caused by overstocking, but also optimizes the resource allocation of upstream and downstream enterprises in the supply chain, making the production, procurement, logistics and other links more coordinated and orderly, thereby improving the response speed and flexibility of the entire supply chain and enhancing the risk resistance of the supply chain in market fluctuations.

[0062] The embodiments of the present application have two beneficial effects. On the one hand, based on the DTW distance between the current data features and the class center of each mode class, the predicted price fluctuation result of the specified product in the next period is determined. When the predicted price fluctuation result is an upward result and the predicted demand quantity is higher than the preset demand quantity, the replenishment quantity of the specified product in the next period is generated according to the pre-defined replenishment quantity model. When the predicted price fluctuation result is a downward result and the predicted demand quantity is lower than the preset demand quantity, the discount rate of the specified product in the next period is generated according to the pre-defined discount rate model. Since it does not need to be obtained manually, the acquisition time of the replenishment quantity or the discount rate is reduced, which is beneficial to improving the acquisition efficiency of the replenishment quantity or the discount rate. On the other hand, since it is not affected by manual intervention, it is beneficial to improving the reliability of the replenishment quantity or the discount rate.

[0063] Please refer to Figure 3 , Figure 3 The flowchart of S205 provided by the embodiments of the present application is described as follows: S301, obtain historical price data of a specified product, process the historical price data by a dynamic time warping algorithm in a prediction model to obtain a distance matrix, group the historical price data according to the distance matrix to obtain a plurality of mode categories; The historical price data is grouped by using the spectral clustering algorithm and the distance matrix to obtain the plurality of mode categories.

[0064] The spectral clustering algorithm (Spectral Clustering) is a nonlinear clustering method based on graph theory and linear algebra. The core logic of the spectral clustering algorithm is to map data to a low-dimensional feature space, mine potential complex structures in the data, and achieve efficient grouping.

[0065] S302, obtain a DTW distance between a current data feature and a category center of each mode category, process the DTW distance between the current data feature and the category center of each mode category to obtain a similarity between the current data feature and each mode category, normalize the similarity between the current data feature and each mode category to obtain a probability value corresponding to each mode category, and select a historical fluctuation result of a mode category with the highest probability value as a predicted price fluctuation result of the specified product in a next period. The DTW distance (Dynamic Time Warping) is used to measure the similarity between the current data feature and the category center of each mode category. The greater the DTW distance, the lower the similarity between the current data feature and the category center of the mode category. The smaller the DTW distance, the higher the similarity between the current data feature and the category center of the mode category.

[0066] The processing of the DTW distance between the current data feature and the category center of each mode category to obtain the similarity between the current data feature and each mode category includes: The reciprocal of the DTW distance between the current data feature and the category center of each mode category is selected as the similarity between the current data feature and each mode category.

[0067] The normalization of the similarity between the current data feature and each mode category includes: The similarity between the current data feature and each mode category is added to obtain a total similarity, and the similarity between the current data feature and each mode category is divided by the total similarity to obtain a probability value corresponding to each mode category.

[0068] For convenience of illustration, the following examples are given: The similarity between the current data feature and the upswing pattern is similarity 1; The similarity between the current data feature and the downswing pattern is similarity 2; The similarity between the current data feature and the swing pattern is similarity 3.

[0069] The similarity total sum is obtained by adding similarity 1, similarity 2 and similarity 3; The probability value corresponding to the upswing pattern is obtained by dividing similarity 1 by the similarity total sum; The probability value corresponding to the downswing pattern is obtained by dividing similarity 2 by the similarity total sum; The probability value corresponding to the swing pattern is obtained by dividing similarity 3 by the similarity total sum.

[0070] The historical fluctuation result of the mode category with the highest probability value is selected as the predicted price fluctuation result of the specified product in the next period.

[0071] For the convenience of description, the following examples are given: The probability value corresponding to the upswing pattern is probability value 1; The probability value corresponding to the downswing pattern is probability value 2; The probability value corresponding to the swing pattern is probability value 3.

[0072] When probability value 1 is the highest, the historical fluctuation result of the upswing pattern is selected as the predicted price fluctuation result of the specified product in the next period; When probability value 2 is the highest, the historical fluctuation result of the downswing pattern is selected as the predicted price fluctuation result of the specified product in the next period; When probability value 3 is the highest, the historical fluctuation result of the swing pattern is selected as the predicted price fluctuation result of the specified product in the next period.

[0073] S303, when the predicted price fluctuation result is an upswing result and the predicted demand quantity is higher than the preset demand quantity, the replenishment quantity of the specified product in the next period is generated according to the predicted demand quantity, the predicted price fluctuation result and a predefined replenishment quantity model; S304, when the predicted price fluctuation result is a downswing result and the predicted demand quantity is lower than the preset demand quantity, the discount rate of the specified product in the next period is generated according to the predicted price fluctuation result and a predefined discount rate model.

[0074] In the embodiments of the present application, the replenishment quantity or the discount rate of the specified product in the next period is determined, which does not need to be obtained manually, thereby reducing the acquisition time of the replenishment quantity or the discount rate and improving the acquisition efficiency of the replenishment quantity or the discount rate.

[0075] Corresponding to the supply chain demand prediction method described in the above embodiments, please refer to Figure 4 ,Figure 4 This is a schematic block diagram of the supply chain demand forecasting device provided in the embodiments of this application. Figure 4 The supply chain demand forecasting device 400 shown can be applied to, for example... Figure 1 The application scenario diagram shows electronic devices. The following section uses electronic devices as an example to illustrate this. Figure 4 The supply chain demand forecasting device 400 shown is described in detail. The supply chain demand forecasting device 400 may include an acquisition module 401, a generation module 402, a fusion module 403, a dimensionality reduction module 404, and a forecasting module 405.

[0076] The acquisition module 401 is used to access the online retail platform of the supply chain and obtain the price time series of a specified product from the online retail platform; The generation module 402 is used to process price time series using the ARIMA model to generate the predicted price of a specified product in the next period, and to process price time series using the Transformer model to generate the predicted demand of a specified product in the next period. The fusion module 403 is used to input the predicted price and predicted demand into the prediction model, use the prediction model to obtain social media text data of the specified product, and fuse the social media text data, user behavior data, industry trend data, predicted price and predicted demand to obtain the first feature matrix. The dimension reduction module 404 is used to process the first feature matrix through the self-attention temporal network in the prediction model to obtain the second feature matrix, and select the dimension-reduced second feature matrix as the current data feature; The prediction module 405 is used to determine the predicted price fluctuation result of a specified product in the next period based on the current data characteristics and the DTW distance between the category centers of each pattern category. When the predicted price fluctuation result is an increase and the predicted demand is higher than the preset demand, the replenishment quantity of the specified product in the next period is generated according to the predefined replenishment quantity model. When the predicted price fluctuation result is a decrease and the predicted demand is lower than the preset demand, the discount rate of the specified product in the next period is generated according to the predefined discount rate model.

[0077] It should be noted that the various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0078] The embodiments of the present application have two advantages. On the one hand, based on the DTW distance between the current data characteristics and the category center of each mode category, the predicted price fluctuation result of the specified product in the next period is determined. When the predicted price fluctuation result is an upward result and the predicted demand is higher than the preset demand, the replenishment quantity of the specified product in the next period is generated according to the predefined replenishment quantity model. When the predicted price fluctuation result is a downward result and the predicted demand is lower than the preset demand, the discount rate of the specified product in the next period is generated according to the predefined discount rate model. Since manual acquisition is not required, the acquisition time of the replenishment quantity or the discount rate is reduced, and the acquisition efficiency of the replenishment quantity or the discount rate is improved. On the other hand, since it is not affected by manual intervention, the reliability of the replenishment quantity or the discount rate is improved.

[0079] Please refer to Figure 5 , Figure 5 The structural schematic diagram of the electronic device provided by the embodiments of the present application is shown.

[0080] As Figure 5 shown, Figure 5 The electronic device 2 comprises at least one processor 20, a memory 21, and a computer program 22 stored in the memory 21 and executable on the at least one processor 20. The processor 20 implements the steps in any of the above method embodiments when executing the computer program 22.

[0081] The electronic device 2 can include, but is not limited to, the processor 20 and the memory 21. Those skilled in the art can understand that Figure 5 The electronic device 2 is only an example and does not constitute a limitation on the electronic device 2, and can include more or fewer components than shown, or combine certain components, or different components, for example, can also include input / output devices, network access devices, etc.

[0082] The processor 20 is configured to run the computer program 22 stored in the memory 21, and implement the following steps when executing the computer program 22: Accessing a network retail platform of a supply chain, and obtaining a price time series of a specified product from the network retail platform; Processing the price time series by using an ARIMA model to generate a predicted price of the specified product in the next period, and processing the price time series by using a Transformer model to generate a predicted demand of the specified product in the next period; Inputting the predicted price and the predicted demand into a prediction model, obtaining social media text data of the specified product by using the prediction model, and performing data fusion on the social media text data, user behavior data, industry trend data, the predicted price, and the predicted demand to obtain a first feature matrix; The first feature matrix is processed through a self-attention time sequence network in the prediction model to obtain a second feature matrix, and the second feature matrix after dimension reduction is selected as the current data feature. Based on the DTW distance between the current data feature and the category center of each mode category, the predicted price fluctuation result of the specified product in the next period is determined, when the predicted price fluctuation result is an upward result and the predicted demand is higher than the preset demand, the replenishment quantity of the specified product in the next period is generated according to the pre-defined replenishment quantity model, when the predicted price fluctuation result is a downward result and the predicted demand is lower than the preset demand, the discount rate of the specified product in the next period is generated according to the pre-defined discount rate model.

[0083] In some embodiments, the processor 20 is configured to implement: Accessing a network retail platform of a supply chain, and obtaining a data file from the network retail platform; Extracting the price of the specified product in the current period from the data file, and constructing a price time series of the specified product by using the price of the specified product in the current period.

[0084] In some embodiments, the processor 20 is configured to implement: Processing the price time series by using an ARIMA model to generate a predicted price of the specified product in the next period; Obtaining an estimate of the Transformer model, obtaining an error value between the estimate and the actual demand, selecting the negative value of the error value as a reward value, and adjusting the parameters of the Transformer model by using a proximal policy optimization algorithm in the reinforcement learning training of the Transformer model, when the reward value converges, stopping adjusting the parameters of the Transformer model, obtaining a trained Transformer model, and processing the price time series by using the trained Transformer model to generate a predicted demand of the specified product in the next period.

[0085] In some embodiments, the processor 20 is configured to implement: Inputting the predicted price and the predicted demand into a prediction model, obtaining social media text data of the specified product by using the prediction model, performing semantic encoding on the social media text data to obtain semantic features, performing feature extraction on user behavior data to obtain behavior features, and performing feature extraction on industry trend data to obtain trend features; Splicing the semantic features, the behavior features, the trend features, the predicted price and the predicted demand to obtain a first feature matrix.

[0086] In some embodiments, the processor 20 is configured to implement: The first feature matrix is processed by a self-attention time sequence network in the prediction model to obtain a second feature matrix. The second feature matrix is processed by using a principal component analysis algorithm to obtain a second feature matrix after dimension reduction, and the second feature matrix after dimension reduction is selected as the current data feature.

[0087] In some embodiments, the processor 20 is configured to: obtain historical price data of the specified product, process the historical price data by a dynamic time warping algorithm in the prediction model to obtain a distance matrix, group the historical price data according to the distance matrix to obtain a plurality of mode categories; obtain a DTW distance between the current data feature and a category center of each mode category, process the DTW distance between the current data feature and the category center of each mode category to obtain a similarity between the current data feature and each mode category, normalize the similarity between the current data feature and each mode category to obtain a probability value corresponding to each mode category, and select a historical fluctuation result of a mode category with the highest probability value as a predicted price fluctuation result of the specified product in a next period; when the predicted price fluctuation result is an upward result and the predicted demand quantity is higher than a preset demand quantity, generating a replenishment quantity of the specified product in the next period according to the predicted demand quantity, the predicted price fluctuation result and a predefined replenishment quantity model; when the predicted price fluctuation result is a downward result and the predicted demand quantity is lower than the preset demand quantity, generating a discount rate of the specified product in the next period according to the predicted price fluctuation result and a predefined discount rate model.

[0088] In some embodiments, the plurality of mode categories are an upward mode, a downward mode and a shock mode.

[0089] The processor 20 can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors, application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0090] The memory 21 can be an internal storage unit of the electronic device 2 in some embodiments, for example, a hard disk or a memory of the electronic device 2. The memory 21 can also be an external storage device of the electronic device 2 in other embodiments, for example, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 2. Further, the memory 21 can include both the internal storage unit and the external storage device of the electronic device 2. The memory 21 is used to store an operating system, an application program, a boot loader, data, and other programs, etc., for example, program codes of the computer program, etc. The memory 21 can also be used to temporarily store data that has been output or will be output.

[0091] It should be noted that the information interaction, execution process, etc. between the above-mentioned devices / units, since based on the same concept as the method embodiments of the present application, the specific functions and the technical effects brought by them can be referred to the method embodiments part, and will not be repeated here.

[0092] The computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the steps in each of the above method embodiments.

[0093] The computer readable storage medium stores program codes, and the program codes can be called and executed by the processor to implement the supply chain demand prediction method described in the above method embodiments.

[0094] The computer readable storage medium has a storage space for program codes.

[0095] The program codes include codes of any steps in the supply chain demand prediction method described in the above method embodiments.

[0096] For example, the program codes are called by the processor, and the following steps can be executed: Accessing a network retail platform of a supply chain, and obtaining a price time series of a specified product from the network retail platform; Processing the price time series by using an ARIMA model to generate a predicted price of the specified product in a next period, and processing the price time series by using a Transformer model to generate a predicted demand of the specified product in the next period; Inputting the predicted price and the predicted demand into a prediction model, obtaining social media text data of the specified product by using the prediction model, and performing data fusion on the social media text data, user behavior data, industry trend data, the predicted price, and the predicted demand to obtain a first feature matrix; The first feature matrix is processed by a self-attention time sequence network in the prediction model to obtain a second feature matrix, and the second feature matrix after dimension reduction is selected as the current data feature; Based on the DTW distance between the current data feature and the category center of each mode category, the predicted price fluctuation result of the specified product in the next period is determined, when the predicted price fluctuation result is an upward result and the predicted demand is higher than the preset demand, the replenishment quantity of the specified product in the next period is generated according to the pre-defined replenishment quantity model, when the predicted price fluctuation result is a downward result and the predicted demand is lower than the preset demand, the discount rate of the specified product in the next period is generated according to the pre-defined discount rate model.

[0097] The specific implementation of each operation can refer to the foregoing embodiments, which will not be repeated here.

[0098] The computer readable storage medium can also be an external storage device of the supply chain demand prediction apparatus or the electronic device, for example, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, a non-transitory computer readable storage medium, etc. equipped on the supply chain demand prediction apparatus or the electronic device.

[0099] Due to the computer program stored in the computer readable storage medium, any data fusion based supply chain demand prediction method provided by the embodiments of the present application can be executed, and thus the computer readable storage medium can achieve the beneficial effects of any data fusion based supply chain demand prediction method provided by the embodiments of the present application. Details can be found in the foregoing embodiments, which will not be repeated here.

[0100] The embodiments of the present application provide a computer program product, when the computer program product runs on the electronic device, the electronic device executes the above-mentioned supply chain demand prediction method.

[0101] The computer program product is loaded by the electronic device, and the following steps can be executed: Accessing a network retail platform of a supply chain, and obtaining a price time sequence of a specified product from the network retail platform; Processing the price time sequence by using an ARIMA model to generate a predicted price of the specified product in the next period, and processing the price time sequence by using a Transformer model to generate a predicted demand of the specified product in the next period; The predicted price and the predicted demand quantity are input into the prediction model, the social media text data of the specified product is acquired by using the prediction model, the social media text data, the user behavior data, the industry trend data, the predicted price and the predicted demand quantity are data fused, and a first feature matrix is obtained; The first feature matrix is processed by a self-attention time sequence network in the prediction model to obtain a second feature matrix, and the second feature matrix after dimension reduction is selected as a current data feature; Based on the DTW distance between the current data feature and the category center of each mode category, a predicted price fluctuation result of the specified product in the next period is determined, when the predicted price fluctuation result is an upward result and the predicted demand quantity is higher than a preset demand quantity, a replenishment quantity of the specified product in the next period is generated according to a predefined replenishment quantity model, and when the predicted price fluctuation result is a downward result and the predicted demand quantity is lower than the preset demand quantity, a discount rate of the specified product in the next period is generated according to a predefined discount rate model.

[0102] The integrated unit, if realized in the form of a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium.

[0103] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above functional units and modules is exemplified, and in actual application, the above functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or software. In addition, the specific names of each functional unit and module are only for easy distinction, and do not limit the protection scope of the present application. The specific working process of the unit and module in the system can refer to the corresponding process in the foregoing method embodiment, which will not be repeated here.

[0104] Based on such understanding, all or part of the processes in the above-mentioned embodiment methods can be implemented by a computer program instructing relevant hardware, the computer program can be stored in a computer readable storage medium, and the computer program can implement the steps of each method embodiment when executed by a processor. The computer program includes computer program code, which can be in the form of source code, object code, executable files or some intermediate forms, etc. The computer readable medium at least includes any entity or device capable of carrying the computer program code to an electronic device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal and a software distribution medium. For example, a U disk, a mobile hard disk, a magnetic disk or an optical disk, etc.

[0105] In the above embodiments, the description of each embodiment has its own focus, and the parts not described or recorded in a certain embodiment can be referred to the relevant description of other embodiments.

[0106] The above is only the preferred embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent process transformation using the content of the specification and drawings, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A data fusion based supply chain demand forecasting method, characterized by, The application is applied to an electronic device, and the electronic device comprises an ARIMA model, a Transformer model, and a prediction model connected with the ARIMA model and the Transformer model respectively. Accessing a network retail platform of a supply chain to obtain a price time sequence of a specified product from the network retail platform; Processing the price time sequence by using the ARIMA model to generate a predicted price of the specified product in a next period, and processing the price time sequence by using the Transformer model to generate a predicted demand of the specified product in the next period; Inputting the predicted price and the predicted demand into the prediction model, obtaining social media text data of the specified product by using the prediction model, and performing data fusion on the social media text data, user behavior data, industry trend data, the predicted price, and the predicted demand to obtain a first feature matrix; Processing the first feature matrix by using a self-attention time sequence network in the prediction model to obtain a second feature matrix, and selecting the second feature matrix after dimension reduction as a current data feature; Based on a DTW distance between the current data feature and a category center of each mode category, determining a predicted price fluctuation result of the specified product in the next period, when the predicted price fluctuation result is an upward result and the predicted demand is higher than a preset demand, generating a replenishment quantity of the specified product in the next period according to a predefined replenishment quantity model, and when the predicted price fluctuation result is a downward result and the predicted demand is lower than the preset demand, generating a discount rate of the specified product in the next period according to a predefined discount rate model.

2. The supply chain demand forecasting method of claim 1, wherein, The method comprises the following steps: Accessing a network retail platform of a supply chain to obtain a data file from the network retail platform; Extracting a price of the specified product in a current period from the data file, and constructing a price time sequence of the specified product by using the price of the specified product in the current period.

3. The supply chain demand forecasting method of claim 1, wherein, The method comprises the following steps: Processing the price time sequence by using the ARIMA model to generate a predicted price of the specified product in a next period; Obtaining an estimated value of the Transformer model, obtaining an error value between the estimated value and an actual demand, selecting a negative value of the error value as a reward value, adjusting parameters of the Transformer model by using a proximal policy optimization algorithm in reinforcement learning training of the Transformer model, stopping adjusting the parameters of the Transformer model when the reward value converges, obtaining a trained Transformer model, and processing the price time sequence by using the trained Transformer model to generate a predicted demand of the specified product in the next period.

4. The supply chain demand forecasting method of claim 1, wherein, The predicted price and the predicted demand quantity are input into a prediction model, social media text data of the specified product is obtained by using the prediction model, the social media text data, user behavior data, industry trend data, the predicted price and the predicted demand quantity are data fused to obtain a first feature matrix, including: The predicted price and the predicted demand quantity are input into a prediction model, social media text data of the specified product is obtained by using the prediction model, the social media text data is semantically encoded to obtain semantic features, the user behavior data is feature extracted to obtain behavior features, and the industry trend data is feature extracted to obtain trend features. The semantic features, the behavior features, the trend features, the predicted price and the predicted demand quantity are spliced to obtain the first feature matrix.

5. The supply chain demand forecasting method of claim 1, wherein, The first feature matrix is processed by a self-attention time sequence network in the prediction model to obtain a second feature matrix, and the second feature matrix after dimension reduction is selected as a current data feature, including: The first feature matrix is processed by a self-attention time sequence network in the prediction model to obtain a second feature matrix. The second feature matrix is dimensionally reduced by using a principal component analysis algorithm to obtain a second feature matrix after dimension reduction, and the second feature matrix after dimension reduction is selected as a current data feature.

6. The supply chain demand forecasting method of claim 1, wherein, The DTW distance between the current data feature and the category center of each mode category is determined to obtain a predicted price fluctuation result of the specified product in a next period, when the predicted price fluctuation result is an upward result and the predicted demand quantity is higher than a preset demand quantity, a replenishment quantity of the specified product in the next period is generated according to a predefined replenishment quantity model, when the predicted price fluctuation result is a downward result and the predicted demand quantity is lower than the preset demand quantity, a discount rate of the specified product in the next period is generated according to a predefined discount rate model, including: The historical price data of the specified product is obtained, the historical price data is processed by a dynamic time warping algorithm in the prediction model to obtain a distance matrix, and the historical price data is grouped according to the distance matrix to obtain a plurality of mode categories; The DTW distance between the current data feature and the category center of each mode category is determined to obtain a predicted price fluctuation result of the specified product in a next period, when the predicted price fluctuation result is an upward result and the predicted demand quantity is higher than a preset demand quantity, a replenishment quantity of the specified product in the next period is generated according to a predefined replenishment quantity model, when the predicted price fluctuation result is a downward result and the predicted demand quantity is lower than the preset demand quantity, a discount rate of the specified product in the next period is generated according to a predefined discount rate model, including: The DTW distance between the current data feature and the category center of each mode category is determined to obtain a predicted price fluctuation result of the specified product in a next period, when the predicted price fluctuation result is an upward result and the predicted demand quantity is higher than a preset demand quantity, a replenishment quantity of the specified product in the next period is generated according to a predefined replenishment quantity model, when the predicted price fluctuation result is a downward result and the predicted demand quantity is lower than the preset demand quantity, a discount rate of the specified product in the next period is generated according to a predefined discount rate model, including: The plurality of mode categories are an upward mode, a downward mode and a shock mode.

7. The supply chain demand forecasting method according to any one of claims 1 to 6, characterized in that, ​ The prediction model comprises a self-attention time series network and an algorithm module connected to the self-attention time series network, and the algorithm module is a module for executing a dynamic time warping algorithm. The replenishment quantity model is: ； wherein, is the predicted demand quantity, is the amplitude value in the predicted price fluctuation result, , is an empirical coefficient; The discount rate model is: ； wherein, is a discount rate, is an adjustment coefficient, is a magnitude value in the predicted price fluctuation result.

8. A data fusion based supply chain demand forecasting apparatus, characterized by, The application is applied to an electronic device, and the electronic device comprises an ARIMA model, a Transformer model, and a prediction model connected to the ARIMA model and the Transformer model. The acquisition module is configured to access a network retail platform of a supply chain and acquire a price time series of a specified product from the network retail platform. The generation module is configured to process the price time series by using the ARIMA model to generate a predicted price of the specified product in a next period, and process the price time series by using the Transformer model to generate a predicted demand quantity of the specified product in the next period. The fusion module is configured to input the predicted price and the predicted demand quantity into the prediction model, acquire social media text data of the specified product by using the prediction model, and perform data fusion on the social media text data, user behavior data, industry trend data, the predicted price, and the predicted demand quantity to obtain a first feature matrix. The dimension reduction module is configured to process the first feature matrix by using a self-attention time series network in the prediction model to obtain a second feature matrix, and select the second feature matrix after dimension reduction as a current data feature. The prediction module is configured to determine a predicted price fluctuation result of the specified product in the next period based on a DTW distance between the current data feature and a category center of each mode category, generate a replenishment quantity of the specified product in the next period according to a predefined replenishment quantity model when the predicted price fluctuation result is an upward result and the predicted demand quantity is higher than a preset demand quantity, and generate a discount rate of the specified product in the next period according to a predefined discount rate model when the predicted price fluctuation result is a downward result and the predicted demand quantity is lower than the preset demand quantity.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the supply chain demand prediction method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1-9. The computer program is executed by the processor to implement the supply chain demand prediction method according to any one of claims 1 to 7.

Images