Prediction method and device applied to supply chain and storage medium

By acquiring sales and historical supply chain data, and utilizing machine learning models and predefined functions, the problem of insufficient accuracy in supply chain logistics forecasting has been solved, enabling more accurate forecasting and resource allocation, thereby improving supply chain operational efficiency and user satisfaction.

CN119048149AInactive Publication Date: 2024-11-29BEIJING WODONG TIANJUN INFORMATION TECH CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410101803.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-01-24
Publication Date
2024-11-29
Estimated Expiration
Not applicable · inactive patent

Smart Images

  • Figure CN119048149A_ABST
    Figure CN119048149A_ABST
Patent Text Reader

Abstract

The invention provides a prediction method and device applied to a supply chain and a storage medium, and relates to the technical field of warehouse logistics. The prediction method applied to the supply chain comprises the following steps: acquiring sales volume historical data and supply chain historical data in the same time period as the sales volume historical data; according to the sales volume historical data and the supply chain historical data, collaborative historical data is determined based on a predetermined function, and the predetermined function has a unique solution under the condition that the sales volume historical data and the collaborative historical data are determined; obtaining a sales volume prediction sequence; obtaining a collaborative data prediction sequence according to the collaborative historical data sequence; and according to the sales prediction sequence and the collaborative data prediction sequence, determining a supply chain prediction sequence according to a predetermined function.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to the field of warehousing logistics technology, and in particular to a forecasting method, device and storage medium applied to a supply chain. Background Art

[0002] Supply chain planning coordination and execution optimization are the core areas of smart supply chains. When realizing the automated and intelligent execution of procurement, warehousing, transportation, fulfillment and other supply chain links, it is necessary to predict the demand for goods, logistics orders and cargo volume for the next month or week. Therefore, the forecasting technology capabilities will ultimately have a significant impact on sales revenue, inventory costs, etc.

[0003] At the same time, with the continuous development of supply chain agility and digitalization, logistics, as the end-execution system of the entire supply chain, the connection between supply chain and logistics is constantly facing new challenges. Logistics execution requires efficient and coordinated algorithms to ensure that logistics resources can respond to changes in the consumer market in a timely and rapid manner and meet fulfillment requirements.

[0004] Taking fast-moving consumer goods companies as an example, the logistics department or third-party logistics companies are responsible for planning the order quantity or cargo volume, and the logistics department prepares warehousing resources and transportation resources based on historical order quantity and cargo volume information. Summary of the invention

[0005] One object of the present disclosure is to improve the accuracy of forecast values ​​applied to supply chains.

[0006] According to one aspect of some embodiments of the present disclosure, a forecasting method applied to a supply chain is proposed, including: acquiring sales history data and supply chain history data of the same time period as the sales history data; determining collaborative historical data based on a predetermined function according to the sales history data and the supply chain history data, wherein the predetermined function has a unique solution when the sales history data and the collaborative historical data are determined; acquiring a sales forecast sequence; acquiring a collaborative data forecast sequence according to the sequence of the collaborative historical data; and determining a supply chain forecast sequence according to a predetermined function according to the sales forecast sequence and the collaborative data forecast sequence.

[0007] In some embodiments, the supply chain historical data includes at least one of order quantity historical data or cargo transportation burden historical data; the supply chain forecast sequence includes at least one of order quantity forecast sequence or cargo transportation burden forecast sequence.

[0008] In some embodiments, obtaining a sales forecast sequence includes: inputting a sequence of historical sales data into a first machine learning model to obtain a sales forecast sequence, wherein the first machine learning model is generated by training using sales sample data.

[0009] In some embodiments, obtaining a collaborative data prediction sequence based on a sequence of collaborative historical data includes: inputting the collaborative historical data into a second machine learning model to obtain a collaborative data prediction sequence, wherein the second machine learning model is generated by training with collaborative sample data, and the collaborative sample data is generated based on a predetermined function according to sales sample data and supply chain sample data.

[0010] In some embodiments, determining collaborative historical data based on a predetermined function according to sales historical data and supply chain historical data includes: determining collaborative historical data of sales volume and order quantity based on a predetermined function according to sales historical data and order quantity historical data; and determining collaborative historical data of sales volume and cargo transportation burden based on a predetermined function according to sales historical data and cargo transportation burden historical data.

[0011] In some embodiments, obtaining a collaborative data prediction sequence based on a sequence of collaborative historical data includes: inputting the collaborative historical data of sales volume and order volume into a third machine learning model to obtain a collaborative data prediction sequence of sales volume and order volume, wherein the third machine learning model is generated by training with collaborative sample data of sales volume and order volume, and the collaborative sample data of sales volume and order volume is generated based on a predetermined function according to the sales volume sample data and the order volume sample data; and inputting the collaborative historical data of sales volume and order volume into a fourth machine learning model to obtain a collaborative data prediction sequence of sales volume and cargo carrying burden, wherein the fourth machine learning model is generated by training with collaborative sample data of sales volume and cargo carrying burden, and the collaborative sample data of sales volume and cargo carrying burden is generated based on a predetermined function according to the sales volume sample data and the cargo carrying burden sample data.

[0012] In some embodiments, determining the supply chain forecast sequence according to a predetermined function based on the sales forecast sequence and the collaborative data forecast sequence includes: performing an inverse operation of the predetermined function based on the sales forecast sequence and the collaborative data forecast sequence of the order quantity to determine the forecast sequence of the order quantity; performing an inverse operation of the predetermined function based on the sales forecast sequence and the collaborative data forecast sequence of the cargo carrying burden to determine the forecast sequence of the cargo carrying burden.

[0013] In some embodiments, obtaining sales historical data and supply chain historical data for the same time period as the sales historical data includes: obtaining order information based on historical data, the order information including order generation time, and the type and quantity of goods in the order; based on a predetermined time granularity and order generation time, counting the order volume in the same time period, as well as the type and quantity of goods in the order, and obtaining a sequence of order volume historical data and a sequence of sales historical data, wherein the supply chain historical data includes order volume historical data.

[0014] In some embodiments, obtaining sales historical data and supply chain historical data for the same time period as the sales historical data includes: obtaining order information based on historical data, the order information including order generation time, the types and quantities of goods in the order, and the cargo transportation burden in the order; according to a predetermined time granularity and order generation time, counting the types and quantities of goods in the orders for the same time period, as well as the cargo transportation burden in the orders, to obtain a sequence of sales historical data and a sequence of cargo transportation burden historical data, wherein the supply chain historical data includes order quantity historical data.

[0015] In some embodiments, the forecasting method also includes: obtaining an updated sales forecast value; determining a collaborative forecast value for a time period corresponding to the sales forecast value in a collaborative data forecast sequence; and determining an updated supply chain forecast value according to a predetermined function based on the updated sales forecast value and the collaborative forecast value.

[0016] In some embodiments, the prediction method also includes: based on the XGBOOST (eXtreme Gradient Boosting) model and the LSTM (Long Short-Term Memory) network deep learning model, a predetermined first number of gradient boosting trees are trained with the MSE (Mean Squared Error) between the true value and the predicted value as the loss function to obtain a first machine learning model.

[0017] In some embodiments, the prediction method further includes: based on the XGBOOST model and the LSTM network deep learning model, training a predetermined second number of gradient boosting trees with the loss function minimizing the MSE between the true value and the predicted value to obtain a second machine learning model.

[0018] According to one aspect of some embodiments of the present disclosure, a forecasting device applied to a supply chain is proposed, comprising: a data acquisition module, configured to acquire sales history data and supply chain history data of the same time period as the sales history data; a data processing module, configured to determine collaborative historical data based on a predetermined function according to the sales history data and the supply chain history data, wherein the predetermined function has a unique solution when the sales history data and the collaborative historical data are determined; a sales forecast data acquisition module, configured to acquire a sales forecast sequence; a collaborative forecast module, configured to acquire a collaborative data forecast sequence according to the sequence of collaborative historical data; and a forecast result determination module, configured to determine a supply chain forecast sequence according to a predetermined function according to the sales forecast sequence and the collaborative data forecast sequence.

[0019] In some embodiments, the sales forecast data acquisition module is further configured to obtain an updated sales forecast value; the prediction result determination module is further configured to: determine the collaborative forecast value for the time period corresponding to the sales forecast value in the collaborative data prediction sequence; and determine the updated supply chain forecast value based on a predetermined function according to the updated sales forecast value and the collaborative forecast value.

[0020] In some embodiments, the prediction device also includes a training module, which is configured to perform at least one of the following: based on the XGBOOST model and the LSTM network deep learning model, a predetermined first number of gradient boosting trees are trained with the loss function of minimizing the MSE between the true value and the predicted value to obtain a first machine learning model, wherein the sales forecast data acquisition module is configured to obtain a sales forecast sequence based on the first machine learning module; or based on the XGBOOST model and the LSTM network deep learning model, a predetermined second number of gradient boosting trees are trained with the loss function of minimizing the MSE between the true value and the predicted value to obtain a second machine learning model, wherein the collaborative prediction module is configured to obtain a supply chain prediction sequence based on the second machine learning module.

[0021] According to one aspect of some embodiments of the present disclosure, a forecasting device applied to a supply chain is proposed, comprising: a memory; and a processor coupled to the memory, wherein the processor is configured to execute any one of the above forecasting methods applied to a supply chain based on instructions stored in the memory.

[0022] According to one aspect of some embodiments of the present disclosure, a non-transitory computer-readable storage medium is provided, on which computer program instructions are stored, and when the instructions are executed by a processor, the steps of any one of the above-mentioned forecasting methods applied to a supply chain are implemented.

[0023] According to one aspect of some embodiments of the present disclosure, a computer program product is provided, including a computer program or instructions, which implement the steps of any one of the above-mentioned forecasting methods applied to a supply chain when the computer program or instructions are executed by a processor. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The drawings described herein are used to provide a further understanding of the present disclosure and constitute a part of the present disclosure. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation on the present disclosure. In the drawings:

[0025] Figure 1 The present invention is a flowchart of some embodiments of the forecasting method applied to the supply chain.

[0026] Figure 2 Flow charts of other embodiments of the forecasting method applied to a supply chain according to the present disclosure.

[0027] Figure 3 Flowcharts of some further embodiments of the forecasting method applied to a supply chain according to the present disclosure.

[0028] Figure 4 It is a schematic diagram of some embodiments of the forecasting device applied to the supply chain of the present disclosure.

[0029] Figure 5 Schematic diagrams of other embodiments of the forecasting device applied to a supply chain according to the present disclosure.

[0030] Figure 6 Schematic diagrams of some further embodiments of the forecasting device applied to a supply chain according to the present disclosure. DETAILED DESCRIPTION

[0031] The technical solution of the present disclosure is further described in detail below through the accompanying drawings and embodiments.

[0032] As the market environment continues to change, the logistics department is unable to cope with future sales fluctuations by preparing warehousing and transportation resources based only on historical order and cargo information. The data isolation and information gap between sales plans and logistics plans lead to imperfect collaborative transparency in the supply chain.

[0033] If the warehousing and fulfillment capacity resources do not have a reasonable order quantity plan as input, the warehousing and fulfillment capacity will not match the transportation and logistics capacity. If the warehousing capacity is insufficient, the goods in the order cannot be prepared in time, which will delay the delivery time, resulting in consumers not receiving the goods in time, resulting in a decline in user experience and customer complaints; if too many warehousing and transportation resources are prepared, it will cause a waste of resources.

[0034] The logistics department only makes predictions based on historical order and cargo information, without sufficient reference to external factors, and is unable to capture changes in external factors. Although factors such as changes in promotions and market demand will not directly affect logistics and transportation, they will be transmitted to the logistics link through changes in sales plans, resulting in a large deviation between the predicted value and the actual value.

[0035] In response to the above problems, the present disclosure proposes a forecasting method, device and storage medium applied to the supply chain to solve the problem of large deviation of the forecast values ​​applied to the supply chain and improve the accuracy of the forecast values ​​applied to the supply chain.

[0036] The flowcharts of some embodiments of the forecasting method of the present disclosure applied to the supply chain are as follows: Figure 1 shown.

[0037] In step S12, the sales history data and the supply chain history data of the same time period as the sales history data are obtained.

[0038] In some embodiments, the sales history data can be obtained by counting the order information. In some embodiments, the historical order information is obtained, and the types of goods and the sales volume of goods in the historical order information are counted to obtain the sales history data. In some embodiments, the types of goods can be determined in units of SKU, and each SKU can be counted separately to improve the accuracy of subsequent predictions. In some embodiments, a predetermined time granularity can be set as needed, and multiple time periods can be divided according to the predetermined time granularity, and the historical order information in the same time period is counted to obtain the sales history data in the time period. In some embodiments, the sales history data obtained in different time periods are arranged in chronological order to obtain a sequence of sales history data.

[0039] In some embodiments, the supply chain historical data includes order volume historical data. In some embodiments, the order volume historical data can be obtained by counting order information. In some embodiments, a predetermined time granularity can be set as needed, multiple time periods can be divided according to the predetermined time granularity, the number of orders in the same time period can be counted, and the order volume historical data in the time period can be obtained. In some embodiments, the order volume historical data obtained in different time periods are arranged in order of precedence to obtain a sequence of the order volume historical data.

[0040] In some embodiments, the supply chain historical data includes historical data on cargo carrying burden. In some embodiments, the cargo carrying burden includes the size of the cargo, such as the length, width, height, or volume of the cargo. Predictions from the perspective of cargo size can provide reference data for storage space preparation for transportation and warehousing processes, thereby improving the accuracy of supply chain resource preparation. In some embodiments, the cargo carrying burden includes the weight of the cargo. Predictions from the perspective of cargo weight can provide reference data for the number and type of transportation vehicles for transportation, thereby improving the accuracy of supply chain resource preparation. In some embodiments, the cargo carrying burden includes the weight and size of the cargo, thereby providing reference data for the supply chain from the two dimensions of space and weight, thereby further improving the accuracy of supply chain resource preparation.

[0041] In some embodiments, the historical data of the cargo transportation burden can be obtained by statistics of order information. In some embodiments, historical order information is obtained, the cargo and the quantity of cargo in the order are determined, and at least one of the weight information or size information of each cargo in the order is extracted. In some embodiments, the weight information of the cargo is the weight after the cargo is packaged, and the size information includes the size information of the packaging box of the cargo, thereby improving the accuracy of the influence of the obtained historical data of the cargo transportation burden on warehousing and logistics. In some embodiments, the historical data of the cargo transportation burden is obtained by statistics of at least one of the weight or size information of the cargo carried in the order. In some embodiments, a predetermined time granularity can be set as needed, and multiple time periods can be divided according to the predetermined time granularity, and the historical order information in the same time period can be counted to obtain the historical data of the cargo transportation burden in the time period. In some embodiments, the historical data of the cargo transportation burden obtained in different time periods are arranged in order of chronological order to obtain a sequence of the historical data of the cargo transportation burden.

[0042] In some embodiments, order information can be obtained based on historical data, and the order information includes the order generation time, the type and quantity of goods in the order, and then the order volume in the same time period, as well as the type and quantity of goods in the order, are counted based on the predetermined time granularity and the order generation time, to obtain the sequence of order volume historical data and the sequence of sales volume historical data. In this way, the sequence of order volume historical data and the sequence of sales volume historical data can be obtained synchronously, thereby improving data acquisition efficiency.

[0043] In some embodiments, order information can be obtained based on historical data, and the types and quantities of goods in the orders in the same time period and the cargo carrying load in the orders can be counted based on the predetermined time granularity and the order generation time, to obtain the sequence of sales history data and the sequence of cargo carrying load history data. In this way, the sequence of cargo carrying load history data and the sequence of sales history data can be obtained synchronously, thereby improving data acquisition efficiency.

[0044] In some embodiments, order information can be obtained based on historical data, and the order information includes the order generation time, the type and quantity of goods in the order, and the cargo carrying burden in the order. Then, according to the predetermined time granularity and the order generation time, the order volume in the same time period is counted, the type and quantity of goods in the order, and the cargo carrying burden in the order are counted, and the sequence of order volume historical data, the sequence of sales volume historical data, and the sequence of cargo carrying burden historical data are obtained. Through such a method, the sequence of order volume historical data, the sequence of sales volume historical data, and the sequence of cargo carrying burden historical data can be obtained simultaneously, while improving the efficiency of data acquisition, predicting supply chain demand in terms of order volume and cargo carrying burden, improving the reference value, and helping to improve the accuracy of preparing warehousing and transportation resources.

[0045] In some embodiments, after obtaining the order information, preprocessing can also be performed. In some embodiments, preprocessing includes removing abnormal values, such as setting a normal range interval to remove order information that exceeds the normal range, or setting a predetermined abnormal condition to determine the need to remove the corresponding order information when the predetermined abnormal condition is met. Through such a method, it is possible to avoid the impact of abnormal data on subsequent predictions and improve the accuracy of predictions. In some embodiments, preprocessing includes filling missing values, such as using the mean of the same type of data as the missing value within a predetermined time length before and after, or using the mean of the same type of data as the missing value in all historical data to fill the missing value, while improving the data integrity, reducing the amount of additional information introduced by the filled data, avoiding the impact on subsequent predictions, and improving the accuracy of predictions.

[0046] In step S13, the collaborative historical data is determined based on a predetermined function according to the sales historical data and the supply chain historical data.

[0047] In some embodiments, the predetermined function has a unique solution when the sales history data and the collaborative history data are determined. The predetermined function is expressed as X=F(a,m), where a is the sales history data, m is the supply chain history data, and X is the collaborative history data. When X and a are determined, a unique m value can be obtained according to X=F(a,m). Based on the method in this embodiment, the sales volume and supply chain information can be integrated through the predetermined function, and the subsequent analysis of the supply chain forecast results is convenient, while improving the forecast accuracy and avoiding affecting the processing efficiency.

[0048] In some embodiments, the predetermined function may be: collaborative historical data = supply chain historical data ÷ sales historical data. The collaborative historical data obtained based on the predetermined function can simultaneously reflect the change trend of supply chain historical data and sales historical data, thereby improving the reference value of collaborative historical data; the computational complexity is low, which is conducive to improving computational efficiency.

[0049] In step S14, a sales forecast sequence is obtained. In some embodiments, the method of step S14 can be performed at any time point before step S16.

[0050] In some embodiments, the sales forecast sequence can be read through an information interaction interface with other parties (such as a sales department), and the read sales forecast sequence can be directly used in subsequent steps, thereby reducing the pressure of data processing.

[0051] In some embodiments, the sales forecast sequence can be predicted and determined by itself, thereby improving the matching degree with the collaborative data prediction. In some embodiments, a first machine learning model for predicting sales can be prepared, and the sequence of sales history data can be input into the first machine learning model to obtain the sales forecast sequence. In some embodiments, the first machine learning model is generated by training with sales sample data, thereby ensuring the matching degree between the model and the sales forecast scenario and improving the accuracy of sales forecast.

[0052] In step S15, a collaborative data prediction sequence is obtained according to the sequence of collaborative historical data.

[0053] In some embodiments, a second machine learning model for predicting sales can be prepared, and collaborative historical data can be input into the second machine learning model to obtain a collaborative data prediction sequence. The second machine learning model is generated by training with collaborative sample data. The collaborative sample data is generated based on sales sample data and supply chain sample data based on a predetermined function, thereby ensuring that the model can match the prediction scenario of collaborative processing of sales and supply chain data, thereby improving prediction accuracy.

[0054] In some embodiments, the model architecture of the first machine learning model is the same or similar to that of the second machine learning model, thereby improving the accuracy of the supply chain forecast sequence subsequently parsed based on the collaborative data forecast sequence and the sales forecast sequence.

[0055] In step S16, the supply chain forecast sequence is determined according to the sales forecast sequence and the collaborative data forecast sequence according to the predetermined function. In some embodiments, the inverse operation of the predetermined function can be determined, and the supply chain forecast value is used as an unknown quantity, and the sales forecast value and the collaborative data forecast value are used as known quantities to be substituted into the inverse operation of the predetermined function to obtain the supply chain forecast sequence.

[0056] In some embodiments, if the supply chain history data includes order quantity history data, the supply chain forecast sequence includes order quantity forecast sequence. In some embodiments, if the supply chain history data includes cargo transportation burden history data, the supply chain forecast sequence includes cargo transportation burden forecast sequence.

[0057] Based on the method in the embodiment shown above, sales data and supply chain data can be collaboratively predicted, and then the supply chain prediction results can be determined based on the sales prediction results and the collaborative prediction results, so that the supply chain prediction can be affected by sales changes, and changes in sales plans can be transmitted to the logistics link, reducing the deviation between the prediction results applied to the supply chain and the actual values, thereby improving the prediction accuracy.

[0058] In some embodiments, the forecasting method applied to the supply chain of the present disclosure also includes pre-training a machine learning model. In some embodiments, the trained machine learning model includes the above-mentioned second machine learning model. In some embodiments, based on the XGBOOST model and the LSTM network deep learning model, a predetermined second number of gradient boosting trees are trained with the loss function of minimizing the MSE of the true value and the predicted value to obtain the second machine learning model. Based on the method in this embodiment, the matching degree between the second machine learning model and the collaborative data prediction can be improved, and the reliability of the prediction can be improved.

[0059] In some embodiments, the trained machine learning model includes the above-mentioned first machine learning model. In some embodiments, based on the XGBOOST model and the LSTM network deep learning model, a predetermined second number of gradient boosting trees are trained with the loss function of minimizing the MSE of the true value and the predicted value to obtain a second machine learning model. Based on the method in this embodiment, the matching degree between the first machine learning model and the sales data prediction can be improved, and the reliability of the prediction can be improved.

[0060] In some embodiments, prediction can be made from two dimensions: order quantity and cargo carrying load, and the prediction results of order quantity and cargo carrying load are applied to the supply chain. Figure 2 shown.

[0061] In step 221, the sales volume historical data, as well as the order volume historical data and the cargo transportation load historical data in the same time period as the sales volume historical data are obtained.

[0062] In some embodiments, taking an online shopping platform as an example, firstly, the orders and goods information of users who purchase goods on the online shopping platform are collected, such as user A generates an order on October 6, 2023, and purchases 2 items of SKU 1 and 1 item of SKU 2; user B generates an order on October 6, 2023, and purchases 1 item of SKU 1 and 1 item of SKU 2. The sales history data includes 3 items of SKU 1 and 2 items of SKU 2, and the supply chain history data information includes the weight and volume information of SKU 1 and SKU 2 respectively.

[0063] Furthermore, the above-obtained information is summarized and counted according to the predetermined time granularity. For example, if the predetermined time granularity is 1 day, the sales history data is summarized by day to obtain the sequence values ​​a1, a2, ...a1 sorted by time. T , the order volume historical data is sorted by time sequence values ​​b1, b2, ...b T , the historical data of cargo transport burden is sorted by time, the sequence values ​​c1, c2, ... c T, where T is the number of time periods. In some embodiments, the cargo carrying burden includes the weight and volume of the cargo, and two sequences can be generated: the cargo weight history data sequence c 11 , c 12 , …c 1T , and the cargo size historical data series c 21 , c 22 , …c 2T .

[0064] In some embodiments, the above historical time series data may also be preprocessed, such as removing outliers, filling missing values ​​with the mean, etc., to update the corresponding sequence.

[0065] In step 231, according to the sales volume history data and the order volume history data, the collaborative history data of the sales volume and the order volume is determined based on a predetermined function.

[0066] In some embodiments, taking the predetermined function as collaborative historical data X = supply chain historical data ÷ sales historical data as an example, the collaborative historical data of order quantity is the ratio of order quantity to sales volume, that is, the collaborative historical data of order quantity By calculating the sequence obtained in step 221 item by item, according to the sequence a1, a2, ...a T With the sequence b1, b2, ... b T , we get the collaborative historical data sequence of order quantity θ1, θ2, ..., θ T ,in, i is a positive integer from 1 to T.

[0067] In step 232, according to the sales volume history data and the cargo transportation burden history data, the coordinated historical data of the sales volume and the cargo transportation burden is determined based on a predetermined function.

[0068] In some embodiments, taking the predetermined function as collaborative historical data X = supply chain historical data ÷ sales historical data as an example, the collaborative historical data of cargo transportation burden is the ratio of cargo transportation burden to sales volume, that is, the collaborative historical data of order quantity By calculating the sequence obtained in step 221 item by item, according to the sequence a1, a2, ...a T With the sequence c1, c2, ... c T , we get the collaborative historical data sequence of cargo transport burden γ1, γ2, ..., γ T ,in, i is a positive integer from 1 to T. In some embodiments, the collaborative historical data sequence γ of the cargo carrying burden based on the cargo weight can be calculated respectively. 11 , γ 12 , ..., γ 1T, and the collaborative historical data series of cargo carrying burden based on cargo volume γ 21 , γ 22 , ..., γ 2T .

[0069] In step 241, a sales forecast sequence is obtained. In some embodiments, the sales forecast sequence can be obtained by the method in any of the embodiments of step S14 above.

[0070] In some embodiments, the sequence of sales history data can be input into the first machine learning model to obtain a sales forecast sequence. The first machine learning model is pre-trained using sales sample data. In some embodiments, the sales forecast sequence includes forecast results for a predetermined number of future periods relative to the time point of the latest data in the historical data. In some embodiments, the time length of each period is the same as the predetermined time granularity. For example, the sales forecast sequence is represented by a T+1 , a T+2 , …a T+h , where h is the predetermined number mentioned above.

[0071] In some embodiments, step 241 may be performed at any time before steps 261 and 261 are performed.

[0072] In step 251, the collaborative historical data of sales volume and order volume is input into the third machine learning model to obtain the collaborative data prediction sequence of sales volume and order volume. The third machine learning model is generated by training with the collaborative sample data of sales volume and order volume, and the collaborative sample data of sales volume and order volume is generated based on the sales volume sample data and the order volume sample data based on a predetermined function. In some embodiments, the collaborative data prediction sequence of order volume includes the prediction results of the future predetermined number of periods relative to the time point of the latest data in the historical data. In some embodiments, the time length of each period is the same as the predetermined time granularity. For example, the collaborative data prediction sequence of order volume is expressed as θ T+1 ,θ T+2 , ..., θ T+h , where h is the predetermined number mentioned above.

[0073] In step 252, the collaborative historical data of sales volume and order volume is input into the fourth machine learning model to obtain a collaborative data prediction sequence of sales volume and cargo carrying burden, wherein the fourth machine learning model is generated by training with collaborative sample data of sales volume and cargo carrying burden, and the collaborative sample data of sales volume and cargo carrying burden is generated based on a predetermined function according to the sales volume sample data and the cargo carrying burden sample data. In some embodiments, the collaborative data prediction sequence of cargo carrying burden includes prediction results for a predetermined number of future periods relative to the time point of the latest data in the historical data, and in some embodiments, the time length of each period is the same as the predetermined time granularity. For example, the collaborative data prediction sequence of order volume is represented as γ T+1 , γ T+2 , ..., γ T+h , where h is the predetermined number mentioned above.

[0074] In some embodiments, the fourth machine learning model can be trained for the collaborative sample data of cargo weight and cargo size, respectively, and a collaborative data prediction sequence of cargo weight and a collaborative data prediction sequence of cargo size can be obtained.

[0075] In step 261, an inverse operation of a predetermined function is performed based on the sales volume forecast sequence and the collaborative data forecast sequence of the order quantity to determine the forecast sequence of the order quantity.

[0076] The predetermined function is For example, the inverse operation of the predetermined function is b i =a i θ i ,i=T+1,T+2,…,T+h, substitute the sales volume forecast sequence and the order volume collaborative data forecast sequence into b i =a i θ i , get the order quantity forecast sequence b T+1 ,b T+2 ,…,b T+h .

[0077] In step 262, based on the sales volume forecast sequence and the coordinated data forecast sequence of the cargo transport burden, an inverse operation of a predetermined function is performed to determine the forecast sequence of the cargo transport burden.

[0078] The predetermined function is For example, the inverse operation of the predetermined function is c i =a i γ i ,i=T+1,T+2,…,T+h, substitute the sales volume forecast sequence and the order volume collaborative data forecast sequence into c i =a i γ i , get the order quantity prediction sequence cT+1 ,c T+2 ,…,c T+h In some embodiments, corresponding prediction sequences may be obtained for cargo weight and cargo size, respectively.

[0079] Based on the method in the embodiment shown above, sales data can be collaboratively predicted with supply chain data including order quantity and cargo transportation load, so that order quantity and cargo transportation load forecasts can be affected by sales changes, and changes in sales plans can be transmitted to the logistics link, reducing the deviation between the predicted results of order quantity and cargo transportation load and the actual value, thereby improving the prediction accuracy.

[0080] In some embodiments, the forecasting method applied to the supply chain of the present disclosure also includes applying the supply chain forecast sequence to the formulation of transportation and warehousing scheduling, guiding the rational allocation of human and material resources, achieving cost savings and efficiency improvements; reducing the problems of insufficient fulfillment capacity and waste of resources, and improving the operational efficiency of the supply chain and user satisfaction.

[0081] In some embodiments, Figure 2 As shown, the forecasting method for supply chain disclosed in the present invention also includes step 211: training the machine learning model. The trained machine learning model may include the third machine learning model mentioned above. In some embodiments, based on the XGBOOST model and the LSTM network deep learning model, a predetermined third number of gradient boosting trees are trained with the loss function of minimizing the MSE of the true value and the predicted value to obtain the third machine learning model. Based on the method in this embodiment, the matching degree between the third machine learning model and the collaborative data forecast of the order quantity can be improved, and the reliability of the forecast can be improved.

[0082] In some embodiments, the trained machine learning model may include the fourth machine learning model described above. In some embodiments, based on the XGBOOST model and the LSTM network deep learning model, a predetermined fourth number of gradient boosting trees are trained with the loss function of minimizing the MSE of the true value and the predicted value to obtain the fourth machine learning model. Based on the method in this embodiment, the matching degree between the fourth machine learning model and the collaborative data prediction of the cargo carrying burden can be improved, and the reliability of the prediction can be improved. In some embodiments, the fourth machine learning model may include two, one of which is trained with data based on cargo weight for collaborative data prediction of cargo weight, and the other is trained with data based on cargo size for collaborative data prediction of cargo size, thereby further improving the specificity and reliability of the model.

[0083] In some embodiments, the trained machine learning model may include the above-mentioned first machine learning model. In some embodiments, based on the XGBOOST model and the LSTM network deep learning model, a predetermined second number of gradient boosting trees are trained with the loss function of minimizing the MSE of the true value and the predicted value to obtain a second machine learning model. Based on the method in this embodiment, the matching degree between the first machine learning model and the sales data forecast can be improved, and the reliability of the forecast can be improved; in addition, such a method can make the models for the sales forecast and the collaborative data forecast similar, and then make the deviations of the two similar, thereby improving the accuracy of the supply chain forecast values ​​obtained subsequently.

[0084] In some embodiments, when the subsequent sales forecast results change, such as when the sales forecast value is updated when the sales forecast is made separately, or when the sales forecast value is adjusted according to the plan, or when the updated sales forecast value is obtained from other interfaces, the supply chain forecast value is updated accordingly. Figure 3 shown.

[0085] In step S37, the updated sales forecast value is obtained.

[0086] In step S38, the collaborative prediction value for the time period corresponding to the sales forecast value is determined in the collaborative data prediction sequence.

[0087] In step S39, an updated supply chain forecast value is determined according to a predetermined function based on the updated sales forecast value and the collaborative forecast value.

[0088] The predictions used in the supply chain in related technologies usually require offline training and execution, which makes it difficult to predict the logistics order quantity and cargo quantity in real time, making it difficult to respond to emergencies or changing demands in a timely manner. When the upstream sales demand plan changes and is adjusted, and the historical order quantity and cargo quantity data has not changed, retraining the historical order quantity and cargo quantity prediction model cannot take into account the factors of upstream commodity sales changes, and manual intervention and adjustment of the logistics order quantity and cargo quantity prediction results are required, resulting in reduced supply chain automation efficiency.

[0089] Based on the method in the above embodiment, when the sales forecast value changes, there is no need to adjust the collaborative forecast value by re-forecasting or the like. Instead, the supply chain forecast value is directly calculated based on a predetermined function and previous collaborative forecast values ​​(such as the collaborative forecast values ​​obtained in the above steps S16, S261, and S262). Since the sales quantity forecast is consistent with the changing trend and law of historical sales before and after the update, and the forecast of each supply chain forecast value is also consistent with the changing law, the adjustment of the sales quantity has little effect on the magnitude of the change. Therefore, it is possible to further improve the accuracy of the forecast value applied to the supply chain and respond to emergencies or changing needs in a timely manner, avoid causing computational burden and improve data acquisition efficiency.

[0090] Compared with directly predicting the supply chain forecast value, the prediction method in the embodiment of the present disclosure provides a more refined modeling process, which can effectively utilize the information of other links in the supply chain, mine through artificial intelligence algorithms, and utilize the potential correlation and influencing factors between the sales quantity scale and the logistics order quantity and cargo quantity scale, so as to improve the coordination between the sales quantity forecast and the logistics order quantity and transportation burden forecast, and achieve the accuracy of the logistics order quantity and transportation burden forecast, which is beneficial to reducing costs and increasing efficiency of supply chain allocation.

[0091] Schematic diagrams of some embodiments of the forecasting device of the present disclosure applied to supply chain are as follows Figure 4 shown.

[0092] The data acquisition module 42 can acquire the sales history data and the supply chain history data of the same time period as the sales history data. In some embodiments, the data acquisition module 42 can execute the method in any of the above steps S12 or S221 to acquire the sales history data and the supply chain history data of the same time period as the sales history data.

[0093] The data processing module 43 can determine the collaborative historical data based on the predetermined function according to the sales historical data and the supply chain historical data, wherein the predetermined function has a unique solution when the sales historical data and the collaborative historical data are determined. In some embodiments, the data processing module 43 executes the method in any of the embodiments of step S13 above, or steps S231 and 232.

[0094] The sales forecast data acquisition module 44 can acquire the sales forecast sequence. In some embodiments, the sales forecast data acquisition module 44 can execute the method in any of the above step S14 or step S24.

[0095] The collaborative prediction module 45 can obtain the collaborative data prediction sequence according to the sequence of collaborative historical data. In some embodiments, the collaborative prediction module 45 can execute the method in any of the above step S15 or steps S251 and 252.

[0096] The forecast result determination module 46 can determine the supply chain forecast sequence according to the sales forecast sequence and the collaborative data forecast sequence according to a predetermined function. In some embodiments, the collaborative forecast module 45 can execute the method in any of the above step S16 or steps S261 and 262.

[0097] The prediction device in the above-mentioned embodiment of the present disclosure can collaboratively predict sales data and supply chain data, and then determine the supply chain prediction results based on the sales prediction results and the collaborative prediction results, so that the supply chain prediction can be affected by sales changes, and the changes in the sales plan can be transmitted to the logistics link, thereby reducing the deviation between the prediction results applied to the supply chain and the actual value, and improving the prediction accuracy.

[0098] In some embodiments, the sales forecast data acquisition module 44 can also acquire updated sales forecast values. When the forecast results of subsequent sales change, such as updating the sales forecast value when forecasting sales separately, adjusting the sales forecast value according to the plan, or acquiring updated sales forecast values ​​from other interfaces, the supply chain forecast value is updated accordingly. In some embodiments, the sales forecast data acquisition module 44 acquires updated sales forecast values ​​by re-forecasting, acquiring data adjusted by staff, or reading sales forecast values ​​provided by other interfaces.

[0099] The forecast result determination module 46 can also determine the collaborative forecast value for the time period corresponding to the sales forecast value in the collaborative data forecast sequence, and determine the updated supply chain forecast value based on a predetermined function according to the updated sales forecast value and the collaborative forecast value.

[0100] The prediction device in the embodiment of the present disclosure does not need to adjust the collaborative prediction value by re-prediction when the sales forecast value changes. Instead, the supply chain prediction value is directly calculated based on a predetermined function and the previous collaborative prediction value. Since the sales quantity forecast is consistent with the changing trend and law of historical sales before and after the update, and the prediction of each supply chain prediction value is also consistent with the changing law, the adjustment of the sales quantity has little effect on the magnitude of the change. Therefore, it is possible to further improve the accuracy of the prediction value applied to the supply chain and respond to emergencies or changing needs in a timely manner, avoid causing computational burden and improve data acquisition efficiency.

[0101] In some embodiments, Figure 4As shown, the forecasting device for supply chain disclosed in the present invention also includes a training module 41, which can train a machine learning model. In some embodiments, the training module 41 is based on the XGBOOST model and the LSTM network deep learning model, and trains a predetermined first number of gradient boosting trees with the loss function of minimizing the MSE of the true value and the predicted value to obtain a first machine learning model, wherein the sales forecast data acquisition module 44 obtains the sales forecast sequence based on the first machine learning module, thereby improving the matching degree between the first machine learning model and the sales data forecast, and improving the reliability of the forecast.

[0102] In some embodiments, the training module 41 is based on the XGBOOST model and the LSTM network deep learning model, and trains a predetermined second number of gradient boosting trees with the loss function of minimizing the MSE between the true value and the predicted value to obtain a second machine learning model, wherein the collaborative prediction module 45 obtains the supply chain prediction sequence based on the second machine learning module, thereby improving the matching degree between the second machine learning model and the collaborative data prediction, and improving the reliability of the prediction.

[0103] In some embodiments, the training module 41 trains a predetermined third number of gradient boosting trees based on the XGBOOST model and the LSTM network deep learning model with the loss function of minimizing the MSE of the true value and the predicted value to obtain a third machine learning model, wherein the collaborative prediction module 45 obtains the prediction sequence of the weight of the goods in the supply chain prediction sequence based on the third machine learning module, thereby improving the matching degree between the third machine learning model and the collaborative data prediction of the order quantity, and improving the reliability of the prediction. In some embodiments, the training module 41 trains a predetermined fourth number of gradient boosting trees based on the XGBOOST model and the LSTM network deep learning model with the loss function of minimizing the MSE of the true value and the predicted value to obtain a fourth machine learning model, wherein the collaborative prediction module 45 obtains the prediction sequence of the cargo carrying burden in the supply chain prediction sequence based on the fourth machine learning module, thereby improving the matching degree between the fourth machine learning model and the collaborative data prediction of the cargo carrying burden, and improving the reliability of the prediction. In some embodiments, the fourth machine learning model may include two, one of which is trained with data based on cargo weight for collaborative data prediction of cargo weight, and the other is trained with data based on cargo size for collaborative data prediction of cargo size, thereby further improving the specificity and reliability of the model.

[0104] A schematic diagram of a structure of an embodiment of the forecasting device of the present disclosure applied to a supply chain is shown in FIG. Figure 5As shown. The forecasting device applied to the supply chain includes a memory 501 and a processor 502. Wherein: the memory 501 can be a disk, a flash memory or any other non-volatile storage medium. The memory is used to store the instructions in the corresponding embodiment of the forecasting method applied to the supply chain mentioned above. The processor 502 is coupled to the memory 501 and can be implemented as one or more integrated circuits, such as a microprocessor or a microcontroller. The processor 502 is used to execute the instructions stored in the memory, which can improve the accuracy of the forecast value applied to the supply chain.

[0105] In one embodiment, it is also possible to Figure 6 As shown, the forecasting device 600 applied to the supply chain includes a memory 601 and a processor 602. The processor 602 is coupled to the memory 601 via a BUS 603. The forecasting device 600 applied to the supply chain can also be connected to an external storage device 605 via a storage interface 604 to call external data, and can also be connected to a network or another computer system (not shown) via a network interface 606. No detailed description will be given here.

[0106] In this embodiment, the accuracy of the forecast value applied to the supply chain can be improved by storing data instructions in a memory and processing the instructions through a processor.

[0107] In another embodiment, a computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement the steps of the method in the corresponding embodiment of the forecasting method applied to the supply chain. It should be understood by those skilled in the art that the embodiments of the present disclosure may be provided as methods, devices, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present disclosure may take the form of a computer program product implemented on one or more computer-usable non-transitory storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0108] The present disclosure is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present disclosure. It should be understood that each process and / or block in the flowchart and / or block diagram and the combination of processes and / or blocks in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0109] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0110] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0111] So far, the present disclosure has been described in detail. In order to avoid obscuring the concept of the present disclosure, some details known in the art are not described. Based on the above description, those skilled in the art can fully understand how to implement the technical solution disclosed here.

[0112] The method and apparatus of the present disclosure may be implemented in many ways. For example, the method and apparatus of the present disclosure may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present disclosure are not limited to the order specifically described above, unless otherwise specifically stated. In addition, in some embodiments, the present disclosure may also be implemented as a program recorded in a recording medium, which includes machine-readable instructions for implementing the method according to the present disclosure. Therefore, the present disclosure also covers a recording medium storing a program for executing the method according to the present disclosure.

[0113] It should be noted that the terms "first", "second", etc. in the specification, claims, and drawings of the present disclosure are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products, or devices.

[0114] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present disclosure rather than to limit it. Although the present disclosure has been described in detail with reference to the preferred embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present disclosure can still be modified or some technical features can be replaced by equivalents without departing from the spirit of the technical solution of the present disclosure, which should be included in the scope of the technical solution for protection of the present disclosure.

Claims

1. A forecasting method applied to a supply chain, comprising: Acquire historical sales data and historical supply chain data in the same time period as the historical sales data; Determine the collaborative historical data based on the sales historical data and the supply chain historical data based on a predetermined function, wherein the predetermined function has a unique solution when the sales historical data and the collaborative historical data are determined; Get sales forecast sequence; Acquire a collaborative data prediction sequence according to the collaborative historical data sequence; According to the sales forecast sequence and the collaborative data forecast sequence, a supply chain forecast sequence is determined according to the predetermined function.

2. The prediction method according to claim 1, wherein: The supply chain historical data includes at least one of order volume historical data or cargo transportation burden historical data; The supply chain forecast sequence includes at least one of an order quantity forecast sequence or a cargo transportation burden forecast sequence.

3. The prediction method according to claim 1, wherein: The obtaining of sales forecast sequence comprises: The sequence of the sales history data is input into a first machine learning model to obtain the sales forecast sequence, wherein the first machine learning model is generated by training with sales sample data.

4. The prediction method according to claim 1, wherein: The acquiring of the collaborative data prediction sequence according to the collaborative historical data sequence comprises: The collaborative historical data is input into a second machine learning model to obtain the collaborative data prediction sequence, wherein the second machine learning model is generated by training with collaborative sample data, and the collaborative sample data is generated based on a predetermined function according to sales sample data and supply chain sample data.

5. The prediction method according to claim 2, wherein: The determining of collaborative historical data based on a predetermined function according to the sales historical data and the supply chain historical data includes: determining the collaborative historical data of sales volume and order volume based on a predetermined function according to the historical data of sales volume and the historical data of order volume; and The method comprises determining the collaborative historical data of sales volume and cargo transportation burden based on a predetermined function according to the sales volume historical data and cargo transportation burden historical data.

6. The prediction method according to claim 5, wherein: The acquiring of the collaborative data prediction sequence according to the collaborative historical data sequence comprises: Inputting the collaborative historical data of sales volume and order volume into a third machine learning model to obtain a collaborative data prediction sequence of sales volume and order volume, wherein the third machine learning model is generated by training with collaborative sample data of sales volume and order volume, and the collaborative sample data of sales volume and order volume is generated based on a predetermined function according to the sales volume sample data and the order volume sample data; and The coordinated historical data of sales volume and order volume are input into a fourth machine learning model to obtain a prediction sequence of coordinated data of sales volume and cargo carrying burden, wherein the fourth machine learning model is generated by training with coordinated sample data of sales volume and cargo carrying burden, and the coordinated sample data of sales volume and cargo carrying burden is generated based on a predetermined function according to the sales volume sample data and the cargo carrying burden sample data.

7. The prediction method according to claim 6, wherein: Determining the supply chain forecast sequence according to the sales forecast sequence and the collaborative data forecast sequence according to the predetermined function includes: According to the sales volume forecast sequence and the collaborative data forecast sequence of the order quantity, an inverse operation of the predetermined function is performed to determine a forecast sequence of the order quantity; According to the sales volume forecast sequence and the coordinated data forecast sequence of the cargo transport burden, an inverse operation of the predetermined function is performed to determine the forecast sequence of the cargo transport burden.

8. The prediction method according to claim 1, wherein: The acquisition of sales history data and supply chain history data in the same time period as the sales history data includes: Acquire order information based on historical data, the order information including order generation time, type and quantity of goods in the order; According to the predetermined time granularity and the order generation time, the order quantity in the same time period, as well as the types and quantities of goods in the orders are counted to obtain a sequence of the order quantity historical data and a sequence of the sales volume historical data, wherein the supply chain historical data includes the order quantity historical data.

9. The prediction method according to claim 1, wherein: The acquisition of sales history data and supply chain history data in the same time period as the sales history data includes: Acquire order information based on historical data, the order information including order generation time, type and quantity of goods in the order, and cargo carrying load in the order; According to the predetermined time granularity and the order generation time, the types and quantities of goods in the orders in the same time period, as well as the cargo transportation burden in the orders are counted to obtain a sequence of the sales volume historical data and a sequence of the cargo transportation burden historical data, wherein the supply chain historical data includes the order quantity historical data.

10. The prediction method according to claim 1, further comprising: Get the updated sales forecast value; Determining a collaborative prediction value for a time period corresponding to the sales volume prediction value in the collaborative data prediction sequence; Based on the updated sales forecast value and the collaborative forecast value, an updated supply chain forecast value is determined according to the predetermined function.

11. The prediction method according to claim 3, further comprising: Based on the extreme gradient boosting tree XGBOOST model and the long short-term memory LSTM network deep learning model, a predetermined first number of gradient boosting trees are trained with the loss function of minimizing the mean square error MSE between the true value and the predicted value to obtain the first machine learning model.

12. The prediction method according to claim 4, further comprising: Based on the extreme gradient boosting tree XGBOOST model and the long short-term memory LSTM network deep learning model, a predetermined second number of gradient boosting trees are trained with the loss function of minimizing the mean square error MSE between the true value and the predicted value to obtain the second machine learning model.

13. A forecasting device for supply chain, comprising: A data acquisition module, configured to acquire sales history data and supply chain history data in the same time period as the sales history data; A data processing module is configured to determine the collaborative historical data based on a predetermined function according to the sales historical data and the supply chain historical data, wherein the predetermined function has a unique solution when the sales historical data and the collaborative historical data are determined; A sales forecast data acquisition module is configured to acquire a sales forecast sequence; A collaborative prediction module, configured to obtain a collaborative data prediction sequence according to the sequence of collaborative historical data; The forecast result determination module is configured to determine the supply chain forecast sequence according to the sales forecast sequence and the collaborative data forecast sequence and according to the predetermined function.

14. The prediction device according to claim 13, wherein: The sales forecast data acquisition module is further configured to acquire an updated sales forecast value; The prediction result determination module is further configured to: Determining a collaborative prediction value for a time period corresponding to the sales volume prediction value in the collaborative data prediction sequence; According to the updated sales forecast value and the collaborative forecast value, an updated supply chain forecast value is determined based on the predetermined function.

15. The prediction device according to claim 13, further comprising a training module configured to perform at least one of the following: Based on the extreme gradient boosting tree XGBOOST model and the long short-term memory LSTM network deep learning model, a predetermined first number of gradient boosting trees are trained with the loss function of minimizing the mean square error MSE between the true value and the predicted value to obtain the first machine learning model, wherein, The sales forecast data acquisition module is configured to acquire the sales forecast sequence based on the first machine learning module; or Based on the extreme gradient boosting tree XGBOOST model and the long short-term memory LSTM network deep learning model, a predetermined second number of gradient boosting trees are trained with the loss function of minimizing the mean square error MSE between the true value and the predicted value to obtain the second machine learning model, wherein the collaborative prediction module is configured to obtain the supply chain prediction sequence based on the second machine learning module.

16. A forecasting device for supply chain, comprising: Memory; as well as A processor coupled to the memory, the processor being configured to execute the method according to any one of claims 1 to 12 based on instructions stored in the memory.

17. A non-transitory computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, implement the steps of the method according to any one of claims 1 to 12.

18. A computer program product, comprising a computer program or instructions, wherein when the computer program or instructions are executed by a processor, the steps of the method according to any one of claims 1 to 12 are implemented.

Citation Information

Cited By

  • Supply chain collaborative management method and system based on artificial intelligence

    CN119599403A