Warehouse inventory demand prediction method and device, and storage medium
Through cluster analysis and autocorrelation moving average model combined with large language model, the items to be predicted are divided into multiple groups, and through the correction of the impact range model and the demand deviation model, the problem of insufficient prediction accuracy of existing warehouse inventory demand is solved, achieving higher prediction accuracy.
Patent Information
- Application Number
- CN202411716742.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-27
- Publication Date
- 2025-02-25
AI Technical Summary
The existing warehouse inventory demand forecasting methods have insufficient accuracy, mainly due to data quality problems, prediction inaccurate caused by a wide variety of items, and the difficulty in accurately capturing sudden demand under the combined influence of multiple factors.
The clustering analysis algorithm is used to divide the items to be predicted into multiple groups, and the autocorrelation moving average model is determined for each group of items, and the impact range model and the demand deviation model are obtained through large language model training, and the prediction results are corrected to improve accuracy.
Through group prediction and model correction, the accuracy of warehouse inventory demand prediction is improved, and the impact of external events on warehouse demand can be captured more accurately, reducing the problem of prediction inaccuracy.
Smart Images
Figure CN119515262A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of warehouse management, and in particular to a warehouse inventory demand forecasting method, equipment and storage medium. Background Art
[0002] With the rapid development of science and technology, technologies such as big data, artificial intelligence and the Internet of Things are increasingly integrated into all walks of life, and warehouse management has ushered in revolutionary changes. The previous manual management methods are gradually being replaced by intelligent management systems. These systems use artificial intelligence prediction models to reveal the changing patterns of warehouse inventory demand over time through in-depth mining and analysis of massive historical data, thereby providing strong support for companies to predict future warehouse inventory demand. However, this prediction method still shows the problem of insufficient prediction accuracy when dealing with warehouse inventory demand in actual applications, mainly due to:
[0003] 1) Some companies’ data collection, storage and processing procedures are not standardized, resulting in low data quality, which in turn affects the accuracy of predictions.
[0004] 2) There are many kinds of items stored in the warehouse, and the usage requirements of different items vary greatly. For example, different materials may be needed to produce a product, and the usage of each material may vary. Using a single forecasting model to forecast the inventory requirements of all items in the warehouse is prone to inaccurate forecasts.
[0005] 3) Warehouse inventory demand is affected by multiple factors, such as supply chain changes, consumer behavior, etc., which makes warehouse inventory demand diverse and sudden. Existing forecasting methods often find it difficult to accurately capture these sudden demands, resulting in deviations between forecast results and actual demand. Summary of the invention
[0006] The present invention provides a warehouse inventory demand forecasting method, device and storage medium, which are used to solve the problem of insufficient accuracy of warehouse inventory demand forecasting.
[0007] The present invention adopts the following technical solutions:
[0008] On the one hand, the present invention provides a method for forecasting warehouse inventory demand, the method comprising: obtaining warehouse management history data of items to be predicted, dividing the items to be predicted into multiple groups according to the warehouse management history data using a preset clustering analysis algorithm, the warehouse management history data at least including the item entry time and the item exit time; obtaining external event information, the external event information at least including supply chain change information; counting the warehouse management history data of the items to be predicted, determining the warehouse demand history data of the items to be predicted, the warehouse demand history data including multiple time points and the warehouse demand corresponding to each time point; determining the autocorrelation moving average model corresponding to each group of items to be predicted according to the warehouse demand history data of each group of items to be predicted, and determining the warehouse demand forecast data of each group of items to be predicted through the autocorrelation moving average model; obtaining an influence range model and a demand deviation model by training a preset large language model according to the warehouse demand history data, external event information and the autocorrelation moving average model; using the influence range model to predict the affected items corresponding to the external event information; using the demand deviation model to predict the demand deviation corresponding to the affected items, and correcting the warehouse demand forecast data of the affected items; integrating the corrected warehouse demand forecast data of the items to be predicted.
[0009] In a feasible implementation, training the impact range model and the demand deviation model includes: extracting multiple historical external event data from external event information as impact range training data; constructing a training data set and a verification data set based on the impact range training data; setting fine-tuning parameters, and using the training data set to fine-tune a preset large language model; after fine-tuning is completed, using the verification data set to verify the accuracy of the affected items predicted by the fine-tuned large language model; if the accuracy does not reach a preset accuracy threshold, adjusting the fine-tuning parameters or reconstructing the training data set, and re-fine-tuning; if the accuracy reaches a preset accuracy threshold or the fine-tuning reaches a preset number of times, stopping the fine-tuning to obtain the impact range model.
[0010] In a feasible implementation, after obtaining the impact range model, the method further includes: based on external event information, using the impact range model to determine the affected items corresponding to the historical external event data at multiple time points; based on the historical warehouse demand data, using the autocorrelation moving average model to determine the actual demand deviation of the affected items after each time point; based on the affected items corresponding to the historical external event data determined by the impact range model, and the actual demand deviation, constructing a demand deviation training data set, and using the demand deviation training data set to fine-tune a preset large language model to obtain a demand deviation model.
[0011] In a feasible implementation, the demand deviation of the affected items after each time point is determined, including: for the affected items, using the autocorrelation moving average model, predicting the warehouse demand after each time point through the historical warehouse demand data before each time point, to obtain warehouse demand prediction data; and determining the actual demand deviation of the affected items through the warehouse demand prediction data and the actual warehouse demand after each time point.
[0012] In a feasible implementation, the items to be predicted are divided into multiple groups, including: counting the outbound data of the items to be predicted, determining the average monthly outbound volume of the items to be predicted, and determining the liquidity index of the items to be predicted based on the average monthly outbound volume; counting the inbound data of the items to be predicted, determining the average interval time between the last batch of inbound items and the next batch of inbound items, and determining the periodicity index of the items to be predicted based on the average interval time; calculating the standard deviation of the outbound quantity of the items to be predicted in multiple time periods, and determining the regularity index of the items to be predicted; and using a preset clustering analysis algorithm based on the liquidity index, periodicity index and regularity index, the items to be predicted are divided into multiple groups.
[0013] In a feasible implementation, a warehouse inventory demand forecasting method also includes: generating unique corresponding RFID tag data for the item to be predicted before the item to be predicted enters the warehouse; obtaining the entry data and exit data of the item to be predicted based on the RFID tag data; integrating the entry data and exit data of the item to be predicted to obtain the warehouse management history data of the item to be predicted.
[0014] In a feasible implementation, obtaining external event information includes: obtaining an API interface or middleware of an enterprise management system; using the API interface or middleware to obtain external event information; or, obtaining an API interface of a third-party organization, using the API interface of the third-party organization to obtain enterprise announcement information contained in the third-party organization, and filtering out external event information based on the enterprise announcement information.
[0015] In a feasible implementation, a warehouse inventory demand forecasting method further includes: after obtaining the warehouse management history data and external event information of the items to be forecasted, storing the warehouse management history data and external event information of the items to be forecasted in a distributed file management system; using preset data cleaning rules to clean the warehouse management history data of the items to be forecasted, and converting the external event information into a plain text format.
[0016] In a second aspect, the present invention also provides a warehouse inventory demand forecasting device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor so that the at least one processor can execute a warehouse inventory demand forecasting method as described in any of the above embodiments.
[0017] In a third aspect, the present invention further provides a non-volatile computer storage medium storing computer executable instructions, wherein the computer executable instructions are configured to be able to execute a warehouse inventory demand forecasting method as described in any of the above embodiments.
[0018] The present invention provides a warehouse inventory demand forecasting method, device and storage medium, which have the following beneficial effects:
[0019] 1. The present invention uses a cluster analysis algorithm to divide the items to be predicted into multiple groups, which can ensure that the items to be predicted in each group have relatively similar warehouse demand characteristics. According to the warehouse demand historical data of each group of items to be predicted, the autocorrelation moving average model corresponding to each group of items to be predicted is determined, which can make the autocorrelation moving average model more targeted, that is, different autocorrelation orders and moving average orders are determined for the warehouse demand characteristics of different groups of items to be predicted, so as to make accurate predictions and avoid the prediction inaccuracy caused by the use of a single prediction model. By obtaining external data, training to obtain a range influence model and a demand deviation model, and using the above model to correct the prediction results of the autocorrelation moving average model, the impact of external data on warehouse inventory demand can be considered in the process of warehouse inventory demand prediction, the sudden changes in warehouse inventory demand can be captured, and the impact of external events on the warehouse demand of the affected items can be determined, thereby further improving the accuracy of warehouse inventory demand prediction.
[0020] 2. After training the influence range model, the present invention uses the influence range model and the autocorrelation moving average model to construct a demand deviation training data set based on the historical external event data and the historical warehouse demand data. A massive demand deviation training data set can be quickly constructed. Through these massive demand deviation training data sets, the demand deviation model can learn the impact of various external event data on warehouse inventory demand from the historical prediction deviation, thereby improving the training effect of the demand deviation model.
[0021] 3. Before the predicted items are put into storage, unique corresponding RFID tag data is generated for the predicted items. The data such as the incoming and outgoing data of the items in the warehouse are recorded according to the above RFID tag data, which can improve the accuracy of the collected incoming and outgoing data. The warehouse management historical data generated according to the above incoming and outgoing data, as well as the external event information are stored in the distributed file management system, which is conducive to the unified management of various data needed in the warehouse inventory demand forecasting, and can ensure the quality of the above various data. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:
[0023] Figure 1 A method flow chart of a warehouse inventory demand forecasting method provided by the present invention;
[0024] Figure 2 A schematic diagram of the structure of a warehouse inventory demand forecasting device provided by the present invention. DETAILED DESCRIPTION
[0025] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.
[0026] The present invention provides a warehouse inventory demand forecasting method, device and storage medium, which solves the technical problem of insufficient accuracy of warehouse inventory demand forecasting by classifying the items to be predicted and then forecasting the warehouse demand, and using a large language model to predict the impact of external events on warehouse demand, and then correcting the warehouse demand forecast value.
[0027] The method of the present invention is described in detail below with reference to the accompanying drawings.
[0028] Figure 1 A method flow chart of a warehouse inventory demand forecasting method provided by the present invention, such as Figure 1 As shown, the method in the present invention at least includes the following execution steps:
[0029] Step 101: Obtain the warehouse management history data of the items to be predicted, and divide the items to be predicted into multiple groups using a preset cluster analysis algorithm based on the warehouse management history data.
[0030] The cluster analysis algorithm is an unsupervised learning method that aims to divide the samples in a data set into several non-overlapping subsets (ie, "groups" in the present invention) so that the samples in the same group are as similar as possible, while the samples in different groups are as different as possible.
[0031] Specifically, first obtain the warehouse management history data of the items to be predicted, and determine the multiple indicators for cluster analysis based on the warehouse entry data and warehouse exit data of each item to be predicted in the warehouse management history data, specifically: count the warehouse exit data of the items to be predicted, determine the average monthly warehouse exit volume of the items to be predicted, and determine the liquidity index of the items to be predicted based on the average monthly warehouse exit volume; count the warehouse entry data of the items to be predicted, determine the average interval time between the last batch of warehouse entry and the next batch of warehouse entry of the items to be predicted, and determine the periodicity index of the items to be predicted based on the average interval time; calculate the standard deviation of the number of warehouse exits of the items to be predicted in multiple time periods, and determine the regularity index of the items to be predicted. After obtaining the above multiple indicators, the items to be predicted are divided into multiple groups using a cluster analysis algorithm such as the k-means algorithm based on these indicators. Using a clustering algorithm to divide the items to be predicted into multiple groups can make the items in the same group have the same warehouse demand model as much as possible, while the warehouse demand models between the items in different groups are as different as possible. On this basis, the corresponding autocorrelation moving average model can be determined for each group in a targeted manner.
[0032] Step 102: Obtain external event information.
[0033] External events refer to events that occur outside the warehouse that may affect warehouse inventory demand, such as supply chain changes, product promotions, abnormal weather and other information.
[0034] Specifically, external event information can be obtained from the enterprise management system of the enterprise through an API interface or middleware. It is also possible to obtain enterprise announcement information contained in a third-party organization through the API interface of the third-party organization, and then filter out external event information from the above announcement information. The above third-party organization can be a stock exchange, that is, the announcement information of listed companies in the stock exchange is obtained through the API interface of the stock exchange, and external event information including supplier announcement information is filtered out from these announcement information. It is also possible to obtain weather information from a weather forecast website, and then filter out abnormal weather information from the above weather information.
[0035] Step 103: Count the warehouse management history data of the items to be predicted, and determine the warehouse demand history data of the items to be predicted.
[0036] The warehouse management historical data here includes the data of each specific item's outbound and inbound storage. The warehouse demand historical data of the items to be predicted refers to the quantity of a type of goods that needs to be stored in the warehouse at different times, which is usually related to the outbound volume. For example, when the outbound volume is high, the corresponding warehouse demand is higher.
[0037] Specifically, firstly, the average monthly outbound volume of the items to be predicted in the past months is counted, and then the average monthly outbound volume is multiplied by a preset coefficient to obtain the historical data of the storage demand of the items to be predicted.
[0038] Step 104: Determine the autocorrelation moving average model corresponding to each group of items to be predicted based on the historical warehouse demand data of each group of items to be predicted, and determine the warehouse demand forecast data of each group of items to be predicted through the autocorrelation moving average model.
[0039] The Autoregressive Moving Average Model (ARMA model) is a time series forecasting model that combines the characteristics of the autoregressive (AR) and moving average (MA) models. The main purpose of the ARMA model is to capture the autocorrelation of time series data and predict future values by fitting historical data.
[0040] Specifically, first check whether the historical data of warehouse demand for each group of items to be predicted is stable. If not, perform differential processing, that is, calculate the difference between adjacent items in the time series until the data becomes stable. Then draw the autocorrelation function graph, and determine the autocorrelation order according to the position of the tail or truncation in the autocorrelation function graph. Then draw the partial autocorrelation function graph, and determine the moving average order according to the truncation position in the above partial autocorrelation function graph. After determining the autocorrelation order and the moving average order, the autocorrelation moving average model corresponding to each group of items to be predicted is determined. Finally, the autocorrelation moving average model corresponding to each group is used to predict the warehouse demand of each group of items to be predicted.
[0041] Step 105: According to the warehouse demand history data, external event information and the autocorrelation moving average model, the influence range model and the demand deviation model are obtained by training the preset large language model.
[0042] The impact range model is used to predict the items to be predicted that may be affected by the external events based on the external event information, that is, which items to be predicted may have their warehouse demand affected after the external events occur. The demand deviation model is used to predict the specific impact of external events on the affected items to be predicted, that is, the specific amount of change in warehouse demand for the affected items after the external events occur.
[0043] Specifically, first, multiple historical external event data are extracted from the external event information as the impact range training data, and a training data set and a validation data set are constructed based on the above impact range data. The above data set mainly includes the specific content of the external event that occurred, and the corresponding affected items to be predicted. Usually, a set of data is formed in the form of questions and answers, and multiple sets of question and answer data constitute a data set. For example: Q: The main downstream distributors issued a corporate announcement indicating that they will pay more attention to inventory turnover and optimize inventory management. According to this announcement, which products of our company will be affected in terms of warehouse demand? A: Product F: The sales cycle of this product is long, and distributors may reduce the inventory of this product. After the training data set and validation data set are constructed, the preset large language model is fine-tuned using the training data set. The fine-tuning here refers to the process of further training and adjusting parameters of the large language model after pre-training using labeled data for specific tasks. This process aims to make the model better adapt to the target task and improve its performance in specific tasks or fields. After fine-tuning is completed, the validation data set is used to verify the accuracy of the affected items predicted by the fine-tuned large language model. If the accuracy does not reach the preset accuracy threshold, the fine-tuning parameters are adjusted or the training data set is reconstructed and fine-tuning is performed again. If the preset accuracy threshold is reached, the fine-tuning is stopped to obtain the impact range model.
[0044] After obtaining the impact range model, according to the external event information, the trained impact range model is used to determine the affected items corresponding to the historical external event data at multiple time points. That is, multiple data records corresponding to the event occurrence events are screened out from the historical external event data, and the trained impact range model is used to determine the affected items corresponding to the above-screened historical external event data. Then, according to the warehouse demand historical data, the autocorrelation moving average model is used to determine the actual demand deviation of the affected items after each time point. Then, according to the above-screened historical external data, the affected items corresponding to the screened historical external data, and the actual demand deviation, a demand deviation training data set is constructed, and the preset large language model is fine-tuned using the above-mentioned demand deviation training data set to obtain a demand deviation model. Using the impact range model and the autocorrelation moving average model to construct a data set for the demand deviation model, a large amount of high-quality training data can be constructed in a short time to ensure that the demand deviation model can learn the specific impact of various external events on the warehouse demand for the predicted items, which can not only improve the training efficiency of the demand deviation model, but also ensure the high-quality training effect of the demand deviation model.
[0045] In one example, a set of data in the constructed demand deviation training data set is as follows: Question: Based on the following external events, calculate the change in warehouse demand for the affected items. The external event is: Class A sports shoes will be on sale for a limited time at 20% off from October 1 to November 30, 2023; the affected items are: Class A sports shoes; the predicted warehouse demand is: 5,000 pairs of warehouse demand from October 1 to November 30, 2023.
[0046] Answer: The actual warehouse demand for Class A sports shoes from October 1 to November 30, 2023 is 6,000 pairs, with a deviation of 1,000 pairs and a deviation ratio of 20%.
[0047] Step 106: Use the impact range model to predict the affected items corresponding to the external event information.
[0048] Specifically, firstly, the external event information is acquired in real time, then the external event information is converted into input data in an input form that conforms to the input form of the impact range model, and then the prediction result of the affected items is determined according to the output result of the impact range model.
[0049] In one example, external event information in json format is obtained from the API, and the content is: From 2023-10-01 to 2023-10-07, a certain place experienced a sharp drop in temperature, with the lowest temperature reaching minus 5 degrees. First, convert the json format into the corresponding txt format, and then convert the above external information into input data in the form of an impact range model: Determine the items that may be affected based on the following external information. The external information is: From 2023-10-01 to 2023-10-07, a certain place experienced a sharp drop in temperature, with the lowest temperature reaching minus 5 degrees.
[0050] Step 107: Use the demand deviation model to predict the demand deviation corresponding to the affected items, and correct the warehouse demand forecast data of the affected items.
[0051] Specifically, after determining the affected items corresponding to the external event information obtained in real time, the external event information obtained in real time, the corresponding affected items and the warehouse demand predicted by the autocorrelation moving average model are converted into an appropriate format and input into the demand deviation model to determine the prediction result of the demand deviation, and the warehouse demand prediction value of the affected items is corrected according to the prediction result of the demand deviation.
[0052] Step 108: Integrate the revised warehouse demand forecast data of the items to be forecasted.
[0053] Specifically, the revised warehouse demand forecast data of all items to be predicted are counted, and the above-stated data are sorted according to preset rules. For example, the revised warehouse demand forecast data of all predicted items can be sorted in order from large to small according to the size of the warehouse demand.
[0054] In a possible implementation of the present invention, there are many items in the warehouse, and it is easy to make mistakes when manually recording the outbound data and inbound data of the items. Therefore, before the predicted items are put into the warehouse, unique corresponding RFID tag data is generated for the predicted items, and the inbound data and outbound data of the predicted items are obtained according to the RFID tag data, and the inbound data and outbound data of the predicted items are integrated to obtain the warehouse management history data of the predicted items. By using RFID tag data to record the inbound data and outbound data of the items, the accuracy of the inbound data and outbound data records can be guaranteed, and the data quality of the original data in the warehouse can be guaranteed.
[0055] In a possible implementation of the present invention, the warehouse management data and external time information of the items to be predicted are often obtained using multiple methods. If they are stored in different systems, data island effects are likely to occur, affecting the prediction effect of warehouse inventory demand. Therefore, after obtaining the warehouse management history data and external event information of the items to be predicted, the present invention stores the above data in a distributed file management system, and uses preset data cleaning rules to clean the warehouse management history data of the items to be predicted, and converts the external event information into a plain text format. By uniformly storing the above-mentioned multiple data in a distributed file system, the standardization of the data storage process can be guaranteed, which is conducive to improving data quality. At the same time, data cleaning is performed on the warehouse management history data of the items to be predicted, which can eliminate the negative impact of abnormal data on the warehouse inventory demand prediction results and improve the accuracy of the prediction. Converting the external event information into a plain text format can ensure that the external event information can be input into the corresponding large language model.
[0056] Based on the same inventive concept, the present invention also provides a warehouse inventory demand forecasting device, the structure of which is as follows: Figure 2 shown.
[0057] Figure 2 This is a schematic diagram of the structure of a warehouse inventory demand forecasting device provided by the present invention. Figure 2 As shown, a warehouse inventory demand forecasting device 200 in the present invention specifically includes: at least one processor 201; and a memory 203 that is communicatively connected to the at least one processor (connected via a bus 202); wherein the memory 203 stores instructions that can be executed by at least one processor 201, so that at least one processor 201 can execute a warehouse inventory demand forecasting method as described in the above embodiment.
[0058] In one or more possible implementations of the present invention, the aforementioned processor is used to execute, obtain the warehouse management history data of the items to be predicted, and divide the items to be predicted into multiple groups according to the warehouse management history data using a preset clustering analysis algorithm; obtain external event information; count the warehouse management history data of the items to be predicted to determine the warehouse demand history data of the items to be predicted; determine the autocorrelation moving average model corresponding to each group of items to be predicted according to the warehouse demand history data of each group of items to be predicted, and determine the warehouse demand forecast data of each group of items to be predicted through the autocorrelation moving average model; obtain the influence range model and the demand deviation model by training a preset large language model according to the warehouse demand history data, external event information and the autocorrelation moving average model; use the influence range model to predict the affected items corresponding to the external event information; use the demand deviation model to predict the demand deviation corresponding to the affected items, and correct the warehouse demand forecast data of the affected items; integrate the corrected warehouse demand forecast data of the items to be predicted.
[0059] In addition, the present invention also provides a non-volatile computer storage medium storing computer executable instructions, wherein the computer executable instructions are configured to be able to execute a warehouse inventory demand forecasting method as described in any of the above embodiments.
[0060] In one or more possible implementations of the present invention, the aforementioned computer executable instructions are configured to execute, obtain the warehouse management history data of the items to be predicted, and divide the items to be predicted into multiple groups based on the warehouse management history data using a preset clustering analysis algorithm; obtain external event information; count the warehouse management history data of the items to be predicted to determine the warehouse demand history data of the items to be predicted; determine the autocorrelation moving average model corresponding to each group of items to be predicted based on the warehouse demand history data of each group of items to be predicted, and determine the warehouse demand forecast data of each group of items to be predicted through the autocorrelation moving average model; obtain the influence range model and the demand deviation model by training a preset large language model based on the warehouse demand history data, external event information and the autocorrelation moving average model; use the influence range model to predict the affected items corresponding to the external event information; use the demand deviation model to predict the demand deviation corresponding to the affected items, and correct the warehouse demand forecast data of the affected items; integrate the corrected warehouse demand forecast data of the items to be predicted.
[0061] Each embodiment of the present invention is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the IoT device and medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.
[0062] The system and medium provided by the present invention correspond one-to-one to the method, and therefore, the system and medium also have similar beneficial technical effects to the corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the system and medium will not be repeated here.
[0063] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0064] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as 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.
[0065] 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 including 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.
[0066] 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 in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0067] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.
[0068] The above are only embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. A warehouse inventory demand forecasting method, characterized in that: The method comprises: Acquire warehouse management history data of the items to be predicted, and divide the items to be predicted into multiple groups using a preset clustering analysis algorithm according to the warehouse management history data, wherein the warehouse management history data at least includes the time when the items enter the warehouse and the time when the items leave the warehouse; Acquiring external event information, wherein the external event information at least includes supply chain change information; Collecting historical warehouse management data of the items to be predicted to determine historical warehouse demand data of the items to be predicted, wherein the historical warehouse demand data includes multiple time points and warehouse demand quantities corresponding to each time point; Determine the autocorrelation moving average model corresponding to each group of items to be predicted based on the historical warehouse demand data of each group of items to be predicted, and determine the warehouse demand forecast data of each group of items to be predicted through the autocorrelation moving average model; According to the warehouse demand history data, the external event information and the autocorrelation moving average model, an impact range model and a demand deviation model are obtained by training a preset large language model; Using the impact range model, predicting the affected items corresponding to the external event information; Using the demand deviation model to predict the demand deviation corresponding to the affected item, and correcting the warehouse demand forecast data of the affected item; The revised warehouse demand forecast data of the items to be forecasted are integrated.
2. A warehouse inventory demand forecasting method according to claim 1, characterized in that: Training impact range models and demand deviation models, including: Extracting a plurality of historical external event data from the external event information as influence range training data; Constructing a training data set and a validation data set according to the impact range training data; Setting fine-tuning parameters, and using the training data set to fine-tune a preset large language model; After fine-tuning is completed, using the verification dataset, verifying the accuracy of the affected items predicted by the fine-tuned large language model; If the accuracy rate does not reach the preset accuracy rate threshold, the fine-tuning parameters are adjusted or the training data set is rebuilt and fine-tuned again; If the accuracy reaches a preset accuracy threshold or the fine-tuning reaches a preset number of times, the fine-tuning is stopped to obtain an impact range model.
3. A warehouse inventory demand forecasting method according to claim 2, characterized in that: After obtaining the impact range model, the method further includes: According to the external event information, using the impact range model, determine the affected items corresponding to the historical external event data at multiple time points; Based on the historical warehouse demand data, the autocorrelated moving average model is used to determine the actual demand deviation of the affected items after each time point; According to the affected items corresponding to the historical external event data determined by the impact range model and the actual demand deviation, a demand deviation training data set is constructed, and the preset large language model is fine-tuned using the demand deviation training data set to obtain a demand deviation model.
4. A warehouse inventory demand forecasting method according to claim 3, characterized in that: Determine the demand deviation of the affected items after each point in time, including: For the affected items, using the autocorrelation moving average model, the warehouse demand after each time point is predicted through the warehouse demand history data before each time point, to obtain warehouse demand prediction data; The actual demand deviation of the affected items is determined by the warehouse demand forecast data and the actual warehouse demand after each time point.
5. A warehouse inventory demand forecasting method according to claim 1, characterized in that: The items to be predicted are divided into multiple groups, including: Counting the outbound data of the items to be predicted, determining the average monthly outbound volume of the items to be predicted, and determining the liquidity index of the items to be predicted according to the average monthly outbound volume; Collecting data on the incoming goods of the items to be predicted, determining an average interval between the last batch of incoming goods and the next batch of incoming goods of the items to be predicted, and determining a periodicity index of the items to be predicted according to the average interval; Calculate the standard deviation of the quantity of the items to be predicted shipped out in multiple time periods to determine the regularity index of the items to be predicted; According to the liquidity index, the periodicity index and the regularity index, the items to be predicted are divided into multiple groups using a preset cluster analysis algorithm.
6. A warehouse inventory demand forecasting method according to claim 1, characterized in that: The method further comprises: Before the predicted item is put into storage, generating unique corresponding RFID tag data for the predicted item; Acquire the inbound and outbound data of the item to be predicted according to the RFID tag data; The warehouse entry data and the warehouse exit data of the items to be predicted are integrated to obtain the warehouse management history data of the items to be predicted.
7. A warehouse inventory demand forecasting method according to claim 1, characterized in that: Get external event information, including: Obtain the API interface or middleware of the enterprise management system; Using the API interface or middleware to obtain the external event information; Alternatively, an API interface of a third-party organization is obtained, and the enterprise announcement information contained in the third-party organization is obtained using the API interface of the third-party organization, and the external event information is filtered out according to the enterprise announcement information.
8. A warehouse inventory demand forecasting method according to claim 1, characterized in that: The method further comprises: After acquiring the warehouse management history data of the item to be predicted and the external event information, storing the warehouse management history data of the item to be predicted and the external event information in a distributed file management system; The warehouse management historical data of the item to be predicted is cleaned using preset data cleaning rules, and the external event information is converted into a plain text format.
9. A warehouse inventory demand forecasting device, characterized in that: The device includes a processor, a memory, and execution instructions stored in the memory, and the execution instructions are configured to enable the device to execute a warehouse inventory demand forecasting method according to any one of claims 1 to 8 when executed by the processor.
10. A non-volatile computer storage medium storing computer executable instructions, characterized in that: The computer executable instructions are configured to be capable of executing a warehouse inventory demand forecasting method as described in any one of claims 1 to 8.