Intelligent delivery and replenishment management method and system based on multi-dimensional inventory early warning

Abstract

The application provides a kind of intelligent delivery and replenishment management method and system based on multi-dimensional inventory early warning, belong to goods management technical field, this method includes: obtaining the global demand data and full-link supply data of inventory object, and determining demand fluctuation intensity sequence and supply interruption frequency sequence;The demand fluctuation intensity sequence and supply interruption frequency sequence are cross recursive analysis, determine the mismatch risk evolution path between the goods demand and goods supply;According to the mismatch risk evolution path, the adaptive adjustment coefficient when safety stock is determined, and the safety stock threshold curve that evolves smoothly with time is obtained;When the inventory level of inventory object is lower than safety stock threshold curve, generate out-of-stock risk early warning, when inventory level is higher than safety stock threshold curve, generate backlog risk early warning.The technical scheme provided by the application can realize the adaptive management of delivery and replenishment under the nonlinear coupling effect of demand fluctuation and supply interruption.

Description

Technical Field

[0001] This application relates to the field of cargo management technology, and more specifically, to an intelligent shipping and replenishment management method and system based on multi-dimensional inventory early warning. Background Technology

[0002] With the increasing complexity of global supply chains and the rapid development of e-commerce and smart manufacturing, traditional cargo management models that rely on manual recording and basic barcode scanning are no longer sufficient to meet enterprises' requirements for real-time performance, accuracy, and traceability. Currently, cargo management technology is accelerating the integration of emerging technologies such as the Internet of Things, radio frequency identification, cloud computing, and big data analytics to achieve automatic cargo identification, dynamic positioning, inventory early warning, and full-process digital tracking. At the same time, automated warehouses are also driving the management focus to shift from simple static storage to intelligent dynamic scheduling, effectively improving cargo management efficiency.

[0003] In existing inventory management, the first step is to inspect, code, and locate goods for shelving based on purchase or production orders, establishing initial inventory records. The second step follows the "first-in, first-out" principle, picking, verifying, and reducing inventory according to instructions. During storage, physical inventory and system data are verified through periodic and dynamic cyclical inventory checks. Inventory upper and lower limit early warning mechanisms are used to optimize warehouse allocation and replenishment strategies to achieve effective inventory management. However, in existing inventory management methods, the setting of safety stock thresholds is usually based on the static assumption that demand and supply are independent or on linear forecasting models. This fails to characterize the nonlinear coupling effect of demand fluctuations and supply interruptions in time (i.e., a complex, non-proportionally amplified temporal interaction between the intensity of demand fluctuations and the frequency of supply interruptions) and its risk propagation path. This results in a phase lag and amplitude deviation between the safety stock threshold and the actual supply-demand mismatch risk evolution trajectory, making it difficult to achieve dynamic balance management of shipments and replenishments. Therefore, how to achieve adaptive management of shipments and replenishments under the nonlinear coupling effect of demand fluctuations and supply interruptions has become a challenge for the industry. Summary of the Invention

[0004] This application provides an intelligent shipping and replenishment management method and system based on multi-dimensional inventory early warning, which can achieve adaptive management of shipping and replenishment under the nonlinear coupling effect of demand fluctuations and supply interruptions.

[0005] Firstly, this application provides an intelligent shipping and replenishment management method based on multi-dimensional inventory early warning, comprising the following steps: Obtain the full-domain demand data and full-link supply data of the inventory objects, and extract the demand fluctuation intensity sequence from the full-domain demand data and the supply interruption frequency sequence from the full-link supply data; By performing cross-recursive analysis on the demand fluctuation intensity sequence and the supply disruption frequency sequence, the evolution path of the mismatch risk characterizing the demand for goods and the supply of goods can be determined. Based on the path evolution and local curvature characteristics of the mismatch risk evolution path within a set risk window, the adaptive adjustment coefficient of the inventory object when setting safety stock is determined, and the current safety stock threshold is adjusted through the adaptive adjustment coefficient, thereby obtaining a safety stock threshold curve that evolves smoothly over time. The system monitors the relative position of the inventory level of an inventory item to the safety stock threshold curve in real time. When the inventory level is lower than the safety stock threshold curve, a stockout risk warning is generated. When the inventory level is higher than the safety stock threshold curve, a backlog risk warning is generated.

[0006] In some embodiments, obtaining the full-domain demand data and full-chain supply data of inventory objects specifically includes: Obtain time-series demand records for inventory objects from omnichannel sales, pre-sales, and returns, and preprocess these time-series demand records to obtain omnichannel demand data. After obtaining the purchase arrival records, production delivery records, and logistics interruption records of the inventory objects and performing correlation processing, the end-to-end supply chain data is obtained.

[0007] In some embodiments, extracting the demand fluctuation intensity sequence from the global demand data and the supply interruption frequency sequence from the end-to-end supply data specifically includes: The global demand data is sliced ​​according to fixed time windows, and the coefficient of variation of demand within each time window is calculated to obtain the initial demand fluctuation sequence. The initial demand fluctuation sequence is smoothed by an exponentially weighted moving average to obtain a demand fluctuation intensity sequence; The procurement arrival records, production delivery records, and logistics interruption records in the entire supply chain data are correlated to construct a supply fulfillment event set indexed by purchase orders; In the set of supply fulfillment events, based on the planned delivery time, supply fulfillment events where the actual delivery time exceeds a preset tolerance threshold are marked as interruption events; The number of occurrences of the interruption event within each fixed time window is obtained, and the density of the occurrences is estimated to obtain the supply interruption frequency sequence.

[0008] In some embodiments, the global demand data is sliced ​​according to fixed time windows, and the coefficient of variation of demand within each time window is calculated to obtain the initial demand fluctuation sequence, specifically including: The time-series demand records corresponding to the global demand data are sorted by timestamp to form an original demand time series sampled at equal intervals; A sliding window of a preset fixed length is used to traverse the original demand time series with a preset step size, and multiple windowed demand subsequences are extracted in sequence. For each windowed demand subsequence, the ratio of the standard deviation to the mean of the demand for all sampling points within the windowed demand subsequence is calculated as the corresponding coefficient of variation. All the calculated coefficients of variation are arranged in time order of the sliding window to form the initial demand fluctuation sequence.

[0009] In some embodiments, cross-recursive analysis is performed on the demand fluctuation intensity sequence and the supply disruption frequency sequence to determine the evolution path of the mismatch risk characterizing the demand for goods and the supply of goods, specifically including: Phase space reconstruction is performed on the demand fluctuation intensity sequence and the supply interruption frequency sequence to obtain the demand phase space trajectory and the supply phase space trajectory, respectively. Calculate the cross recursion matrix of the demand phase space trajectory and the supply phase space trajectory, where the elements of the cross recursion matrix indicate whether the states of the two phase space trajectories at corresponding times fall into each other's neighborhood. A statistical analysis of the diagonal segment length distribution is performed on the cross recursive matrix, and the quantiles of the diagonal segment length are extracted as a quantitative indicator of mismatch risk. The quantitative indicators of the mismatch risk are arranged in chronological order to obtain the evolution path of the mismatch risk.

[0010] In some embodiments, phase space reconstruction is performed on the demand fluctuation intensity sequence and the supply interruption frequency sequence to obtain the demand phase space trajectory and the supply phase space trajectory, specifically including: The first embedding delay of the demand fluctuation intensity sequence is determined by mutual information method, and the first embedding dimension of the demand fluctuation intensity sequence is determined by spurious nearest neighbor method. Based on the first embedding delay and the first embedding dimension, the demand fluctuation intensity sequence is embedded with delay coordinates to obtain a demand phase space trajectory composed of multiple first delay vectors; The second embedding delay of the supply interruption frequency sequence is determined using the mutual information method, and the second embedding dimension of the supply interruption frequency sequence is determined using the spurious nearest neighbor method. The supply interruption frequency sequence is embedded with delay coordinates according to the second embedding delay and the second embedding dimension to obtain a supply phase space trajectory composed of multiple second delay vectors.

[0011] In some embodiments, determining the adaptive adjustment coefficient of the inventory object when setting up safety stock based on the path evolution trend and local curvature characteristics of the mismatch risk evolution path within a set risk window specifically includes: On the mismatch risk evolution path, a path segment of a preset duration is extracted back from the current moment as a reference to set the risk window; Linear fitting is performed on the path segments within the defined risk window, and the slope of the resulting fitted line is taken as the path evolution direction. Local curvature calculation is performed on the path segments within the defined risk window, and the mean value of the obtained curvature is used as the local curvature feature; The path evolution direction and the local curvature features are input into a preset fuzzy logic controller, and the fuzzy logic controller outputs the adaptive adjustment coefficient of the inventory object when setting up safety stock.

[0012] Secondly, this application provides an intelligent shipping and replenishment management system based on multi-dimensional inventory early warning, used to execute an intelligent shipping and replenishment management method based on multi-dimensional inventory early warning. The system includes: The acquisition module is used to acquire the overall demand data and the entire supply chain data of the inventory objects, and extract the demand fluctuation intensity sequence from the overall demand data and the supply interruption frequency sequence from the entire supply chain data. The processing module is used to perform cross-recursive analysis on the demand fluctuation intensity sequence and the supply interruption frequency sequence, thereby determining the evolution path of the mismatch risk characterizing the demand for goods and the supply of goods. The processing module is further configured to determine the adaptive adjustment coefficient of the inventory object when carrying out safety stock based on the path evolution direction and local curvature characteristics of the mismatch risk evolution path within the set risk window, and adjust the current safety stock threshold through the adaptive adjustment coefficient to obtain a safety stock threshold curve that evolves smoothly over time. The execution module is used to monitor the relative position of the inventory level of the inventory object with the safety stock threshold curve in real time. When the inventory level is lower than the safety stock threshold curve, a stockout risk warning is generated. When the inventory level is higher than the safety stock threshold curve, a backlog risk warning is generated.

[0013] Thirdly, this application provides a computer device, the computer device including a memory and a processor, the memory storing code, the processor being configured to acquire the code and execute the above-described intelligent shipment and replenishment management method based on multi-dimensional inventory early warning.

[0014] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned intelligent shipping and replenishment management method based on multi-dimensional inventory early warning.

[0015] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: The intelligent shipping and replenishment management method and system based on multi-dimensional inventory early warning provided in this application first acquires the full-domain demand data and full-link supply data of the inventory object, and extracts the demand fluctuation intensity sequence from the full-domain demand data and the supply interruption frequency sequence from the full-link supply data; secondly, it performs cross-recursive analysis on the demand fluctuation intensity sequence and the supply interruption frequency sequence to determine the evolution path of the mismatch risk between goods demand and goods supply; then, based on the path evolution trend and local curvature characteristics of the mismatch risk evolution path within a set risk window, it determines the adaptive adjustment coefficient of the inventory object when maintaining safety stock, and adjusts the current safety stock threshold through the adaptive adjustment coefficient to obtain a safety stock threshold curve that evolves smoothly over time; finally, it monitors the relative position of the inventory level of the inventory object with the safety stock threshold curve in real time, and generates a stockout risk warning when the inventory level is lower than the safety stock threshold curve, and generates a backlog risk warning when the inventory level is higher than the safety stock threshold curve.

[0016] Therefore, this application can achieve adaptive management of shipments and replenishment under the nonlinear coupling effect of demand fluctuations and supply disruptions. First, by integrating full-domain demand data from all channels (sales, pre-sales, and returns) and full-chain supply data from procurement, production, and logistics disruptions, the application extracts demand fluctuation intensity sequences and supply disruption frequency sequences respectively. This eliminates the overestimation of demand scale and the omission of supply risks from a single data source, allowing the uncertainties on both the supply and demand sides to be independently quantified in the form of time-series indicators, providing comparable dual-source inputs for subsequent cross-analysis. Second, by using cross-recursive analysis to map the demand fluctuation intensity sequence and the supply disruption frequency sequence to the same phase space for coupling analysis, this application can identify the dynamic mismatch process of demand fluctuations and supply disruptions moving from synchronization to divergence in the time dimension, and quantify this process as a continuous mismatch risk evolution path. This avoids the shortcomings of traditional methods that treat supply and demand risks as static independent variables and fail to capture the superposition effect of the two. Then, by extracting the mismatch risk evolution path in... This paper defines the evolutionary trajectory and local curvature characteristics within a risk window, and uses an adaptive adjustment coefficient output by a fuzzy logic controller to correct the safety stock threshold. This allows the adjustment of the safety stock to respond to both the changing direction of risk and the acceleration of risk changes, enabling the adjustment of the inventory buffer level in advance before the critical turning point of supply-demand mismatch. This achieves dynamic safety stock decision-making with proactive response capabilities, avoiding management errors caused by phase lag and amplitude deviation between the static safety stock threshold and the actual evolution trajectory of supply-demand mismatch risk. Finally, a smooth evolutionary safety stock threshold curve is generated based on B-spline interpolation, and the relative position of the real-time inventory level to this curve is used as the early warning criterion to output the corresponding risk warning. This avoids repeated oscillations of the warning signal caused by static discrete threshold jumps, and simultaneously improves the timeliness and stability of inventory risk assessment, thereby achieving dynamic balance management of shipments and replenishment. In summary, the technical solution provided in this application can achieve adaptive management of shipments and replenishment under the nonlinear coupling effect of demand fluctuations and supply interruptions. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating an intelligent shipping and replenishment management method based on multi-dimensional inventory early warning, according to some embodiments of this application. Figure 2 Exemplary flowchart of an intelligent shipping and replenishment management method based on multi-dimensional inventory early warning, as shown in some embodiments of this application; Figure 3 This is an exemplary flowchart illustrating the determination of the mismatch risk evolution path according to some embodiments of this application; Figure 4 This is a schematic diagram of the structure of an intelligent shipping and replenishment management system based on multi-dimensional inventory early warning, as shown in some embodiments of this application; Figure 5This is a schematic diagram of the structure of a computer device that implements an intelligent shipping and replenishment management method based on multi-dimensional inventory early warning, according to some embodiments of this application. Detailed Implementation

[0018] To better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] refer to Figure 1 This diagram is a flowchart of an intelligent shipping and replenishment management method based on multi-dimensional inventory early warning, according to some embodiments of this application. From left to right, the diagram shows data input, central analysis and processing, inventory monitoring, and dual early warning output. The data input section collects demand data and supply chain data from the entire domain and extracts demand fluctuation intensity sequences and supply interruption frequency sequences to provide basic data sources for subsequent analysis. The central analysis and processing section serves as the core computing unit of the entire system. It performs cross-recursive analysis on the two types of input sequences to determine the evolution path of the supply and demand mismatch risk. Then, it calculates the adaptive adjustment coefficient of the safety stock by combining the path direction and local curvature characteristics within the set risk window, generating a dynamic safety stock threshold curve that evolves smoothly over time. The inventory monitoring section is responsible for monitoring the actual inventory level of the inventory objects in real time and comparing its position with the dynamic safety stock threshold curve in real time. The dual early warning output section serves as the final execution triggering unit. When the inventory level is detected to be lower than the safety threshold, it triggers a stockout early warning to start the intelligent replenishment process. When the inventory level is higher than the safety threshold, it triggers a backlog early warning to start the intelligent shipping and allocation process.

[0020] refer to Figure 2 This diagram is an exemplary flowchart of an intelligent shipping and replenishment management method based on multi-dimensional inventory early warning, according to some embodiments of this application. The diagram mainly includes the following steps: In step S101, the full-domain demand data and full-link supply data of the inventory object are obtained, and the demand fluctuation intensity sequence in the full-domain demand data and the supply interruption frequency sequence in the full-link supply data are extracted.

[0021] It should be noted that, in this application, the inventory object refers to an independent inventory management unit that requires safety stock monitoring, shipment scheduling, and replenishment decisions. This includes, but is not limited to, a single homogeneous material group, key spare parts, or inventory entities aggregated according to specific business granularity. This inventory object has a complete full-domain demand data traceability link and a full-link supply data mapping relationship. It is the direct carrier for multi-dimensional inventory early warning, adaptive adjustment of safety thresholds, and judgment of stockout / overstock risk.

[0022] In some embodiments, obtaining the overall demand data and end-to-end supply data of inventory objects is achieved through the following steps: Obtain time-series demand records for inventory objects from omnichannel sales, pre-sales, and returns, and preprocess these time-series demand records to obtain omnichannel demand data. After obtaining the purchase arrival records, production delivery records, and logistics interruption records of the inventory objects and performing correlation processing, the end-to-end supply chain data is obtained.

[0023] Specifically, when acquiring the overall demand data for inventory items, time-series demand records collected from all sales channels, pre-sales channels, and return channels are aligned by timestamps and measured using a unified unit of measurement to form an original demand record set. The original demand record set is then filled with missing values ​​using a moving average method, and outliers (beyond three standard deviations) are removed to obtain a cleaned demand sequence. This cleaned demand sequence is then linked by sales order number and return order number, and the corresponding return volume is deducted from the sales demand to obtain a net demand sequence. This net demand sequence is used as the overall demand data. The overall demand data refers to a single time-series data reflecting the true net demand for inventory items in the market. Specifically, it involves aligning completed order records, locked but not yet shipped pre-sale records, and completed or ongoing return records for inventory items across all sales channels with a unified time granularity. The data, after deducting returns, forms a single time series that fully reflects the net demand for the inventory item in the market. For example, an inventory item generates 800 paid orders on an e-commerce platform, 150 completed transactions in the store's POS system, and 200 pre-sale orders with deposits paid but to be shipped in 30 days in the pre-sale system. The total number of returns across all channels on that day is 50. When calculating the total demand data for this inventory item on that day, the 800 e-commerce orders, 150 store sales, and 200 pre-sale orders are first summed after aligning with their daily timestamps, resulting in a total demand of 1150 items. Then, the 50 returned items are deducted as a negative deduction from the total demand data for that day, ultimately yielding a net demand of 1100 items for the inventory item on that day.

[0024] Specifically, when acquiring end-to-end supply chain data, the purchase arrival records, production delivery records, and logistics interruption records are left-joined using the purchase order number as the association key to obtain end-to-end supply chain data. A left join means that the purchase order is used as the main table, and all purchase records are retained when associating production and logistics records. The end-to-end supply chain data refers to the data after associating and integrating purchase arrival records, production delivery records, and logistics interruption records with the purchase order as the main association line. For example, an inventory object may have three purchase orders within a certain week, with the first order having a planned arrival date of Monday... The first order's actual delivery date was Monday, with a deviation of 0 days and a delivery quantity of 100 units. The second order's planned delivery date was Wednesday, and the actual delivery date was Saturday, with a deviation of 3 days and exceeding the preset tolerance threshold by 2 days, so it was marked as an interruption event, with a delivery quantity of 200 units. The third order's planned delivery date was Friday, and the actual delivery date was Friday, with a deviation of 0 days and a delivery quantity of 150 units. When calculating the end-to-end supply chain data for this inventory item, the purchase and delivery records, production and delivery records, and logistics interruption records of the three orders are associated by the purchase order number to obtain the end-to-end supply chain data for this inventory item.

[0025] In some embodiments, the extraction of the demand fluctuation intensity sequence from the global demand data and the supply interruption frequency sequence from the end-to-end supply data is achieved using the following steps: The global demand data is sliced ​​according to fixed time windows, and the coefficient of variation of demand within each time window is calculated to obtain the initial demand fluctuation sequence. The initial demand fluctuation sequence is smoothed by an exponentially weighted moving average to obtain a demand fluctuation intensity sequence; The procurement arrival records, production delivery records, and logistics interruption records in the entire supply chain data are correlated to construct a supply fulfillment event set indexed by purchase orders; In the set of supply fulfillment events, based on the planned delivery time, supply fulfillment events where the actual delivery time exceeds a preset tolerance threshold are marked as interruption events; The number of occurrences of the interruption event within each fixed time window is obtained, and the density of the occurrences is estimated to obtain the supply interruption frequency sequence.

[0026] In some embodiments, the global demand data is sliced ​​according to fixed time windows, and the coefficient of variation of demand within each time window is calculated to obtain the initial demand fluctuation sequence. This is achieved through the following steps: The time-series demand records corresponding to the global demand data are sorted by timestamp to form an original demand time series sampled at equal intervals; A sliding window of a preset fixed length is used to traverse the original demand time series with a preset step size, and multiple windowed demand subsequences are extracted in sequence. For each windowed demand subsequence, the ratio of the standard deviation to the mean of the demand for all sampling points within the windowed demand subsequence is calculated as the corresponding coefficient of variation. All the calculated coefficients of variation are arranged in time order of the sliding window to form the initial demand fluctuation sequence.

[0027] In specific implementation, after acquiring the global demand data, firstly, the time-series demand records of the global demand data are sorted in ascending order by the timestamp field to form the original demand time series arranged chronologically. The original demand time series refers to the set of continuous time-series demand observations without window segmentation. Secondly, a fixed length of 7 sampling points and a step size of 1 sampling point are set for the sliding window. Starting from the beginning of the original demand time series, the sliding window is used to truncate the data once every time it slides to the right by 1 sampling point. During truncation, the demand data of 7 consecutive sampling points are taken starting from the left boundary of the sliding window to form the first windowed demand subsequence. After that, the window is shifted to the right by 1 sampling point to continue truncation until the right boundary of the window exceeds the end of the sequence. This process is repeated to obtain multiple windowed demand subsequences equal to the length of the original sequence minus 6 sampling points. The sequence, the windowed demand subsequence, refers to a data segment of a fixed-length continuous subset in the original sequence; then, for each windowed demand subsequence, statistical formulas are used for calculation. First, the mean function is called to calculate the arithmetic mean of the demand of all sampling points in the subsequence, and then the standard deviation function is called to calculate the root mean square of the deviation of the demand of all sampling points in the subsequence from the mean. The quotient obtained by dividing the standard deviation by the mean is used as the coefficient of variation of the window. The coefficient of variation is a dimensionless ratio of the dispersion of a set of data to the central tendency, used to characterize the magnitude of the demand fluctuation under a unit demand level. Finally, the coefficients of variation calculated by all sliding windows in chronological order are stored in a one-dimensional array, which is the initial demand fluctuation sequence, the initial demand fluctuation sequence being the sequence formed by arranging the demand fluctuation coefficients of each window in chronological order.

[0028] In specific implementation, firstly, the overall demand data is sliced ​​according to fixed time windows, and the coefficient of variation of demand within each time window is calculated to obtain an initial demand fluctuation sequence. Secondly, the initial demand fluctuation sequence is iteratively calculated point by point using an exponentially weighted moving average with a smoothing coefficient α of 0.3. The smoothed value at the current moment is a weighted synthesis of the current original value and the smoothed value at the previous moment to obtain a demand fluctuation intensity sequence that eliminates high-frequency random noise. The exponentially weighted moving average refers to a moving average method that weights historical data with exponential decay weights. The demand fluctuation intensity sequence refers to a continuous index sequence reflecting the degree of instability on the demand side after smoothing. Furthermore, the purchase arrival records, production delivery records, and logistics interruption records in the full-link supply data are joined using the purchase order number as the primary key. The planned arrival time, planned delivery time, actual delivery time, and logistics anomaly event records of the same purchase order are merged into one row to construct a supply fulfillment event set. The inner join refers to retaining only the intersection of records that can be successfully matched in all three tables. In this method, the supply fulfillment event set refers to a data set that integrates complete fulfillment process records at the granularity of purchase orders. Then, in the supply fulfillment event set, for each fulfillment event, the number of days the actual delivery time is delayed relative to the planned delivery time is calculated based on the planned delivery time. The number of delay days is compared with a preset tolerance threshold, which is set to 2 days. In addition, the preset tolerance threshold can be set according to actual needs, which is not limited here. Fulfillment events with a delay of more than 2 days are marked as interruption events. Finally, the number of occurrences of the events marked as interruptions within each fixed time window is counted to obtain the interruption frequency count value for each window. Then, the interruption frequency count value sequence is subjected to kernel density estimation method for density estimation. Kernel density estimation uses Gaussian kernel function and the thumb rule to determine the bandwidth. The estimated probability density function is taken at the count value of each window to obtain the supply interruption frequency sequence. Kernel density estimation is a nonparametric method that uses a continuous kernel function to smooth discrete observations to estimate their probability density function.

[0029] It should be noted that the supply disruption frequency sequence in this application refers to a continuous indicator sequence reflecting the intensity of supply-side disruptions. Furthermore, it should be noted that a single-dimensional static indicator cannot capture the degree of uncertainty in the dynamic changes of both the demand and supply sides. The demand fluctuation intensity sequence is used to quantify the relative instability of the demand side after eliminating scale effects, while the supply disruption frequency sequence is used to quantify the intensity of disruptions in the supply-side fulfillment process. Both provide dual-source input sequences with temporal evolution characteristics from the demand and supply sides, respectively, for subsequent cross-recursive analysis. The demand fluctuation intensity sequence reflects the evolutionary trend of demand uncertainty over time, while the supply disruption frequency sequence reflects the accumulation and release of supply risks over time. By cross-recursively analyzing the two sequences, the dynamic evolution path of mismatch risk formed by the superposition of demand fluctuations and supply disruptions in the time dimension can be identified, thus providing a basis for adaptively adjusting the safety stock threshold.

[0030] In step S102, the demand fluctuation intensity sequence and the supply interruption frequency sequence are cross-recursively analyzed to determine the evolution path of the mismatch risk characterizing the demand for goods and the supply of goods.

[0031] In some embodiments, reference Figure 3 As shown in the figure, this is an exemplary flowchart illustrating the determination of the mismatch risk evolution path according to some embodiments of this application. In this embodiment, the demand fluctuation intensity sequence and the supply interruption frequency sequence are cross-recursively analyzed to determine the mismatch risk evolution path characterizing the relationship between commodity demand and commodity supply. This can be achieved through the following steps: In step S1021, the demand fluctuation intensity sequence and the supply interruption frequency sequence are reconstructed in phase space to obtain the demand phase space trajectory and the supply phase space trajectory, respectively. In step S1022, the cross recursion matrix of the demand phase space trajectory and the supply phase space trajectory is calculated. The elements of the cross recursion matrix indicate whether the states of the two phase space trajectories at the corresponding time fall into each other's neighborhood. In step S1023, a statistical analysis of the diagonal segment length distribution is performed on the cross recursive matrix, and the quantiles of the diagonal segment lengths are extracted as a quantitative indicator of mismatch risk. In step S1024, the quantitative indicators of the mismatch risk are arranged in chronological order to obtain the evolution path of the mismatch risk.

[0032] In some embodiments, the phase space reconstruction of the demand fluctuation intensity sequence and the supply interruption frequency sequence to obtain the demand phase space trajectory and the supply phase space trajectory is achieved by the following steps: The first embedding delay of the demand fluctuation intensity sequence is determined by mutual information method, and the first embedding dimension of the demand fluctuation intensity sequence is determined by spurious nearest neighbor method. Based on the first embedding delay and the first embedding dimension, the demand fluctuation intensity sequence is embedded with delay coordinates to obtain a demand phase space trajectory composed of multiple first delay vectors; The second embedding delay of the supply interruption frequency sequence is determined using the mutual information method, and the second embedding dimension of the supply interruption frequency sequence is determined using the spurious nearest neighbor method. The supply interruption frequency sequence is embedded with delay coordinates according to the second embedding delay and the second embedding dimension to obtain a supply phase space trajectory composed of multiple second delay vectors.

[0033] In specific implementation, firstly, after normalizing the demand fluctuation intensity sequence and the supply interruption frequency sequence through Z-score standardization, when determining the first embedding delay for the demand fluctuation intensity sequence using the mutual information method, the mutual information value between the demand fluctuation intensity sequence and itself at different delay steps is calculated. The mutual information value decays as the delay step increases. The delay step corresponding to the first drop of the mutual information value to its initial value 1 / e is taken as the first embedding delay. The mutual information method is a nonlinear method that determines the optimal delay by measuring the amount of shared information between the sequence and its delayed replicas. The first embedding delay refers to the time delay used when reconstructing the demand phase space. For example, for a demand fluctuation intensity sequence with 500 sampling points, the mutual information value is calculated sequentially with delay steps from 1 to 50. If the mutual information value is 0.85 when the delay step is 1, and the mutual information value drops to 0.85 / e≈0.313 for the first time when the delay step is 6, then the first embedding delay is taken as 6 sampling intervals. At the same time, the false nearest neighbor method is used to determine the first embedding dimension. Starting from the embedding dimension 1, the demand fluctuation intensity sequence is embedded with delayed coordinates. The ratio of the nearest neighbor distance of each point in the phase space under this embedding dimension to the corresponding nearest neighbor distance after dimensionality increase is calculated. If the ratio exceeds a preset threshold, the neighbor is a false nearest neighbor. The embedding dimension is gradually increased. When the proportion of false nearest neighbors first falls below 5%, the embedding dimension at this point is the first embedding dimension. The false nearest neighbor method refers to a method to determine the minimum embedding dimension by detecting the phenomenon that falsely adjacent points in the low-dimensional projection are separated in the high-dimensional projection due to insufficient dimension. The first embedding dimension refers to the minimum number of embedding dimensions required to reduce the proportion of falsely adjacent points in the phase space trajectory caused by projection to too low a dimension to below a preset threshold when reconstructing the phase space of the demand fluctuation intensity sequence. Secondly, based on the first embedding delay of 6 and the first embedding dimension of 3, the demand fluctuation intensity sequence x(t) is embedded with delayed coordinates to construct the first delay vector as [x(t), x(t-6)]. [x(t-12)], arrange the first delay vectors of all times in time to obtain the demand phase space trajectory, which refers to the path of the state points formed by the demand fluctuation intensity sequence in the reconstructed phase space as a function of time; similarly, calculate the self-mutual information sequence of the supply interruption frequency sequence using the same mutual information method and take the first delay step that drops to the initial value 1 / e as the second embedding delay. Use the false nearest neighbor method to search from dimension 1 incrementally until the proportion of false nearest neighbors is less than 5% to obtain the second embedding dimension. Then, embed the delay coordinates according to the second embedding delay and the second embedding dimension to construct the second delay vector and arrange it to obtain the supply phase space trajectory.

[0034] It should be noted that, in this embodiment, the supply phase space trajectory refers to the path of the state points formed by the supply interruption frequency sequence in the reconstructed phase space as it evolves over time. By determining the demand phase space trajectory and the supply phase space trajectory, the one-dimensional demand fluctuation intensity sequence and the supply interruption frequency sequence can be mapped to a high-dimensional phase space to restore the complete dynamic structure contained in their respective sequences. This allows the nonlinear evolution modes of the demand system and the supply system, which are difficult to observe from a single variable time series, to be unfolded in the high-dimensional phase space, thereby providing a topologically comparable state space expression for the subsequent calculation of the cross recursion matrix between the two phase space trajectories.

[0035] It should also be noted that the cross-recursive analysis in this application refers to the method of constructing a binary cross-recursive matrix by calculating the distance between the state points of the demand phase space trajectory and the supply phase space trajectory, which are two sequences of state points from different dynamic systems, and comparing it with the neighborhood radius. Then, the length distribution statistics of the continuous neighboring structures along the diagonal direction in the matrix are statistically analyzed to quantify the degree of synchronization of the evolution mode of the two systems in the phase space and its nonlinear analysis over time.

[0036] In specific implementation, firstly, the demand fluctuation intensity sequence and the supply interruption frequency sequence are reconstructed in phase space to obtain the demand phase space trajectory and the supply phase space trajectory, respectively. Secondly, after obtaining the demand phase space trajectory and the supply phase space trajectory, the neighborhood radius ε is set to 10% of the mean standard deviation of each trajectory. The Euclidean distance is calculated between the first delay vector at time i in the demand phase space trajectory and the second delay vector at time j in the supply phase space trajectory. If the distance is less than ε, the element at (i, j) of the cross recursion matrix is ​​set to 1; otherwise, it is set to 0. After traversing all i and j, a binary cross recursion matrix consisting of 0 and 1 is obtained. The neighborhood radius refers to the distance threshold for determining whether two phase space state points are adjacent. The cross recursion matrix is ​​a two-dimensional matrix that records the proximity relationship between the states of two different phase space trajectories at any two times. Then, the line segment consisting of consecutive 1 elements along the diagonal direction in the cross recursion matrix is ​​identified, and a diagonal traversal algorithm is used to identify the line segment from the cross recursion matrix. Extract all diagonal segments consisting of consecutive 1s from each diagonal of the recursive matrix and record the length of each segment to form a diagonal segment length sample set. The diagonal segment length refers to the number of consecutive adjacent state pairs appearing along the 45-degree diagonal direction in the cross-recursive matrix. Then, sort the diagonal segment length sample set in ascending order of length value and take the 95th percentile as the quantitative indicator of mismatch risk. The larger the quantile value, the higher the degree to which the two phase space trajectories maintain a similar evolution pattern over a longer period of time, that is, the demand and supply are currently in a synchronous and stable state. Conversely, the smaller the quantile value, the more mismatch occurs between demand and supply. For example, if there are 100 segments in the diagonal segment length sample set within a certain time window, and the length of the 95th segment after sorting by length in ascending order is 8 sampling points, then the quantitative indicator of mismatch risk at that time is 8. Finally, calculate the corresponding quantitative indicator value of mismatch risk at each sampling time along the time axis and arrange them in chronological order to obtain the mismatch risk evolution path.

[0037] It should be noted that the mismatch risk evolution path in this application refers to a continuous quantitative curve showing the change in the degree of mismatch between demand and supply over time. Traditional inventory early warning methods typically treat demand fluctuations and supply disruptions as two independent variables and set thresholds for each, failing to capture the dynamic mismatch process formed by their coupling and superposition over time. However, this embodiment maps the demand phase space trajectory and the supply phase space trajectory to the same phase space and calculates the cross-recursive matrix through cross-recursive analysis. This can identify the turning point from synchronization to divergence in the evolution of demand fluctuation patterns and supply disruption patterns. As a result, the mismatch risk is not presented as a static value at a single moment, but as a path with continuous evolution characteristics. This provides a temporally causal pre-judgment basis for adjusting the safety stock threshold in advance based on the direction and curvature of this path, enabling the adjustment of safety stock to respond in advance to structural changes in supply and demand mismatch rather than just reacting to inventory deviations after the fact.

[0038] In step S103, the adaptive adjustment coefficient of the inventory object when carrying out safety stock is determined based on the path evolution direction and local curvature characteristics of the mismatch risk evolution path within the set risk window, and the current safety stock threshold is adjusted by the adaptive adjustment coefficient to obtain the safety stock threshold curve that evolves smoothly over time.

[0039] In some embodiments, determining the adaptive adjustment coefficient of the inventory object when setting up safety stock based on the path evolution trend and local curvature characteristics of the mismatch risk evolution path within a set risk window is achieved through the following steps: On the mismatch risk evolution path, a path segment of a preset duration is extracted back from the current moment as a reference to set the risk window; Linear fitting is performed on the path segments within the defined risk window, and the slope of the resulting fitted line is taken as the path evolution direction. Local curvature calculation is performed on the path segments within the defined risk window, and the mean value of the obtained curvature is used as the local curvature feature; The path evolution direction and the local curvature features are input into a preset fuzzy logic controller, and the fuzzy logic controller outputs the adaptive adjustment coefficient of the inventory object when setting up safety stock.

[0040] In specific implementation, firstly, after obtaining the mismatch risk evolution path, a path segment of 14 past sampling points is extracted backward as a set risk window. The set risk window refers to a fixed-duration observation interval used to capture recent risk evolution dynamics. In addition, in other embodiments, it can be set according to actual needs, which is not limited here. The path segment within the set risk window is subjected to first-order linear fitting using the least squares method. The time values ​​of the 14 sampling points in the path segment are used as independent variables and the mismatch risk quantification index value is used as the dependent variable. The fitted line that minimizes the sum of squared residuals is calculated, and the slope of the fitted line is extracted as the path evolution trend. The path evolution trend refers to the overall increase or decrease trend of mismatch risk within the set risk window. The path segment within the set risk window is used to calculate the local curvature at each internal sampling point using the three-point arc length method. The 2nd to 13th sampling points in the path segment are selected, and an arc is fitted to each point and its three adjacent points, and the curvature of this arc is calculated. The arithmetic mean of the 12 local curvature values ​​is then used to obtain the local curvature feature. The three-point arc length method is a discrete curvature calculation method that approximates the curvature of the curve at the midpoint using the arc curvature determined by three adjacent points. The local curvature feature is a mean measure of the overall curvature of the mismatch risk evolution path within the set risk window. The path evolution direction and local curvature feature are used as two input variables to a preset Mamdani-type fuzzy logic controller. In the controller, the path evolution direction is fuzzified into five fuzzy sets {negative large, negative small, zero, positive small, positive large} using a triangular membership function. Local curvature features are fuzzified into three fuzzy sets {low, medium, high} using the same triangular membership function. Taking the path evolution direction as an example, the triangular membership function parameters for the "positive small" fuzzy set are set with the vertex at 0.3, the left base at 0.1, and the right base at 0.5. The membership function parameters for the "medium" fuzzy set are set with the vertex at 0.6, the left base at 0.3, and the right base at 0.9. The membership function parameters for other fuzzy sets are similarly preset. Then, inference is performed based on a preset IF-THEN fuzzy rule base. For example, in the rule base, if the path evolution direction is positive large A... The local curvature feature of ND is high, and the adaptive adjustment coefficient of THEN is also high. Finally, the centroid method is used to defuzzify the inference result, resulting in a continuous numerical adaptive adjustment coefficient. The Mamdani-type fuzzy logic controller refers to a fuzzy control system in which both input and output use fuzzy sets and generate accurate outputs through rule inference and defuzzification. For example, assuming that at the current moment, the calculated value of the path evolution direction is 0.38 and the calculated value of the local curvature feature is 0.62, after fuzzification, the membership degree of the path evolution direction 0.38 belonging to "small" is 0.76 and the membership degree of belonging to "zero" is 0.24. The membership degree of the local curvature feature 0.62 belonging to "medium" is 0.76 and the membership degree of belonging to "high" is 0.24; Then, rule reasoning is performed, using the minimum value operation to determine the activation degree of the four activated rules. The activation degree of the rule "If the direction is positively small and the curvature is medium, then the coefficient is relatively high" is min(0.76, 0.76) = 0.76; the activation degree of the rule "If the direction is positively small and the curvature is high, then the coefficient is high" is min(0.76, 0.24) = 0.24; the activation degree of the rule "If the direction is zero and the curvature is medium, then the coefficient is medium" is min(0.24, 0.76) = 0.24; the activation degree of the rule "If the direction is zero and the curvature is high, then the coefficient is relatively high" is min(0.24, 0.24) = 0.24; finally, the centroid method is used for defuzzification. The membership functions of the four rule output fuzzy sets are truncated by their respective activation degrees, and the union is calculated. The abscissa value corresponding to the geometric center of the area of ​​the union is calculated, and the adaptive adjustment coefficient is obtained as 1.086.

[0041] It should be noted that the adaptive adjustment coefficient in this application refers to the adjustment factor used to dynamically correct the safety stock threshold. The adjustment of the safety stock threshold needs to simultaneously perceive the changing direction and acceleration characteristics of the supply-demand mismatch risk. Relying solely on a single absolute value of risk or a simple threshold trigger cannot capture the inflection point of risk turning from a slow increase to a sharp increase in advance. Unlike existing technologies, this solution uses the direction of the mismatch risk evolution path and the local curvature as two independent dimensions to input the fuzzy logic controller. The direction reflects the current increase or decrease in risk, and the curvature reflects the acceleration of risk change, i.e., the severity of the turning point. The combined reasoning of the two enables the adaptive adjustment coefficient to increase the safety stock increase in advance when the curvature has increased significantly but the risk has not yet deteriorated significantly. This achieves an advanced response to the critical turning point in the non-stationary evolution of mismatch risk, rather than passively adjusting only after the risk has actually increased.

[0042] In some embodiments, adjusting the current safety stock threshold using the adaptive adjustment coefficient to obtain a safety stock threshold curve that evolves smoothly over time is achieved through the following steps: The baseline safety stock level calculated based on the classic inventory model is used as the current safety stock threshold. The adaptive adjustment parameter is used as a weighting factor and multiplied by the current safety stock threshold to obtain the initially adjusted safety stock value. Using the initially adjusted safety stock value as the target node, the initially adjusted safety stock values ​​of multiple consecutive times prior to the current time are used as a sequence of control points; B-spline curve interpolation is performed on the control point sequence and the target node to obtain a continuous and smooth safety stock threshold curve.

[0043] In practice, the first step is to obtain the inventory object based on the classic safety stock formula. The calculated baseline safety stock level SS is used, where Z is the standard normal distribution quantile corresponding to the preset service level, σ is the recent mean of the demand fluctuation intensity sequence, and L is the average replenishment lead time. This baseline safety stock level is used as the current safety stock threshold. The classic safety stock formula is a mathematical statistical formula for determining the buffer stock level based on the service level under the condition that both demand and lead time are random. Next, the adaptive adjustment coefficient is used as a multiplicative weighting factor and multiplied by the current safety stock threshold. For example, if the current safety stock threshold is 200 units and the adaptive adjustment coefficient is 1.086, the product of the two yields a preliminary adjusted safety stock value of 217 units. The weighting factor is a coefficient that acts in a multiplicative manner to scale the adjusted variable. The preliminary adjusted safety stock value refers to the discrete safety stock target point that has been corrected for risk conditions but has not yet undergone time-series smoothing. Then... Using the preliminary adjusted safety stock value of 217 units calculated at the current moment as the target node, the preliminary adjusted safety stock values ​​of the five consecutive sampling moments preceding the current moment are used as the control point sequence. In other embodiments, the multiple consecutive sampling moments preceding the current moment can be selected according to actual needs, which is not limited here. The control point sequence refers to a set of discrete value points used to constrain the shape of the interpolation curve. Finally, the six discrete values ​​of the control point sequence and the target node are interpolated using a cubic B-spline curve. These six values ​​are used as the control points of the B-spline curve. Intermediate points are generated between adjacent control points using a linear combination of cubic polynomial basis functions, so that the entire curve satisfies second-order continuous differentiability at the control points, resulting in a continuous and smooth safety stock threshold curve. The cubic B-spline curve refers to a piecewise polynomial curve recursively defined by the control points and cubic basis functions.

[0044] It should be noted that the safety stock threshold curve in this application refers to a function curve in which the safety stock threshold changes continuously and smoothly over time. The safety stock threshold curve is determined because the initial adjustment value after adaptive adjustment coefficient correction may have a step change between adjacent time points due to abrupt changes in the coefficient. If discrete points are directly used as the warning threshold, it will cause frequent short-term crossings between the inventory level and the threshold, causing repeated jumps in the warning signal. By using B-spline curve interpolation to make the discrete adjustment points continuous, the safety stock threshold is transformed from a numerical sequence of point-by-point jumps into a curve that evolves continuously and smoothly over time. The relative position of the inventory level and the curve can output a stable warning signal without high-frequency jitter. At the same time, the threshold at any future time on the curve can be obtained analytically through the interpolation equation, providing a dynamic benchmark with time continuity for shipping and replenishment decisions.

[0045] In step S104, the relative position of the inventory level of the inventory object and the safety stock threshold curve is monitored in real time. When the inventory level is lower than the safety stock threshold curve, a stockout risk warning is generated. When the inventory level is higher than the safety stock threshold curve, a backlog risk warning is generated.

[0046] In some embodiments, the relative position of the inventory level of an inventory object to the safety stock threshold curve is monitored in real time. When the inventory level is lower than the safety stock threshold curve, a stockout risk warning is generated; when the inventory level is higher than the safety stock threshold curve, an overstock risk warning is generated. This is achieved through the following steps: The real-time inventory level of the inventory object is obtained according to a preset sampling period to obtain a discrete inventory level sampling sequence; Align each sampling time in the inventory level sampling sequence with the safety stock threshold curve in time to extract the safety stock threshold corresponding to each sampling time. Calculate the difference between the real-time inventory level and the corresponding safety stock threshold at each sampling time to obtain the inventory deviation sequence; Traverse the inventory deviation sequence. If the inventory deviation at the current sampling time is negative, generate the stockout risk warning. If the inventory deviation at the current sampling time is positive, generate the overstock risk warning.

[0047] In specific implementation, firstly, the inventory module of the warehouse management system calls the inventory query interface every hour at a preset sampling period to obtain the real-time inventory level of the inventory object. The values ​​obtained from multiple consecutive samplings are stored in a one-dimensional array according to their timestamps, forming a discrete inventory level sampling sequence. The real-time inventory level refers to the actual available quantity of the inventory object in the warehouse at the current moment, and the inventory level sampling sequence refers to the set of time-series observations of inventory quantity collected at fixed time intervals. Secondly, the timestamp of each sampling point in the inventory level sampling sequence is aligned with the time axis of the safety stock threshold curve. The safety stock threshold curve is analytically expressed by a B-spline interpolation equation. The current sampling time is substituted into this interpolation equation to calculate the safety stock threshold corresponding to that sampling time. Time alignment refers to the process of precisely matching the discrete sampling time with the corresponding time on the continuous curve to obtain the curve value at that time. For example, if a sampling time is 10:00 AM on a certain day, substituting this time into the B-spline interpolation equation of the safety stock threshold curve yields a safety stock threshold of 72 items at that time. Then, for each sampling... At any given time, the real-time inventory level is subtracted from the corresponding safety stock threshold. For example, subtracting the safety stock threshold of 72 from the real-time inventory level of 68 units yields a difference of -4 units. The differences from all sampling times are arranged in chronological order to form an inventory deviation sequence. This inventory deviation sequence refers to the time-series sequence composed of the differences between the actual inventory level and the dynamic safety stock threshold at each sampling time. Finally, each element in the inventory deviation sequence is iterated. If the inventory deviation at the current sampling time is negative, a stockout risk warning is triggered, and a replenishment suggestion notification is pushed to the purchasing module. If the inventory deviation at the current sampling time is positive, an overstock risk warning is triggered, and a promotion suggestion notification is pushed to the sales module. The stockout risk warning is a supply shortage warning signal issued when the inventory level is lower than the dynamic safety stock threshold. When the inventory deviation is 0, it means that the real-time inventory level and the safety stock threshold curve coincide at that sampling time. At this time, neither a stockout risk warning nor an overstock risk warning is generated. The system determines that the inventory object is in an ideal state of supply and demand balance and only maintains regular inventory monitoring without triggering any warning push action.

[0048] It should be noted that the overstock risk warning in this application refers to the warning signal issued when the inventory level is higher than the dynamic safety stock threshold, indicating the risk of capital occupation and slow sales.

[0049] Furthermore, in another aspect of this application, in some embodiments, this application provides an intelligent shipping and replenishment management system based on multi-dimensional inventory early warning, referencing... Figure 4The figure is a schematic diagram of the structure of an intelligent shipping and replenishment management system based on multi-dimensional inventory early warning, according to some embodiments of this application. This intelligent shipping and replenishment management system based on multi-dimensional inventory early warning includes: an acquisition module 201, a processing module 202, and an execution module 203, which are described below: The acquisition module 201 in this application is mainly used to acquire the full-domain demand data and full-link supply data of the inventory object, and extract the demand fluctuation intensity sequence in the full-domain demand data and the supply interruption frequency sequence in the full-link supply data. Processing module 202, in this application, is mainly used to perform cross-recursive analysis on the demand fluctuation intensity sequence and the supply interruption frequency sequence, thereby determining the evolution path of the mismatch risk characterizing the demand for goods and the supply of goods. The processing module 202 is further configured to determine the adaptive adjustment coefficient of the inventory object when carrying out safety stock based on the path evolution direction and local curvature characteristics of the mismatch risk evolution path within the set risk window, and adjust the current safety stock threshold through the adaptive adjustment coefficient to obtain a safety stock threshold curve that evolves smoothly over time. The execution module 203 in this application is mainly used to monitor the relative position of the inventory level of the inventory object and the safety stock threshold curve in real time. When the inventory level is lower than the safety stock threshold curve, a stockout risk warning is generated. When the inventory level is higher than the safety stock threshold curve, a backlog risk warning is generated.

[0050] In addition, this application also provides a computer device, the computer device including a memory and a processor, the memory storing code, the processor being configured to acquire the code and execute the above-described intelligent shipment and replenishment management method based on multi-dimensional inventory early warning.

[0051] In some embodiments, reference Figure 5 The figure is a schematic diagram of the structure of a computer device implementing an intelligent shipping and replenishment management method based on multi-dimensional inventory early warning, according to some embodiments of this application. The intelligent shipping and replenishment management method based on multi-dimensional inventory early warning in the above embodiments can be implemented through... Figure 5 The computer device shown is used to implement this, and the computer device includes at least one processor 301, a communication bus 302, a memory 303, and at least one communication interface 304.

[0052] The processor 301 can be a general-purpose central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more devices used to control the execution of the intelligent shipping and replenishment management method based on multi-dimensional inventory warning in this application.

[0053] The communication bus 302 can be used to transmit information between the aforementioned components.

[0054] The memory 303 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 303 may exist independently and be connected to the processor 301 via the communication bus 302. The memory 303 may also be integrated with the processor 301.

[0055] The memory 303 stores program code for executing the solution of this application, and its execution is controlled by the processor 301. The processor 301 executes the program code stored in the memory 303. The program code may include one or more software modules. In the above embodiments, the determination of the intelligent shipping and replenishment management method based on multi-dimensional inventory early warning can be implemented by the processor 301 and one or more software modules in the program code in the memory 303.

[0056] Communication interface 304 uses any transceiver-like device for communicating with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0057] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0058] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device can be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0059] In addition, this application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-mentioned intelligent shipping and replenishment management method based on multi-dimensional inventory early warning.

[0060] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0061] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A smart shipping and replenishment management method based on multi-dimensional inventory early warning, characterized in that, Includes the following steps: Obtain the full-domain demand data and full-link supply data of the inventory objects, and extract the demand fluctuation intensity sequence from the full-domain demand data and the supply interruption frequency sequence from the full-link supply data; By performing cross-recursive analysis on the demand fluctuation intensity sequence and the supply interruption frequency sequence, the evolution path of the mismatch risk between commodity demand and commodity supply is determined. Specifically, this includes: reconstructing the phase space of the demand fluctuation intensity sequence and the supply interruption frequency sequence to obtain the demand phase space trajectory and the supply phase space trajectory, respectively; calculating the cross-recursive matrix of the demand phase space trajectory and the supply phase space trajectory, where the elements of the cross-recursive matrix indicate whether the states of the two phase space trajectories at corresponding times fall into each other's neighborhoods; performing statistical analysis on the diagonal segment length distribution of the cross-recursive matrix and extracting the quantiles of the diagonal segment length as a quantitative indicator of mismatch risk; and arranging the quantitative indicators of mismatch risk in chronological order to obtain the mismatch risk evolution path. On the path of supply and demand mismatch risk evolution, a path segment of a preset duration is extracted backward from the current moment as a set risk window; the path segment within the set risk window is linearly fitted, and the slope of the fitted line is taken as the path evolution direction; the path segment within the set risk window is local curvature calculated, and the mean of the obtained curvature is taken as the local curvature feature; the path evolution direction and the local curvature feature are input into a preset fuzzy logic controller, and the fuzzy logic controller outputs the adaptive adjustment coefficient of the inventory object when setting safety stock. The adaptive adjustment coefficient is an adjustment factor used to dynamically correct the safety stock threshold. The adjustment of the safety stock threshold needs to simultaneously sense the changing direction and acceleration characteristics of the supply and demand mismatch risk, and adjust the current safety stock threshold through the adaptive adjustment coefficient to obtain a safety stock threshold curve that evolves smoothly over time. The system monitors the relative position of the inventory level of an inventory item to the safety stock threshold curve in real time. When the inventory level is lower than the safety stock threshold curve, a stockout risk warning is generated. When the inventory level is higher than the safety stock threshold curve, a backlog risk warning is generated.

2. The method as described in claim 1, characterized in that, Obtaining full-domain demand data and full-chain supply data for inventory objects specifically includes: Obtain time-series demand records for inventory objects from omnichannel sales, pre-sales, and returns, and preprocess these time-series demand records to obtain omnichannel demand data. After obtaining the purchase arrival records, production delivery records, and logistics interruption records of the inventory objects and performing correlation processing, the end-to-end supply chain data is obtained.

3. The method as described in claim 1, characterized in that, Extracting the demand fluctuation intensity sequence from the overall demand data and the supply interruption frequency sequence from the end-to-end supply data specifically includes: The global demand data is sliced ​​according to fixed time windows, and the coefficient of variation of demand within each time window is calculated to obtain the initial demand fluctuation sequence. The initial demand fluctuation sequence is smoothed by an exponentially weighted moving average to obtain a demand fluctuation intensity sequence; The procurement arrival records, production delivery records, and logistics interruption records in the entire supply chain data are correlated to construct a supply fulfillment event set indexed by purchase orders; In the set of supply fulfillment events, based on the planned delivery time, supply fulfillment events where the actual delivery time exceeds a preset tolerance threshold are marked as interruption events; The number of occurrences of the interruption event within each fixed time window is obtained, and the density of the occurrences is estimated to obtain the supply interruption frequency sequence.

4. The method as described in claim 3, characterized in that, The overall demand data is sliced ​​according to fixed time windows, and the coefficient of variation of demand within each time window is calculated to obtain the initial demand fluctuation sequence, which specifically includes: The time-series demand records corresponding to the global demand data are sorted by timestamp to form an original demand time series sampled at equal intervals; A sliding window of a preset fixed length is used to traverse the original demand time series with a preset step size, and multiple windowed demand subsequences are extracted in sequence. For each windowed demand subsequence, the ratio of the standard deviation to the mean of the demand for all sampling points within the windowed demand subsequence is calculated as the corresponding coefficient of variation. All the calculated coefficients of variation are arranged in time order of the sliding window to form the initial demand fluctuation sequence.

5. The method as described in claim 1, characterized in that, Phase space reconstruction is performed on the demand fluctuation intensity sequence and the supply interruption frequency sequence respectively to obtain the demand phase space trajectory and the supply phase space trajectory, specifically including: The first embedding delay of the demand fluctuation intensity sequence is determined by mutual information method, and the first embedding dimension of the demand fluctuation intensity sequence is determined by spurious nearest neighbor method. Based on the first embedding delay and the first embedding dimension, the demand fluctuation intensity sequence is embedded with delay coordinates to obtain a demand phase space trajectory composed of multiple first delay vectors; The second embedding delay of the supply interruption frequency sequence is determined using the mutual information method, and the second embedding dimension of the supply interruption frequency sequence is determined using the spurious nearest neighbor method. The supply interruption frequency sequence is embedded with delay coordinates according to the second embedding delay and the second embedding dimension to obtain a supply phase space trajectory composed of multiple second delay vectors.

6. An intelligent shipping and replenishment management system based on multi-dimensional inventory early warning, used to execute the intelligent shipping and replenishment management method based on multi-dimensional inventory early warning as described in any one of claims 1 to 5, characterized in that, The system includes: The acquisition module is used to acquire the overall demand data and the entire supply chain data of the inventory objects, and extract the demand fluctuation intensity sequence from the overall demand data and the supply interruption frequency sequence from the entire supply chain data. The processing module is used to perform cross-recursive analysis on the demand fluctuation intensity sequence and the supply interruption frequency sequence, thereby determining the evolution path of the mismatch risk characterizing the demand for goods and the supply of goods. The processing module is further configured to determine the adaptive adjustment coefficient of the inventory object when carrying out safety stock based on the path evolution direction and local curvature characteristics of the mismatch risk evolution path within the set risk window, and adjust the current safety stock threshold through the adaptive adjustment coefficient to obtain a safety stock threshold curve that evolves smoothly over time. The execution module is used to monitor the relative position of the inventory level of the inventory object with the safety stock threshold curve in real time. When the inventory level is lower than the safety stock threshold curve, a stockout risk warning is generated. When the inventory level is higher than the safety stock threshold curve, a backlog risk warning is generated.

7. A computer device, characterized in that, The computer device includes a memory and a processor, the memory storing code, and the processor being configured to retrieve the code and execute the intelligent shipment and replenishment management method based on multi-dimensional inventory early warning as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the intelligent shipping and replenishment management method based on multi-dimensional inventory early warning as described in any one of claims 1 to 5.

Images