Supply chain demand prediction and inventory optimization method and system based on AI

Abstract

The invention relates to an AI-based supply chain demand prediction and inventory optimization method and system, and the method comprises the following steps: calling a sales record from a supply chain, recognizing a periodic demand rule of a corresponding commodity, and predicting an estimated demand quantity in a future time period according to the periodic demand rule; carrying out supply and demand difference analysis by combining the current warehouse commodity stock margin, and judging the stock state; if the commodity stock surplus exists, making a corresponding stock digestion strategy; if a shortage quantification result appears, carrying out key degree evaluation and urgency sorting based on a shortage amount in combination with market factors, and generating a key commodity list and an out-of-stock urgency index; and finally, in combination with a preset commodity constraint condition, according to the key commodity list and the urgency index, an optimal replenishment scheme meeting resource and time limit is formulated, and the technical problem of how to realize conversion from passive response to active pre-judgment in a supply chain environment with large demand fluctuation and complex influence factors is solved.

Description

Technical Field

[0001] This invention relates to the field of supply chain technology, and in particular to an AI-based method and system for supply chain demand forecasting and inventory optimization. Background Technology

[0002] In the context of globalization and rapid market changes, supply chain management faces increasingly complex challenges in addressing demand fluctuations and inventory control. Traditional forecasting methods often rely on historical averages or simple time series analysis, making it difficult to accurately capture the cyclical patterns and sudden changes in commodity demand, resulting in significant forecasting errors. Simultaneously, enterprises generally suffer from delayed response times and inefficient replenishment strategies in inventory management. This can lead to resource waste and capital tied up due to excess inventory, or negatively impact customer satisfaction and corporate reputation due to stockouts. Existing systems lack dynamic assessment and tiered response mechanisms for supply-demand discrepancies, making it particularly difficult to achieve refined and intelligent inventory adjustment and replenishment decisions, especially when dealing with complex supply chains involving multiple product categories and nodes. The above-mentioned technical solution integrates AI-driven demand forecasting, supply-demand gap analysis, criticality assessment, and constraint optimization. The main technical problem it addresses is how to shift from "passive response" to "proactive prediction" in a supply chain environment with large demand fluctuations and complex influencing factors. In particular, it aims to build an intelligent decision-making mechanism that can automatically identify key shortage commodities and generate the optimal replenishment plan based on their urgency and constraints, thereby improving inventory turnover efficiency, reducing operating costs, and ensuring supply continuity. Summary of the Invention

[0003] This invention aims to address at least one of the technical problems existing in the prior art. To this end, this invention proposes an AI-based method and system for supply chain demand forecasting and inventory optimization.

[0004] The technical solution adopted in this invention is: On one hand, embodiments of the present invention provide an AI-based supply chain demand forecasting and inventory optimization method, including the following steps: Retrieve sales records of goods from the supply chain and identify the cyclical demand patterns of the corresponding goods based on the sales records; Based on the cyclical demand pattern, the estimated demand for future periods is determined, the current inventory balance of goods in the warehouse is obtained, and a supply-demand difference analysis is performed based on the inventory balance and the estimated demand to obtain the supply-demand difference analysis results. If the result of the supply and demand difference analysis is a surplus of commodity inventory, then an inventory digestion strategy is formulated based on the surplus of commodity inventory. When the supply and demand difference analysis result is a shortage quantification result, the keyness and urgency of the commodities are assessed and ranked based on the shortage quantification result to obtain a list of key commodities and the corresponding shortage urgency index. By using preset product constraints, the optimal replenishment plan is formulated based on the list of key products and the corresponding shortage urgency index.

[0005] Furthermore, the step of retrieving sales records of goods from the supply chain and identifying the cyclical demand patterns of the corresponding goods based on the sales records includes: The sales records of the goods are retrieved through the database interface of the supply chain, and the daily sales data in the sales records are decomposed into trends to obtain the basic sales curve. The peaks and troughs of the basic sales curve are identified to obtain the periodic fluctuation range. Based on the periodic fluctuation range, the sales records are rhythmically aligned to obtain standardized periodic segments, and demand patterns are identified in the standardized periodic segments to obtain periodic demand patterns.

[0006] Furthermore, the method of determining the estimated demand for future periods based on cyclical demand patterns includes: Based on the cyclical demand pattern, the basic sales curve is extrapolated to obtain a trend extension line. Future time period nodes are marked on the trend extension line to obtain predicted time nodes. The cyclical fluctuation range is extended using these predicted time nodes to obtain future fluctuation ranges. The peak and trough positions of these future fluctuation ranges are estimated to obtain peak and trough prediction points. Based on these peak and trough prediction points, the future demand for the product is estimated to obtain the predicted demand.

[0007] Furthermore, the supply-demand gap analysis based on the remaining inventory and the estimated demand, to obtain the supply-demand gap analysis results, includes: The remaining inventory of the goods is summarized by multiple goods to obtain total inventory data, and the estimated demand is broken down by goods and future time periods to obtain demand segmentation data. Based on the total inventory data and the demand segmentation data, supply and demand are matched on a product-by-product and time-by-time basis to obtain a preliminary supply and demand comparison table. When there are discrepancies in the preliminary supply and demand comparison table, the discrepancies are marked in the preliminary supply and demand comparison table to generate a discrepancy mark table, and the discrepancy mark table is subjected to a discrepancy quantification analysis to obtain the supply and demand difference analysis results.

[0008] Furthermore, the step of formulating an inventory reduction strategy based on the surplus of commodity inventory includes: The surplus of the commodity inventory is classified according to the shelf life of the commodity to obtain a near-expiry classification table, and the surplus quantity in the near-expiry classification table is marked with a time limit for digestion to obtain a time constraint table. Based on the aforementioned time constraint table, the unit cost of the goods is calculated based on various preset digestion methods to obtain a cost accounting sheet. The cost accounting sheet is then matched with the historical sales increase data of the corresponding digestion method to obtain the cost-effectiveness ratio. Based on the cost-effectiveness ratio, the digestion methods of each commodity are combined and calculated to obtain a set of combined solutions. Based on a preset AI model, the total digestion cost and expected digestion volume in the set of combined solutions are balanced and verified to obtain a feasible solution group. The feasible schemes are prioritized according to their digestion cycle length to obtain the scheme ranking results, and the optimal combination scheme is selected from the scheme ranking results as the inventory digestion strategy.

[0009] Furthermore, the assessment of the criticality and urgency of goods based on the shortage quantification results, resulting in a list of key goods and corresponding shortage urgency indices, includes: The shortage quantification results are summarized by commodity category to obtain a total commodity shortage table, and recent market dynamics are collected to obtain a set of market influencing factors. Based on the total shortage table of the commodities and the set of market influencing factors, a shortage risk score is given for each category of commodities to obtain the risk score result. The urgency of the goods is weighted by risk scoring results to obtain weighted urgency parameters, and the goods are ranked based on the weighted urgency parameters to obtain an urgency ranking table. Based on the urgency ranking table, products with urgency higher than a preset threshold are selected to generate a key product list, and each product in the key product list is assigned a corresponding shortage urgency index based on the risk scoring results.

[0010] Furthermore, the step of formulating an optimal replenishment plan for goods based on a list of key goods and their corresponding shortage urgency index, using preset goods constraints, includes: The preset commodity constraints are classified and analyzed to obtain resource constraints and timeliness constraints. The corresponding replenishment cost parameters are matched to the commodities in the key commodity list to obtain the commodity cost table. Based on the stockout urgency index, the replenishment priority of the goods in the commodity cost table is sorted to obtain a priority sequence. The feasibility of the priority sequence is then verified according to the resource constraint item to obtain feasible replenishment items. Based on the feasible replenishment items and the timeliness constraints, the replenishment quantity is adjusted and calculated to obtain the adjusted replenishment quantity. The adjusted replenishment quantity is then summarized and arranged according to the shortage urgency index to obtain the optimal replenishment plan.

[0011] This invention also provides an AI-based supply chain demand forecasting and inventory optimization system, comprising: The mining module is used to retrieve the sales records of goods from the supply chain and identify the cyclical demand patterns of the corresponding goods based on the sales records. The analysis module is used to determine the estimated demand for future periods based on the cyclical demand pattern, obtain the current inventory balance of goods in the warehouse, and perform supply and demand difference analysis based on the inventory balance and the estimated demand to obtain the supply and demand difference analysis results. The first formulation module is used to formulate an inventory digestion strategy based on the commodity inventory surplus when the supply and demand difference analysis result is a commodity inventory surplus. The sorting module is used to assess the criticality and urgency of commodities based on the shortage quantification result when the supply and demand difference analysis result is a shortage quantification result, so as to obtain a list of key commodities and the corresponding shortage urgency index. The second module is used to formulate the optimal replenishment plan for goods based on a list of key goods and the corresponding shortage urgency index, according to preset goods constraints.

[0012] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the methods described above.

[0013] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of any of the methods described above.

[0014] This invention provides an AI-based supply chain demand forecasting and inventory optimization method, comprising the following steps: retrieving sales records of goods from the supply chain and identifying the cyclical demand patterns of the corresponding goods based on the sales records; determining the estimated demand for future periods based on the cyclical demand patterns, obtaining the current inventory balance in the warehouse, and performing a supply-demand difference analysis based on the inventory balance and the estimated demand to obtain the supply-demand difference analysis result; when the supply-demand difference analysis result is a surplus of goods inventory, formulating an inventory reduction strategy based on the surplus of goods inventory; when the supply-demand difference analysis result is a shortage quantification result, then based on... The shortage quantification results are used to assess the criticality and urgency of goods, resulting in a list of key goods and corresponding shortage urgency indices. Based on preset goods constraints and the list of key goods and their corresponding shortage urgency indices, optimal replenishment plans are formulated. This solves the technical problem of shifting from passive response to proactive prediction in a supply chain environment with large demand fluctuations and complex influencing factors. It also enables the development of targeted inventory reduction strategies (such as promotions and transfers) based on inventory surpluses after identification, preventing goods from accumulating, expiring, or depreciating, and effectively reducing inventory holding costs and capital tied up. Attached Figure Description

[0015] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a flowchart illustrating the steps of the AI-based supply chain demand forecasting and inventory optimization method in this embodiment of the invention. Figure 2 This is a structural block diagram of the AI-based supply chain demand forecasting and inventory optimization system in this embodiment of the invention; Figure 3 This is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0016] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0018] In the description of this invention, it should be understood that the orientation descriptions, such as up, down, front, back, left, right, etc., are based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0019] In the description of this invention, "several" means one or more, "more than" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. The use of "first" and "second" in the description is merely for distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.

[0020] In the description of this invention, unless otherwise explicitly defined, terms such as "setting," "installing," and "connecting" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this invention in conjunction with the specific content of the technical solution.

[0021] The embodiments of this application will be further described below with reference to the accompanying drawings.

[0022] Reference Figure 1This invention provides an AI-based supply chain demand forecasting and inventory optimization method, comprising the following steps: Step S1: Retrieve the sales records of the goods from the supply chain, and identify the cyclical demand patterns of the corresponding goods based on the sales records.

[0023] Specifically, sales records of goods are retrieved from the supply chain, and the cyclical demand patterns of the corresponding goods are identified based on these records. This process obtains historical sales records of the target goods through the data interface of the enterprise supply chain management system. The data includes the product number, sales date, and daily sales volume, ensuring the continuity of the time series. Missing dates are interpolated to complete the data. Subsequently, time series decomposition techniques, such as STL decomposition, are applied to the daily sales data to separate the trend term, seasonal term, and residual term. The seasonal term reflects repetitive fluctuation characteristics and forms the basic sales curve. Then, the basic sales curve is used to... The sliding window method is used to identify adjacent peaks and troughs and calculate their time intervals. If multiple intervals approach a fixed number of days, such as a beverage product experiencing a sales peak on Saturdays for several consecutive weeks with a stable cycle of 7 days, then the product is confirmed to have a weekly cycle. Historical sales are then aligned according to this cycle, and sales segments of the same phase each week are extracted to form standardized cycle segments. The average of multiple segments is aggregated to obtain the relative demand ratio of the product in each period within the cycle, such as Saturday being 1.4 times the baseline value and Tuesday being 0.8 times. Finally, the cycle demand pattern of the product is summarized and used to predict future demand.

[0024] Step S2: Determine the estimated demand for future periods based on the cyclical demand pattern, obtain the current inventory balance of goods in the warehouse, and perform a supply-demand difference analysis based on the inventory balance and the estimated demand to obtain the supply-demand difference analysis results.

[0025] Specifically, based on cyclical demand patterns, the estimated demand for future periods is determined, the current inventory level in the warehouse is obtained, and a supply-demand gap analysis is performed based on the inventory level and the estimated demand to obtain the supply-demand gap analysis results. This step first utilizes the cyclical demand patterns discovered in the previous steps to map the patterns to a specific future period. For example, considering the cyclical demand pattern of a certain beverage product where sales increase by 40% every Friday to Sunday in summer, combined with the actual sales baseline of the most recent period, the estimated demand for this product in the target warehouse next Saturday is calculated to be 1200 units. Subsequently, the current inventory level of this product in the warehouse is read in real time through the warehouse management system, and the current inventory level is found to be 80 units. 0 units. Next, the estimated demand is compared with the remaining inventory, and the difference between the two is calculated, i.e., 1200 units minus 800 units, resulting in a supply-demand difference of -400 units. A negative value indicates that the supply is insufficient to meet the demand. This value is the result of the supply-demand difference analysis. This result not only reflects the absolute amount of shortage or surplus, but also serves as a basis for subsequent judgment of inventory status. For example, when the supply-demand difference is positive, such as an estimated demand of 600 units and an inventory balance of 900 units, the difference is +300 units, indicating that there is an inventory surplus. All calculations are based on a unified time granularity and product dimension to ensure that the analysis results are consistent with the actual operational rhythm and serve as a direct input to trigger subsequent inventory digestion strategies or replenishment processes.

[0026] Step S3: If the result of the supply and demand difference analysis is a surplus of commodity inventory, then an inventory digestion strategy is formulated based on the surplus of commodity inventory.

[0027] Specifically, when the supply-demand difference analysis result is a surplus in commodity inventory, an inventory clearance strategy is formulated based on this surplus. This step is triggered when the aforementioned supply-demand difference analysis result is positive. For example, if the estimated demand for a beverage is 600 units, and the current inventory surplus in the warehouse is 900 units, the calculated supply-demand difference is +300 units, indicating a surplus in commodity inventory. At this time, the system uses this surplus of 300 units as input data and combines it with commodity attribute information such as shelf life, storage costs, and historical turnover rate for matching and judgment. If the commodity is nearing its shelf life or has high storage costs, the clearance strategy is initiated first. The process then calls upon a pre-defined inventory clearance rule library, which includes strategy options such as promotional activity templates, cross-regional allocation paths, and bundled sales schemes. The system recommends the optimal strategy based on the current sales channel status and customer profiles. For example, it might push a "buy two get one free" promotion to online platforms to accelerate shipments, or allocate surplus goods to nearby warehouses with higher estimated demand based on regional demand differences. The entire process is executed through an automated decision engine that generates strategies and outputs an inventory clearance strategy that includes specific execution actions, target channels, time windows, and expected consumption, ensuring that surplus goods are effectively handled without affecting the overall supply chain balance.

[0028] Step S4: When the supply and demand difference analysis result is the shortage quantification result, the keyness assessment and urgency ranking of the commodities are carried out based on the shortage quantification result to obtain a list of key commodities and the corresponding shortage urgency index.

[0029] Specifically, when the supply-demand gap analysis result is a shortage quantification result, the criticality of the goods is assessed and their urgency is ranked based on the shortage quantification result, resulting in a list of key goods and corresponding shortage urgency indices. This step is initiated when the aforementioned supply-demand gap analysis result is negative. For example, if the estimated demand for a beverage is 1200 units and the current inventory in the warehouse is 800 units, the calculated supply-demand gap is -400 units. This negative value constitutes the shortage quantification result. The system uses this shortage quantification result of 400 units as the basic input and combines it with the classification attributes, unit cost, historical sales, and customer order priority in the product master data to assess the criticality. During the assessment, a weighted scoring model is used to evaluate the product's position in the supply chain. The impact of the product is quantified. For example, if a beverage is a key summer product and its daily order coverage exceeds 70%, it is classified as high criticality. The criticality score is then correlated with the shortage quantification result. Taking into account the duration of the shortage, the restocking cycle, and the availability of substitutes, a shortage urgency index is generated through a preset algorithm. For example, products with large shortages and no substitutes are assigned higher index values. The system sorts all products with shortage quantification results and outputs a list containing product number, shortage quantity, criticality level, and corresponding shortage urgency index, which is the critical product list. This list is arranged from high to low index and is used to guide the priority allocation of subsequent restocking resources to ensure that high-impact and high-urgency products are responded to first.

[0030] Step S5: Based on the key product list and the corresponding shortage urgency index, formulate the optimal replenishment plan for the products by using the preset product constraints.

[0031] Specifically, based on preset product constraints and a list of key products with corresponding stockout urgency indices, the system formulates optimal replenishment plans for each product. This step is initiated after obtaining the list of key products and the stockout urgency index for each product. First, the system loads the preset product constraints, which include supplier delivery cycle, maximum single delivery quantity, procurement cost budget, transportation resource availability, warehouse receiving time window, and minimum order quantity. For example, if a beverage product has a stockout urgency index of 92 (out of 100), ranking first in the list of key products, it indicates that its replenishment priority is the highest. The system uses the quantified shortage result of 400 units for this product as the replenishment target quantity, combined with its supplier's delivery cycle... The system is matched and verified under constraints such as a 3-day supply period and a maximum daily supply capacity of 500 units. Then, linear programming or heuristic optimization algorithms are used to comprehensively solve for all products to be replenished. Under the premise of satisfying the constraints of each product, the algorithm prioritizes the allocation of resources to products with high shortage urgency index and adjusts the replenishment batches and quantities to balance cost and timeliness. For example, high-index products are arranged in the most recent transportation batch and loaded first, while orders for products with lower indexes are merged to reduce procurement costs. Finally, the system outputs a complete replenishment plan that includes product number, replenishment quantity, supplier name, estimated arrival time, transportation batch, and total cost. This plan achieves the optimal balance between overall replenishment efficiency and response speed under the constraints.

[0032] In a specific embodiment, retrieving sales records of goods from the supply chain and identifying the cyclical demand patterns of the corresponding goods based on the sales records includes: The sales records of the goods are retrieved through the database interface of the supply chain, and the daily sales data in the sales records are decomposed into trends to obtain the basic sales curve. The peaks and troughs of the basic sales curve are identified to obtain the periodic fluctuation range. Based on the periodic fluctuation range, the sales records are rhythmically aligned to obtain standardized periodic segments, and demand patterns are identified in the standardized periodic segments to obtain periodic demand patterns.

[0033] Specifically, the sales records of the product are retrieved through the supply chain database interface, and the daily sales data in the sales records are decomposed into a basic sales curve. Peaks and troughs are then identified on the basic sales curve to obtain cyclical fluctuation ranges. Based on these cyclical fluctuation ranges, the sales records are rhythmically aligned to obtain standardized cyclical segments. Demand patterns are then identified on these standardized cyclical segments to obtain cyclical demand patterns. This process first uses a standard database connection protocol to extract historical sales records of the target product from the enterprise's ERP or WMS system, obtaining a structured data sequence containing product number, sales date, and corresponding daily sales volume. This ensures the data covers a sufficiently long time range to capture potential cycles. Missing or abnormal daily sales data is repaired using front-to-back averaging or linear interpolation. Subsequently, a trend decomposition operation is performed on the repaired daily sales data using STL (Seasonal and Trend Decomposition Using...) The Loess algorithm decomposes the original sales sequence into a trend term, a seasonal term, and a residual term, retaining the seasonal term as the basic sales curve. This curve reflects the repetitive fluctuations in sales over fixed time intervals. Then, a local extremum detection algorithm is applied to the basic sales curve, setting a sliding window width of 7 days. It scans the peaks and troughs on the curve point by point, identifying the time span between adjacent peaks and troughs. If the spans of multiple consecutive periods are close to 7 days—for example, if a beverage product has consistently shown high sales on Saturdays and low sales on Tuesdays for the past 8 weeks, with the peak-trough interval remaining stable within 5 to 7 days—then this time span is defined as a periodic fluctuation range. Subsequently, the start and end patterns of this periodic fluctuation range are used as a benchmark. Historical sales records are rhythmically aligned by mapping the start point of each cycle to the peak time of Saturday each week. Sales data from different cycles are sliced ​​according to the same time phase to form multiple standardized cycle segments of consistent length. For example, after aligning the 8 weekly cycles, they are cut into 8 "Saturday to Friday" 7-day segments. The daily average of these standardized cycle segments is then calculated to obtain the average sales percentage of each day in each cycle, such as 140% on Saturday, 130% on Sunday, and 80%-90% during the week. Finally, based on this distribution characteristic, the cyclical demand pattern of the product is summarized, and the demand fluctuation pattern of each period in the weekly cycle is clarified for subsequent estimation of the expected demand in future periods.

[0034] In a specific embodiment, determining the estimated demand for future periods based on cyclical demand patterns includes: Based on the cyclical demand pattern, the basic sales curve is extrapolated to obtain a trend extension line. Future time period nodes are marked on the trend extension line to obtain predicted time nodes. The cyclical fluctuation range is extended using these predicted time nodes to obtain future fluctuation ranges. The peak and trough positions of these future fluctuation ranges are estimated to obtain peak and trough prediction points. Based on these peak and trough prediction points, the future demand for the product is estimated to obtain the predicted demand.

[0035] Specifically, determining the estimated demand for future periods based on cyclical demand patterns includes: extrapolating the basic sales curve based on the cyclical demand patterns to obtain a trend extension line; marking future period nodes on the trend extension line to obtain prediction time nodes; extending the cyclical fluctuation range through the prediction time nodes to obtain a future fluctuation range; and estimating the peak and trough positions of the future fluctuation range to obtain peak and trough prediction points; and estimating the future demand for the product based on the peak and trough prediction points to obtain the estimated demand. This process first, based on the obtained cyclical demand patterns, involves extrapolating the aforementioned steps... The trend term in the generated basic sales curve is mathematically modeled, and time series forecasting methods such as linear regression, exponential smoothing, or ARIMA are used to extrapolate the trend term, extending the historical trend into the future time axis to form a continuous numerical curve, which is the trend extension line. This trend extension line reflects the long-term growth or decline direction of the overall sales of the product. For example, if the average weekly sales of a beverage product have increased by 2% in the past 12 weeks, the trend extension line will extend into the future according to this growth rate. Subsequently, the trend extension line is marked according to a preset time granularity, such as daily or weekly, marking the specific time points of the next week or the next seven days. These marked positions are the forecast time. Interval nodes are used to define the time range for subsequent demand forecasts. Then, based on these forecast time nodes, the previously identified cyclical fluctuation ranges are sequentially extended over time. For example, if the cyclical fluctuation range for this beverage product is 7 days, and the most recent fluctuation range ended this Sunday, then the future fluctuation range is the continuous 7-day period from next Monday to next Sunday. This range is the future fluctuation range. Subsequently, within this future fluctuation range, the position is estimated based on the relative positions of peaks and troughs in historical cycles. For example, if historical data shows that Saturday is the peak point and Tuesday is the trough point, then the peak and trough prediction points are marked on the corresponding dates of the following week, i.e., next Saturday is the peak point. The peak is estimated at 100 units, and next Tuesday is estimated at 100 units. Then, the fluctuation range ratio of each time point in the historical standardized cycle segment is combined with the base value. For example, the peak is 1.4 times the base value and the trough is 0.8 times the base value. The baseline value of the trend extension line at the prediction time point is multiplied by the corresponding proportional coefficient to calculate the specific demand at each peak and trough estimated point. For example, if the baseline value of the trend extension line at next Saturday is 1000 units, then multiplying by 1.4 will give the estimated demand for that day as 1400 units. Finally, the demand at all prediction time points is calculated and summarized to form a complete estimated demand sequence for future periods. This sequence is used for subsequent supply and demand difference analysis with the current inventory balance in the warehouse.

[0036] In a specific embodiment, the step of performing supply-demand difference analysis based on the remaining inventory of goods and the estimated demand to obtain the supply-demand difference analysis results includes: The remaining inventory of the goods is summarized by multiple goods to obtain total inventory data, and the estimated demand is broken down by goods and future time periods to obtain demand segmentation data. Based on the total inventory data and the demand segmentation data, supply and demand are matched on a product-by-product and time-by-time basis to obtain a preliminary supply and demand comparison table. When there are discrepancies in the preliminary supply and demand comparison table, the discrepancies are marked in the preliminary supply and demand comparison table to generate a discrepancy mark table, and the discrepancy mark table is subjected to a discrepancy quantification analysis to obtain the supply and demand difference analysis results.

[0037] Specifically, the supply-demand difference analysis based on the remaining inventory of goods and the estimated demand to obtain the supply-demand difference analysis results includes summarizing the remaining inventory of goods for multiple goods to obtain total inventory data, and splitting the estimated demand by goods and future time periods in a two-dimensional manner to obtain demand segmentation data; and performing supply-demand matching on a goods-by-goods and time-by-time basis based on the total inventory data and the demand segmentation data to obtain a preliminary supply-demand comparison table.When the preliminary supply and demand comparison table contains discrepancies, these discrepancies are marked to generate a discrepancy marking table. A quantitative analysis of the discrepancy marking table is then performed to obtain the supply and demand difference analysis results. This process first reads the real-time inventory data of all relevant products from the warehouse management system, aggregates the available inventory quantity of each product to form an inventory data set covering multiple products, and then performs a unified summary operation on the remaining inventory of products in this set to generate total inventory data reflecting the overall inventory status. For example, a regional central warehouse may contain three products: beverages A, B, and C, with current remaining inventory of 800 units, 600 units, and so on. Starting with 500 units, after aggregating multiple products, the total inventory data is 1900 units. Simultaneously, the estimated demand output from the previous steps is further refined according to two dimensions: product type and future time interval. For example, the estimated demand for beverage A over the next 7 days is 300, 250, 280, 320, 350, 450, and 400 units per day from Monday to Sunday; for beverage B, it's 200, 220, 240, 230, 260, 300, and 280 units; and for beverage C, it's 180, 190, 200, 210, 220, 240, and 230 units. This forms structured demand breakdown data, including product number, date, and... Based on daily demand, the total inventory data and detailed demand data are then input into the matching engine. The engine retrieves the remaining inventory for each product code and compares it daily with the demand for that product in each future time period. This constructs a preliminary supply-demand comparison table containing four columns: product, time period, remaining inventory, and estimated demand. For example, the table might show that beverage A has a remaining inventory of 800 units on Monday and an estimated demand of 300 units (the difference should not be negative). On Saturday, the estimated demand for beverage A is 450 units. If the remaining inventory after the day's consumption is 400 units, there will be a difference of 50 units. This difference constitutes the surplus item. The system then applies this item to the preliminary supply-demand comparison table. Records showing supply-demand imbalances are marked, for example, by highlighting them in red or adding a "shortage" label, generating a difference marking table. Then, each difference in the difference marking table is quantified. If inventory exceeds demand, the difference is positive, indicating a surplus; if inventory is less than demand, the difference is negative, indicating a shortage. For example, beverage A's difference next Saturday might be -50 units, while beverage B's might be +20 units next Wednesday. All quantified differences form a set of values. After being aggregated and processed by the system, the final supply-demand difference analysis result is formed. This result includes the specific difference value for each product in each future time period, serving as a direct basis for determining whether to trigger inventory clearance strategies or stockout response procedures.

[0038] In a specific embodiment, the step of formulating an inventory reduction strategy based on the surplus of commodity inventory includes: The surplus of the commodity inventory is classified according to the shelf life of the commodity to obtain a near-expiry classification table, and the surplus quantity in the near-expiry classification table is marked with a time limit for digestion to obtain a time constraint table. Based on the aforementioned time constraint table, the unit cost of the goods is calculated based on various preset digestion methods to obtain a cost accounting sheet. The cost accounting sheet is then matched with the historical sales increase data of the corresponding digestion method to obtain the cost-effectiveness ratio. Based on the cost-effectiveness ratio, the digestion methods of each commodity are combined and calculated to obtain a set of combined solutions. Based on a preset AI model, the total digestion cost and expected digestion volume in the set of combined solutions are balanced and verified to obtain a feasible solution group. The feasible schemes are prioritized according to their digestion cycle length to obtain the scheme ranking results, and the optimal combination scheme is selected from the scheme ranking results as the inventory digestion strategy.

[0039] Specifically, the inventory clearance strategy based on the surplus of commodity inventory includes: classifying the surplus of commodity inventory according to the shelf life of the commodities to obtain a near-expiration classification table; marking the clearance time limit for the surplus quantity in the near-expiration classification table to obtain a time constraint table; based on the time constraint table, calculating the unit cost of commodities based on various preset clearance methods to obtain a cost accounting sheet; matching the cost accounting sheet with the historical sales increase data of the corresponding clearance method to obtain a cost-effectiveness ratio; performing combination calculations on clearance methods for each commodity based on the cost-effectiveness ratio to obtain a set of combination schemes; and balancing and verifying the total clearance cost and expected clearance volume in the set of combination schemes based on a preset AI model to obtain a group of feasible schemes; prioritizing the group of feasible schemes according to the length of the clearance cycle to obtain a scheme ranking result; and selecting the optimal combination scheme from the scheme ranking result as the inventory clearance strategy. This process first involves retrieving the batch information of a commodity after confirming that there is a surplus of inventory. The shelf-life data is used to categorize surplus goods into different levels based on the number of days remaining until their expiration date. For example, goods with less than 30 days remaining are classified as Level 1 (near expiration), those with 31 to 60 days remaining are Level 2, and those with more than 61 days remaining are Level 3, creating a near-expiration grading table. For instance, if a beverage product currently has 300 surplus units, 100 units with 25 days remaining are classified as Level 1, and the remaining 200 units with 50 days remaining are classified as Level 2. Then, based on this near-expiration grading table, a time limit is set for each type of surplus quantity to be sold off, i.e., a sales deadline. Level 1 surpluses must be processed within 7 days, and Level 2 within 14 days. This generates a time constraint table, which serves as the time boundary for subsequent strategy execution. Then, for various preset digestion methods, including online promotions, bundled sales, cross-warehouse transfers, and offline special sales, the direct costs required to process one unit of goods under each method are calculated, such as promotional discount losses, logistics and transfer fees, and personnel execution costs, forming a cost accounting sheet. For example, the unit cost of an online "buy one get one free" promotion is 2 yuan / item, and the unit cost of cross-warehouse transfer is 1 yuan / item.The cost is calculated at 5 yuan per item. This cost accounting sheet is then matched with historical data on the sales increase of similar products for the corresponding digestion method. For example, "buy one get one free" once increased the daily sales of similar beverages by 120%, and the average consumption cycle for items transferred to high-demand areas was shortened to 5 days. The cost-effectiveness ratio of each method is then calculated, i.e., the efficiency of expected sales growth or inventory reduction per unit cost. Based on this cost-effectiveness ratio, the system performs combined calculations for all feasible digestion methods. For example, it tries a combination of "online promotion + in-store special offer" for 100 near-expiry items in the first-tier category, and a combination of "regional transfer + bundled sales" for 200 items in the second-tier category, generating multiple possible combination schemes. These schemes are then input into a preset AI. The optimization model uses minimizing total digestion cost and maximizing expected digestion volume as its objective functions. Feasibility is verified by incorporating operational constraints such as storage capacity and channel capacity. Solutions that satisfy both timeliness constraints and achieve a balance between cost and effectiveness are selected to form a feasible solution group. Finally, the solutions in the feasible solution group are sorted in ascending order of their expected digestion cycle, with shorter cycles ranking higher. The solution with the highest ranking and comprehensive score is selected as the final inventory digestion strategy for implementation. Furthermore, a pre-set AI model is used to balance the total digestion cost and expected digestion volume in the combined solution set to obtain the feasible solution group. Specifically, the AI... The model includes a multi-scale feature fusion layer, a dual-objective attention decision layer, and a constraint verification layer. The total digestion cost and expected digestion amount from the combined scheme set are input into the multi-scale feature fusion layer, which contains three parallel convolutional branches (using 5×5, 3×3, and 1×1 convolutional kernels respectively, with 64 kernels per branch, a stride of 1, and padding of 2). After feature concatenation, the output is obtained through a 128-dimensional fully connected layer, resulting in a cost-digestion fusion feature vector, which includes cost fluctuation features, digestion gradient features, and spatiotemporal correlation features. This cost-digestion fusion feature vector is then input into the dual-objective attention decision layer, which employs a dual-channel self-attention mechanism (8 attention heads each for the cost channel and the digestion channel, with a head dimension of 16). After processing by a layered Transformer encoder (256 hidden dimensions per layer, 512 dimensions of the feedforward network, and Mish activation function), a Pareto optimal solution candidate set is output, which includes a subset of solutions that meet the cost threshold, a subset of solutions that meet the digestion threshold, and a subset of cost-digestion balanced solutions, etc. The Pareto optimal solution candidate set is then input into a constraint verification layer, which uses a gated recurrent unit (GRU) network (128 hidden dimensions, dropout rate 0).3) Using a 10×10 fuzzy matrix, verify each option in the candidate set against near-expiration time constraints, inventory turnover constraints, and channel capacity constraints, outputting a feasibility score (0-100 points). Select options with scores ≥80 points to obtain a feasible option group, which includes the verification passed items, constraint satisfaction degree, and execution priority of each option.

[0040] In a specific embodiment, the step of assessing the criticality and urgency of goods based on the shortage quantification results to obtain a list of key goods and corresponding shortage urgency indices includes: The shortage quantification results are summarized by commodity category to obtain a total commodity shortage table, and recent market dynamics are collected to obtain a set of market influencing factors. Based on the total shortage table of the commodities and the set of market influencing factors, a shortage risk score is given for each category of commodities to obtain the risk score result. The urgency of the goods is weighted by risk scoring results to obtain weighted urgency parameters, and the goods are ranked based on the weighted urgency parameters to obtain an urgency ranking table. Based on the urgency ranking table, products with urgency higher than a preset threshold are selected to generate a key product list, and each product in the key product list is assigned a corresponding shortage urgency index based on the risk scoring results.

[0041] Specifically, the process of assessing the criticality and urgency of goods based on the shortage quantification results to obtain a list of key goods and corresponding shortage urgency indices includes: summarizing the shortage quantities of the shortage quantification results by goods category to obtain a total shortage table; collecting recent market dynamics to obtain a set of market influencing factors; scoring the shortage risk of each category of goods based on the total shortage table and the set of market influencing factors to obtain risk score results; calculating the urgency weight of the goods based on the risk score results to obtain weighted urgency parameters; and ranking the goods based on the weighted urgency parameters to obtain an urgency ranking table.Based on the urgency ranking table, products with urgency levels higher than a preset threshold are selected to generate a key product list. Then, based on the risk scoring results, a corresponding shortage urgency index is assigned to each product in the key product list. This process first involves obtaining the shortage quantification results for each product and then classifying and aggregating them according to product category. For example, the shortage quantities of multiple SKUs (such as A, B, and C) under the beverage category are counted and accumulated separately. If beverage A is in short supply of 400 units, beverage B is in short supply of 300 units, and beverage C is in short supply of 100 units, then the total shortage quantity for the beverage category is 800 units, forming a total product shortage table. Simultaneously, this is verified through external data interfaces or internal market data. The market intelligence system collects recent market dynamics information, including the popularity of discussions about this type of product on social media, competitor stockouts, seasonal consumption trends, weather forecasts (such as high-temperature warnings), and the approach of major holidays. This information is structured into a set of market influencing factors. For example, given the current continuous high temperatures and the five-day National Day holiday, the system includes "continued high temperatures" and "pre-holiday stocking peak" as two high-weight factors in the market influencing factor set. Then, combining the total product shortage table with the market influencing factor set, a weighted scoring model is used to score the stockout risk for each product category. The scoring dimensions include the scale of the shortage, the category's sales share, etc. Factors such as customer complaint history and the combined strength of market influencing factors are considered. For example, beverages, due to their large overall shortage volume, accounting for 35% of warehouse shipments, and the significant increase in demand due to high temperatures, received a risk score of 92, while other categories received lower scores. This score is the risk score result. This risk score result is then used as a weighting factor, and its urgency is calculated by combining it with the quantified shortage results of each product. For example, if beverage A has a shortage of 400 units, multiplying it by its category's risk score of 92, and then combining this with its influence coefficient in order fulfillment rate, the weighted urgency parameter is calculated to be 36800. The parameter values ​​for other products are calculated similarly. Finally, all... Products are ranked from highest to lowest based on a weighted urgency parameter, generating an urgency ranking table. Beverage A ranks first, followed by beverage B. A preset threshold is set; products with a weighted urgency parameter higher than 30,000 are considered high urgency. The system filters these products from the urgency ranking table, generating a list of key products. Finally, based on the aforementioned risk scoring results, a shortage urgency index is assigned to each product in the list. For example, the risk score is directly mapped to an index value, making the shortage urgency index for beverage A 92 and for beverage B 85. This index is used for priority determination in subsequent replenishment plans, ensuring resources are allocated to high-risk, high-impact products.

[0042] In a specific embodiment, the step of formulating an optimal replenishment plan for goods based on a list of key goods and the corresponding stockout urgency index, using preset goods constraints, includes: The preset commodity constraints are classified and analyzed to obtain resource constraints and timeliness constraints. The corresponding replenishment cost parameters are matched to the commodities in the key commodity list to obtain the commodity cost table. Based on the stockout urgency index, the replenishment priority of the goods in the commodity cost table is sorted to obtain a priority sequence. The feasibility of the priority sequence is then verified according to the resource constraint item to obtain feasible replenishment items. Based on the feasible replenishment items and the timeliness constraints, the replenishment quantity is adjusted and calculated to obtain the adjusted replenishment quantity. The adjusted replenishment quantity is then summarized and arranged according to the shortage urgency index to obtain the optimal replenishment plan.

[0043] Specifically, the step of formulating an optimal replenishment plan for goods based on a list of key goods and a corresponding shortage urgency index, using preset goods constraints, includes classifying and analyzing the preset goods constraints to obtain resource constraints and timeliness constraints, matching the corresponding replenishment cost parameters for the goods in the list of key goods to obtain a goods cost table; prioritizing the goods in the goods cost table based on the shortage urgency index to obtain a priority sequence, and performing feasibility verification on the priority sequence according to resource constraints to obtain feasible replenishment items;Based on the feasible replenishment items and time constraints, the replenishment quantity is adjusted to obtain the adjusted replenishment quantity. This adjusted replenishment quantity is then summarized and arranged according to the stockout urgency index to obtain the optimal replenishment plan. This process first loads preset product constraints from the supply chain management system. These constraints cover operational limitations such as supplier supply capacity, transportation vehicle capacity, procurement budget ceiling, minimum order quantity per order, and maximum daily warehouse receiving batches. The system categorizes and analyzes these constraints by nature, classifying those involving limited resources such as funds, transportation capacity, and warehousing capacity as resource constraints. For example, a supplier can ship a maximum of 5000 units per week, and transportation capacity... The delivery fleet supports a maximum of 3 deliveries per day. Items with time requirements are categorized as time-sensitive constraints. For example, the supplier's delivery cycle for beverage A is 3 days, and the warehouse has a receiving window from Tuesday to Thursday of the following week; receiving is not available at other times. Then, for each item in the key product list, such as beverages A, B, and C, the system retrieves its corresponding purchase price, transportation rate, loading and unloading costs, etc., to form a structured data table containing product number, unit purchase cost, unit logistics cost, and total replenishment cost—this is the product cost table. Next, the system sorts the key product list items in descending order based on their shortage urgency index; the higher the index, the higher the ranking. For example, for beverages… The urgency index for beverage A is 92, for beverage B it's 85, and for beverage C it's 78. After sorting, a priority sequence is formed with beverage A as the first priority. This priority sequence is then input into the resource verification module item by item, and its replenishment feasibility is verified according to resource constraints. For example, if beverage A needs replenishment of 400 units, which does not exceed the supplier's weekly limit of 5000 units, and there are available transport vehicles, it is considered feasible. However, if the replenishment quantity of a certain product exceeds the budget or transport capacity, it is marked as infeasible. After filtering, feasible replenishment items containing only those meeting the conditions are generated. These feasible replenishment items are then matched with time constraints to determine whether their replenishment cycle meets delivery requirements. For example, if beverage A must be replenished... The goods must arrive by next Thursday, and the supplier's delivery cycle is 3 days. If the order is placed no later than next Monday, the demand can be met. The system dynamically adjusts the replenishment quantity accordingly. For example, the original plan to replenish 400 units might be split into two batches: the first batch of 300 units will arrive within 3 days, while the second batch of 100 units is delayed due to shipping schedules. The adjusted replenishment quantity will then be 300 units. Finally, all adjusted replenishment quantities are summarized and arranged according to their urgency index, forming a structured output file containing the product, replenishment quantity, supplier, estimated arrival time, shipping batch, and total cost. This is the optimal replenishment plan, ensuring priority for high-urgency items, no resource overruns, and timely delivery.

[0044] The above describes the AI-based supply chain demand forecasting and inventory optimization method in the embodiments of the present invention. The following describes the AI-based supply chain demand forecasting and inventory optimization system in the embodiments of the present invention. Please refer to [link / reference]. Figure 2 One embodiment of the AI-based supply chain demand forecasting and inventory optimization system of the present invention includes: The mining module 21 is used to retrieve the sales records of goods from the supply chain and identify the cyclical demand patterns of the corresponding goods based on the sales records. Analysis module 22 is used to determine the estimated demand for future periods based on the cyclical demand pattern, obtain the current inventory balance of goods in the warehouse, and perform supply and demand difference analysis based on the inventory balance and the estimated demand to obtain the supply and demand difference analysis results. The first formulation module 23 is used to formulate an inventory digestion strategy based on the commodity inventory surplus when the result of the supply and demand difference analysis is a commodity inventory surplus. The sorting module 24 is used to assess the criticality and urgency of commodities based on the shortage quantification result when the supply and demand difference analysis result is the shortage quantification result, so as to obtain a list of key commodities and the corresponding shortage urgency index. The second formulation module 25 is used to formulate the optimal replenishment plan for goods based on the list of key goods and the corresponding shortage urgency index through preset goods constraints.

[0045] In this embodiment, the specific implementation of each unit in the above system embodiment is described in the above method embodiment, and will not be repeated here.

[0046] like Figure 3 As shown in the diagram, this embodiment of the invention provides a structural schematic block diagram of a computer device, including: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the above-described AI-based supply chain demand forecasting and inventory optimization method.

[0047] It is evident that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented in this device embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0048] Furthermore, this application also discloses a computer program product or computer program stored in a computer-readable storage medium. A processor of a computer device can read the computer program from the computer-readable storage medium and execute the computer program, causing the computer device to perform the aforementioned AI-based supply chain demand forecasting and inventory optimization method. Similarly, the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0049] The above is a detailed description of the preferred embodiments of the present invention. However, the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention. All such equivalent modifications or substitutions are included within the scope defined by the claims of this application.

Claims

1. An AI-based supply chain demand forecasting and inventory optimization method, characterized in that, Includes the following steps: Retrieve sales records of goods from the supply chain and identify the cyclical demand patterns of the corresponding goods based on the sales records; Based on the cyclical demand pattern, the estimated demand for future periods is determined, the current inventory balance of goods in the warehouse is obtained, and a supply-demand difference analysis is performed based on the inventory balance and the estimated demand to obtain the supply-demand difference analysis results. If the result of the supply and demand difference analysis is a surplus of commodity inventory, then an inventory digestion strategy is formulated based on the surplus of commodity inventory. When the supply and demand difference analysis result is a shortage quantification result, the keyness and urgency of the commodities are assessed and ranked based on the shortage quantification result to obtain a list of key commodities and the corresponding shortage urgency index. By using preset product constraints, the optimal replenishment plan is formulated based on the list of key products and the corresponding shortage urgency index.

2. The AI-based supply chain demand forecasting and inventory optimization method according to claim 1, characterized in that, The step of retrieving sales records of goods from the supply chain and identifying the cyclical demand patterns of the corresponding goods based on the sales records includes: The sales records of the goods are retrieved through the database interface of the supply chain, and the daily sales data in the sales records are decomposed into trends to obtain the basic sales curve. The peaks and troughs of the basic sales curve are identified to obtain the periodic fluctuation range. Based on the periodic fluctuation range, the sales records are rhythmically aligned to obtain standardized periodic segments, and demand patterns are identified in the standardized periodic segments to obtain periodic demand patterns.

3. The AI-based supply chain demand forecasting and inventory optimization method according to claim 2, characterized in that, The method of determining the estimated demand for future periods based on cyclical demand patterns includes: Based on the cyclical demand pattern, the basic sales curve is extrapolated to obtain a trend extension line. Future time period nodes are marked on the trend extension line to obtain predicted time nodes. The cyclical fluctuation range is extended using these predicted time nodes to obtain future fluctuation ranges. The peak and trough positions of these future fluctuation ranges are estimated to obtain peak and trough prediction points. Based on these peak and trough prediction points, the future demand for the product is estimated to obtain the predicted demand.

4. The AI-based supply chain demand forecasting and inventory optimization method according to claim 1, characterized in that, The supply-demand gap analysis based on the remaining inventory and the estimated demand yields the following results: The remaining inventory of the goods is summarized by multiple goods to obtain total inventory data, and the estimated demand is broken down by goods and future time periods to obtain demand segmentation data. Based on the total inventory data and the demand segmentation data, supply and demand are matched on a product-by-product and time-by-time basis to obtain a preliminary supply and demand comparison table. When there are discrepancies in the preliminary supply and demand comparison table, the discrepancies are marked in the preliminary supply and demand comparison table to generate a discrepancy mark table, and the discrepancy mark table is subjected to a discrepancy quantification analysis to obtain the supply and demand difference analysis results.

5. The AI-based supply chain demand forecasting and inventory optimization method according to claim 1, characterized in that, The inventory clearance strategy based on the commodity inventory surplus includes: The surplus of the commodity inventory is classified according to the shelf life of the commodity to obtain a near-expiry classification table, and the surplus quantity in the near-expiry classification table is marked with a time limit for digestion to obtain a time constraint table. Based on the aforementioned time constraint table, the unit cost of the goods is calculated based on various preset digestion methods to obtain a cost accounting sheet. The cost accounting sheet is then matched with the historical sales increase data of the corresponding digestion method to obtain the cost-effectiveness ratio. Based on the cost-effectiveness ratio, the digestion methods of each commodity are combined and calculated to obtain a set of combined solutions. Based on a preset AI model, the total digestion cost and expected digestion volume in the set of combined solutions are balanced and verified to obtain a feasible solution group. The feasible schemes are prioritized according to their digestion cycle length to obtain the scheme ranking results, and the optimal combination scheme is selected from the scheme ranking results as the inventory digestion strategy.

6. The AI-based supply chain demand forecasting and inventory optimization method according to claim 1, characterized in that, The process of assessing the criticality and urgency of goods based on the shortage quantification results, resulting in a list of key goods and corresponding shortage urgency indices, includes: The shortage quantification results are summarized by commodity category to obtain a total commodity shortage table, and recent market dynamics are collected to obtain a set of market influencing factors. Based on the total shortage table of the commodities and the set of market influencing factors, a shortage risk score is given for each category of commodities to obtain the risk score result. The urgency of the goods is weighted by risk scoring results to obtain weighted urgency parameters, and the goods are ranked based on the weighted urgency parameters to obtain an urgency ranking table. Based on the urgency ranking table, products with urgency higher than a preset threshold are selected to generate a key product list, and a corresponding shortage urgency index is assigned to each product in the key product list based on the risk scoring results.

7. The AI-based supply chain demand forecasting and inventory optimization method according to claim 1, characterized in that, The process of formulating an optimal replenishment plan for goods based on a list of key goods and their corresponding stockout urgency index, using preset goods constraints, includes: The preset commodity constraints are classified and analyzed to obtain resource constraints and timeliness constraints. The corresponding replenishment cost parameters are matched to the commodities in the key commodity list to obtain the commodity cost table. Based on the stockout urgency index, the replenishment priority of the goods in the commodity cost table is sorted to obtain a priority sequence. The feasibility of the priority sequence is then verified according to the resource constraint item to obtain feasible replenishment items. Based on the feasible replenishment items and the timeliness constraint, the replenishment quantity adjustment calculation is performed to obtain the adjusted replenishment quantity. The adjusted replenishment quantity is then summarized and arranged according to the shortage urgency index to obtain the optimal replenishment plan.

8. An AI-based supply chain demand forecasting and inventory optimization system, characterized in that, include: The mining module is used to retrieve the sales records of goods from the supply chain and identify the cyclical demand patterns of the corresponding goods based on the sales records. The analysis module is used to determine the estimated demand for future periods based on the cyclical demand pattern, obtain the current inventory balance of goods in the warehouse, and perform supply and demand difference analysis based on the inventory balance and the estimated demand to obtain the supply and demand difference analysis results. The first formulation module is used to formulate an inventory digestion strategy based on the commodity inventory surplus when the supply and demand difference analysis result is a commodity inventory surplus. The sorting module is used to assess the criticality and urgency of commodities based on the shortage quantification result when the supply and demand difference analysis result is a shortage quantification result, so as to obtain a list of key commodities and the corresponding shortage urgency index. The second module is used to formulate the optimal replenishment plan for goods based on a list of key goods and the corresponding shortage urgency index, according to preset goods constraints.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images