Sensor-based dynamic demand forecasting and supply chain adjustment method

Abstract

The application discloses a sensor-based dynamic demand prediction and supply chain adjustment method, relates to the technical field of supply chain adjustment of logistics and storage, and comprises the following steps: analyzing demand behavior data and outputting a prediction result of dynamic demand prediction; and analyzing the logistics data based on the prediction result and outputting a supply chain adjustment strategy. The demand behavior vector model is constructed by fusing the length deep analysis of browsing, the comment tendency empowerment, the dynamic urgency quantification of shopping carts and the consumption capacity, high-precision quantification of the purchase willingness of users and prospective prediction of the demand quantity are realized, and meanwhile, a dynamic inventory control system is designed based on the prediction result, a regional precise allocation strategy is realized through a multi-constraint intelligent optimization algorithm, the accuracy of demand prediction and the inventory turnover rate are improved, and the comprehensive cost of storage and logistics is greatly reduced.

Description

Technical Field

[0001] This invention relates to the field of logistics warehousing supply chain adjustment technology, and in particular to a sensor-based dynamic demand forecasting and supply chain adjustment method. Background Technology

[0002] With the rapid development of the Internet of Things, big data, and sensor technologies, e-commerce and modern retail have accumulated massive amounts of user behavior and logistics data. How to effectively mine this data and apply it to demand forecasting and supply chain adjustments has become an important research direction in the supply chain field.

[0003] Existing methods utilize user behavior data for demand forecasting and supply chain scheduling. For example, patent CN120471412A discloses an AI-based virtual goods supply chain scheduling system and method. This system collects user behavior data and inventory status data from various suppliers through a data acquisition module, constructs a prediction model using an LSTM network and attention mechanism, obtains demand forecast data based on user behavior data, and then determines weight adjustment coefficients based on inventory risk levels to schedule inventory supply for each supplier. However, this patent suffers from limitations: its analysis of behavioral data largely remains at a macro level, lacking in-depth analysis of individual behavioral data, and thus failing to accurately reflect user purchasing intentions. Furthermore, existing methods lack tiered processing of inventory redundancy and shortages, making it difficult to achieve optimal inventory adjustments. Summary of the Invention

[0004] In view of this, the present invention provides a sensor-based dynamic demand forecasting and supply chain adjustment method to solve the problems mentioned in the background art.

[0005] The objective of this invention can be achieved through the following technical solution: a sensor-based dynamic demand forecasting and supply chain adjustment method, comprising:

[0006] Collect demand behavior data and logistics data through the Internet of Things and sensors;

[0007] The demand behavior data is analyzed to output the prediction results of dynamic demand forecasting;

[0008] Based on the prediction results, the logistics data is analyzed to output supply chain adjustment strategies; wherein the supply chain adjustment strategies include adjustment strategy one, adjustment strategy two, and adjustment strategy three.

[0009] In some embodiments, the analysis of the demand behavior data specifically includes:

[0010] The browsing data is parsed to obtain the item details browsing time, positive review viewing time, negative review viewing time, and other comment viewing time, and then standardized.

[0011] Obtain the detail view values ​​of all purchased items for a user and take the average value to obtain the average purchase view value. Compare and analyze the detail view values ​​with the average purchase view value and output the detail view coefficient.

[0012] If the details browsing coefficient is less than the minimum coefficient threshold, the system will redirect to the shopping cart data analysis, and the item browsing index will be set to the default value.

[0013] If the details browsing coefficient is greater than or equal to the minimum coefficient threshold, then obtain the weights corresponding to the positive review viewing value, negative review viewing value, and other comment viewing value, calculate the sum of the details browsing coefficient and the comment viewing coefficient, output the item browsing index, and jump to the add to cart data analysis;

[0014] By analyzing the data from items added to the shopping cart, an "add to cart index" for each item is output.

[0015] A demand behavior vector is constructed based on item browsing index, shopping cart addition index, and monthly spending amount.

[0016] Calculate the dot product of the demand behavior vector and the behavior weight vector, and output the demand value of the item;

[0017] When the demand value is greater than or equal to the demand threshold, users are marked as demanding users, and the number of demanding users is counted to obtain the predicted demand quantity of the item, which is recorded as the prediction result.

[0018] In some embodiments, the shopping cart data analysis includes:

[0019] Parse the shopping cart data to obtain the number of days items have been added to the cart, and obtain the number of items added to the cart by the user and the average number of days since the items were purchased.

[0020] Compare the number of days an item is added to the cart to the average number of days it is purchased:

[0021] If the number of days an item has been added to the cart is less than the average number of days it has been purchased, then the number of days the item has been added to the cart is subtracted from the average number of days it has been purchased to get the remaining number of days. The remaining number of days is then multiplied by the corresponding remaining coefficient to output the cart addition base.

[0022] If the number of days an item has been added to the cart is greater than or equal to the average number of days it has been purchased, then the number of days the item has been added to the cart is subtracted from the average number of days it has been purchased to get the number of days the item has been over-purchased. The reciprocal of the number of days the item has been over-purchased is recorded as the cart addition base.

[0023] The "Add to Cart Index" is calculated by dividing the base number of items added to the cart by the total number of items added to the cart.

[0024] In some embodiments, the analysis of the logistics data based on the prediction results specifically includes:

[0025] Obtain orders for items, identify the orders to obtain the sales quantity of items, and add the sales quantity to the predicted demand quantity to obtain the pre-sale quantity of items.

[0026] Compare the pre-sale quantity with the total warehouse quantity:

[0027] If the total quantity in storage is greater than the pre-sale quantity, the difference between the two is calculated to obtain the over-pre-sale quantity of the item. If the over-pre-sale quantity exceeds the maximum value of quantity range one, the adjustment strategy one corresponding to the item is executed.

[0028] If the excess quantity falls within quantity range one, then the corresponding adjustment strategy two for the item will be executed;

[0029] If the total quantity in storage is less than or equal to the pre-sale quantity, the difference between the two is calculated to obtain the lower pre-sale quantity of the item. If the lower pre-sale quantity exceeds the lower pre-sale quantity threshold, the corresponding adjustment strategy three for the item is executed.

[0030] In some embodiments, implementing adjustment strategy one corresponding to the item includes:

[0031] Set constraints and construct several adjustment schemes for the item;

[0032] Each adjustment plan is coded;

[0033] Targeted transfers between multiple warehouses are represented by matrix encoding;

[0034] Initial population generation;

[0035] Construct a fitness function and calculate the fitness value of the adjustment scheme;

[0036] Randomly select several adjustment schemes, select the adjustment scheme with the highest fitness value and perform cross operation, repeat the random selection step until a certain number of adjustment schemes with the highest fitness value are selected.

[0037] Perform cross operations on the selected adjustment schemes and output sub-adjustment schemes;

[0038] Calculate the total cost required for each sub-adjustment plan. If the total cost required is less than or equal to the preset budget threshold, mark the sub-adjustment plan as the optimal adjustment plan, output the optimal adjustment plan and execute it.

[0039] In some embodiments, the fitness function is specifically:

[0040] YFj=1 / (Mj+λ×P penalty );

[0041] Where Mj is the total cost required for the adjustment plan, λ is the penalty coefficient, and P penalty To constrain violations and penalties; , where su1i represents the safety stock quantity of items in warehouse i, su2i represents the adjusted stock quantity of items in warehouse i, and su3 is the excess quantity minus the maximum value of the quantity range.

[0042] In some embodiments, the adjustment strategy two corresponding to the execution item includes:

[0043] The system analyzes the pre-sale quantity and the corresponding user's delivery location in the order, marks the delivery location, obtains the coverage area of ​​each warehouse, counts the quantity of delivery locations within the warehouse's coverage area, identifies the sales quantity of the corresponding item in the warehouse, compares the sales quantity with the warehouse's inventory quantity, if the inventory quantity is greater than the sales quantity and the difference is greater than a preset quantity, subtracts the safety stock quantity from the warehouse's inventory quantity to obtain the item transfer quantity, and marks the warehouse as a pending transfer warehouse; if the inventory quantity is less than the sales quantity and the difference is greater than a preset quantity, subtracts the inventory quantity from the warehouse's sales quantity to obtain the item receiving quantity, and marks the warehouse as a pending receipt warehouse; based on the pending transfer warehouse and the transfer quantity, a logistics order for the item is generated, and the corresponding item transfer quantity is sent to the pending receipt warehouse, so that the inventory quantity of the item in the pending receipt warehouse is equal to the sales quantity in the corresponding coverage area of ​​the pending receipt warehouse.

[0044] In some embodiments, the adjustment strategy three corresponding to the execution item specifically includes:

[0045] Implementation of Adjustment Strategy Two is applied to each warehouse. Then, the shortage quantity of corresponding items after allocation is determined. If the shortage quantity is greater than or equal to the shortage threshold one, a severe shortage instruction is generated. The item name, warehouse location, and severe shortage instruction are sent to the smart terminal of the corresponding manufacturer. Upon receiving the severe shortage instruction, the manufacturer immediately generates a corresponding production order and delivers the items corresponding to the shortage quantity to the warehouse. If the shortage quantity is greater than or equal to the shortage threshold two but less than the shortage threshold one, a moderate shortage instruction is generated. The item name, warehouse location, and moderate shortage instruction are sent to the smart terminal of the corresponding manufacturer. Upon receiving the moderate shortage instruction, the manufacturer produces the corresponding shortage quantity of items within a preset time period (one week or half a month) and delivers them to the warehouse within the preset time period. When the shortage quantity is less than the shortage threshold two, a minor shortage instruction is generated and sent to the manufacturer's smart terminal; no replenishment is performed.

[0046] Compared with the prior art, the beneficial effects of the present invention are:

[0047] 1. This invention uses the average purchase browsing value as a benchmark for individual user consumption decisions. By comparing the current item details browsing time with the consumption decision benchmark, it achieves sensitive capture and value amplification of users' high-attention behaviors, effectively identifies the user's likelihood of purchasing the item, and improves the accuracy and sensitivity of demand signal extraction.

[0048] 2. This invention analyzes users' sensitivity to negative information and its potential inhibitory effect on purchasing decisions by measuring the viewing time of positive reviews, negative reviews, and other comments, and assigning the largest negative weight to negative reviews. It then synthesizes the detail browsing coefficient and the comment viewing coefficient to output an item browsing index, reflecting the intensity of users' attention and decision-making tendencies during the information gathering stage, thus mapping browsing behavior to purchasing intention.

[0049] 3. This invention constructs an "Add to Cart Index" by segmenting and mapping the difference between the number of days a user adds items to their cart and the user's historical average purchase days, and then dividing this difference by the number of items added to the cart. This "Add to Cart Index" not only reflects the urgency with which users convert their intentions into actual purchases, but also considers the dilution effect of competition among items in the cart. The more items in the cart, the more diluted the index becomes, indicating a lower probability of purchasing that item, thus making the judgment of the user's current purchase intention more accurate.

[0050] 4. This invention uses matrix encoding to intuitively represent the directional allocation relationship between multiple warehouses, and adopts a hybrid initial population generation strategy to ensure the diversity of solution space exploration while using heuristic information to accelerate convergence. By constructing a fitness function with multiple constraints and an evaluation model with a penalty term, it ensures that the final output optimal adjustment scheme achieves the best balance between total cost and operational feasibility.

[0051] 5. This invention addresses the technical problems in traditional supply chain management, such as relying on extensive historical data for demand forecasting, lacking the ability to capture user behavioral intentions, and the static lag in inventory allocation and production replenishment, making it difficult to cope with structural supply-demand imbalances and dynamic market fluctuations. It constructs a demand behavior vector model that integrates in-depth analysis of browsing time, weighted comment preferences, dynamic urgency measurement of shopping carts, and purchasing power. This model achieves high-precision quantification of user purchasing intentions and forward-looking forecasting of demand quantities. Simultaneously, based on the forecast results, a dynamic inventory control system is designed. Through multi-constraint intelligent optimization algorithms, a precise regional allocation strategy is implemented to improve the accuracy of demand forecasting and inventory turnover rate, significantly reducing overall warehousing and logistics costs. Attached Figure Description

[0052] Further details, features, and advantages of this application are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which:

[0053] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0054] Several embodiments of this application will now be described in more detail with reference to the accompanying drawings to enable those skilled in the art to implement this application. This application may be embodied in many different forms and for various purposes and should not be limited to the embodiments set forth herein. These embodiments are provided to make this application thorough and complete, and to fully convey the scope of this application to those skilled in the art. The embodiments described do not limit this application.

[0055] Unless otherwise defined, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It will be further understood that terms such as those defined in commonly used dictionaries shall be interpreted as having a meaning consistent with their meaning in the relevant field and / or the context of this specification, and shall not be interpreted in an idealized or overly formal sense unless expressly defined herein.

[0056] Please see Figure 1 Sensor-based dynamic demand forecasting and supply chain adjustment methods include:

[0057] Demand behavior data and logistics data are collected through the Internet of Things (IoT) and sensors and stored on servers. Demand behavior data includes user browsing data, shopping cart data, and purchase data for specific items. Logistics data includes order information, warehouse location, inventory quantity, and transportation information for items. By leveraging IoT and sensor technologies, real-time and automated collection of demand behavior and logistics data is achieved, ensuring the data source is comprehensive, real-time, and traceable.

[0058] The analysis of demand behavior data outputs dynamic demand forecasting results, specifically including:

[0059] The browsing data is parsed to obtain the item details browsing time, positive review viewing time, negative review viewing time, and other comment viewing time. If any of the durations is less than the minimum duration threshold or there is no duration, the default duration is used to facilitate subsequent analysis and execution, and to avoid the inability to analyze the results due to excessively short or nonexistent durations.

[0060] The time spent browsing item details, positive reviews, negative reviews, and other comments is processed using the Z-Score standardization method to obtain the details browsing value, positive review viewing value, negative review viewing value, and other comment viewing value. Duration information for dimensions such as details, positive reviews, and negative reviews is extracted, and error tolerance is performed based on a minimum duration threshold to effectively solve the analysis breakpoint problem caused by data sparsity or anomalies. The Z-Score standardization method eliminates the interference of different units and data distribution differences on the analysis results, making browsing behavior across users and items comparable and providing high-quality input features for subsequent analysis.

[0061] The system obtains the average view count of all purchased items by the user and calculates the mean value. It then compares the average view count with the mean view count: if the average view count is greater than or equal to the mean view count, a preset view weight 1 is applied, and the value is multiplied by the preset view weight 1 to obtain the view count coefficient; if the average view count is less than the mean view count, a preset view weight 2 is applied, and the value is multiplied by the preset view weight 2 to obtain the view count coefficient. The preset view weight 1 is greater than or equal to twice the preset view weight 2. Using the mean view count as a benchmark for the user's personal consumption decisions, and comparing the current item's view count duration with this benchmark, the system can sensitively capture and amplify the value of high-attention user behaviors, effectively identifying the user's likelihood of purchasing an item and improving the accuracy and sensitivity of demand signal extraction.

[0062] If the details browsing coefficient is less than the minimum threshold, the system redirects to the "Add to Cart" data analysis, and the item browsing index uses the default value. If the details browsing coefficient is greater than or equal to the minimum threshold, the system obtains the weights corresponding to positive review views, negative review views, and other review views. These weights are then multiplied by their respective weights and summed to output the review view coefficient. The weights for positive review views and other review views are positive, while the weight for negative review views is negative. The absolute value of the negative review view weight is greater than the weight of the positive review view weight, and the weight of the positive review view weight is greater than the weight of other review views. The sum of the details browsing coefficient and the review view coefficient is calculated to output the item browsing index, and the system redirects to the "Add to Cart" data analysis. By analyzing the viewing duration of positive reviews, negative reviews, and other reviews, and assigning the largest negative weight to the negative review view value, the system analyzes the user's sensitivity to negative information and its potential inhibitory effect on purchase decisions. The details browsing coefficient and the review view coefficient are combined to output the item browsing index, reflecting the user's attention intensity and decision-making tendency during the information gathering stage. This mapping from browsing behavior to purchase intention provides behavioral characteristics for demand forecasting.

[0063] The analysis of shopping cart data includes: parsing the shopping cart data to obtain the number of days items have been added to the cart, obtaining the number of items added to the cart by the user, and the average purchase days for the corresponding items in the cart; comparing the number of days items have been added to the cart with the average purchase days: if the number of days items have been added to the cart is less than the average purchase days, the remaining days are obtained by subtracting the number of days items have been added to the cart from the average purchase days, and the remaining days are multiplied by the corresponding remaining coefficient to output the shopping cart base; if the number of days items have been added to the cart is greater than or equal to the average purchase days, the over-purchase days are obtained by subtracting the average purchase days from the number of days items have been added to the cart, and the reciprocal of the over-purchase days is recorded as the shopping cart base; dividing the shopping cart base by the number of items added to the cart yields the shopping cart index; and constructing the shopping cart index by segmenting the difference between the number of days items have been added to the cart and the user's historical average purchase days, and dividing by the number of items added to the cart. The Add to Cart Index not only reflects the urgency with which users convert their intentions into actual purchases, but also takes into account the dilution effect of competition among items in the cart. The more items there are, the more diluted the index becomes, indicating a lower probability of purchasing that item, making the judgment of users' current purchase intentions more accurate.

[0064] The purchase data is analyzed to obtain the user's monthly spending amount; a demand behavior vector is constructed based on the monthly spending amount, item browsing index, and shopping cart addition index. , Where Ct, Vt, and At are the values ​​of monthly spending amount, item browsing index, and add-to-cart index, respectively; obtain the preset behavior weight vector. , , , , The weights are equal to the sum of the three.

[0065] Calculate the dot product of the demand behavior vector and the behavior weight vector, and output the demand value X of the item, i.e.:

[0066] ;

[0067] When the demand value is greater than or equal to the demand threshold, users are marked as demanding users, and the number of demanding users is counted to obtain the predicted demand quantity of the item, which is recorded as the prediction result.

[0068] The monthly spending amount, browsing index, and shopping cart index are used to construct a demand behavior vector, which is then multiplied by a preset behavior weight vector to output the demand value of the item. This allows for the analysis of demand intensity based on users' spending power, cognitive preferences, and urgency of action, overcoming the one-sidedness of single-indicator evaluation. This accurately identifies the user group that purchases the item, making the predicted demand quantity closer to the actual quantity, thereby improving the reliability of demand forecasting.

[0069] Based on the forecast results, logistics data is analyzed to output supply chain adjustment strategies, specifically including:

[0070] Obtain orders for items, identify the orders to obtain the sales quantity of items, and add the sales quantity to the predicted demand quantity to obtain the pre-sale quantity of items.

[0071] Obtain the storage quantity of each item in each warehouse and summarize it to obtain the total storage quantity. Compare the pre-sale quantity with the total storage quantity. If the total storage quantity is greater than the pre-sale quantity, calculate the difference between the two to obtain the over-pre-sale quantity of the item. If the over-pre-sale quantity exceeds the maximum value of quantity range one, execute the corresponding adjustment strategy one for the item. For inventory redundancy, use an intelligent optimization algorithm under multiple constraints to solve the optimal combination of allocation, promotion and return scheme, effectively control inventory holding costs, and prevent capital backlog and storage waste.

[0072] If the excess quantity falls within quantity range one, then the corresponding adjustment strategy two for the item will be implemented; for situations where inventory and regional demand are basically matched, a refined regional allocation based on geographical radiation area is adopted, which accurately realizes the pre-deployment of inventory and allocates goods near the demand terminal in advance, which can significantly improve order fulfillment efficiency and reduce last-mile logistics and distribution costs.

[0073] If the total warehouse quantity is less than or equal to the pre-sold quantity, the difference between the two is calculated to obtain the lower pre-sale quantity of the item. If the lower pre-sale quantity exceeds the lower pre-sale quantity threshold, the corresponding adjustment strategy three for the item is executed. For inventory shortages, a three-tiered shortage warning and differentiated response mechanism (severe / moderate / minor) is constructed. Based on the severity of the shortage, a tiered supply response is dynamically triggered, ranging from emergency production and planned production to no-processing, ensuring market supply while reasonably balancing production and replenishment costs.

[0074] It should be noted that the values ​​at both ends of the quantity range are included. The quantity range is defined by those skilled in the art, and the minimum value of both is zero.

[0075] The first adjustment strategy for the corresponding item includes: setting constraints and constructing several adjustment plans for the item. These constraints include inventory balance constraints, safety stock constraints, capacity constraints, non-negative constraints, and excess inventory disposal constraints. The adjustment plans include inter-warehouse transfers, promotional diversions, and returns, along with the corresponding item quantities.

[0076] Inventory balance constraint: After adjustment, the inventory of each warehouse = original inventory + inbound transfer volume - outbound transfer volume - sales volume - return volume.

[0077] Safety stock constraint: After adjustment, the inventory of each warehouse shall be greater than or equal to the minimum safety stock of that warehouse.

[0078] Capacity constraint: After adjustment, the inventory of each warehouse shall be less than or equal to the maximum capacity of the warehouse.

[0079] Non-negative constraint: All transfer quantities, promotional quantities, and return quantities are ≥0.

[0080] Overstock digestion constraint: After adjustment, the total quantity of stored goods minus the quantity of pre-sold goods should be less than or equal to the maximum value of the quantity range.

[0081] Each adjustment scheme is coded as Fj=[xi, yi, zi], where Fj represents the adjustment scheme, j is the index of the adjustment scheme, xi represents the net amount transferred from warehouse i to other warehouses, with a value range of [0, the excess quantity of warehouse i]; yi represents the promotional consumption amount of warehouse i, with a value range of [0, the promotional quantity of warehouse i]; zi represents the return processing amount of warehouse i, with a value range of [0, the return quantity of warehouse i]; i=1,2,……N; N represents the total number of warehouses.

[0082] Targeted transfers between multiple warehouses are represented by matrix encoding, i.e.:

[0083] x ii This represents the quantity transferred from warehouse i to other warehouse i, where i = 1, 2, ..., N; when two warehouses overlap, the value is 0, i.e., X11.

[0084] Initial population generation: 70% of the population is generated randomly within the feasible region, satisfying the constraints; the remaining 30% is generated using a greedy strategy to generate better initial solutions. Priority is given to allocating resources to warehouses that are nearby and have large demand gaps; priority is also given to promoting items with large inventories. The population size is dynamically adjusted based on the number of warehouses; this application provides an example of 200 warehouses.

[0085] Construct a fitness function to evaluate the merits of each adjustment scheme, specifically as follows:

[0086] YFj=1 / (Mj+λ×P penalty ), where Mj is the total cost required for the adjustment plan, λ is the penalty coefficient, and P penalty To constrain violations and penalties; Where su1i represents the safety stock quantity of items in warehouse i, su2i represents the adjusted stock quantity of items in warehouse i, and su3 is the excess quantity minus the maximum value of the quantity range. YFj is the fitness value of the adjustment plan; the larger the fitness value, the better the adjustment plan.

[0087] Randomly select several adjustment schemes, choose the one with the highest fitness value, and perform a crossover operation. Repeat the random selection step until a certain number of adjustment schemes with the highest fitness value are selected, and then perform the crossover operation. Specifically:

[0088] Sub-adjustment scheme 1 = 0.5 × [(1+β) × Fa + (1-β) × Fb], a, b ∈ j, Fa represents adjustment scheme a, and Fb represents adjustment scheme b;

[0089] Sub-adjustment scheme 2 = 0.5 × [(1-β) × Fa + (1+β) × Fb]; β is a random number, with a value of 0.4;

[0090] Calculate the total cost required for each sub-adjustment plan. If the total cost required is less than or equal to the preset budget threshold, mark the sub-adjustment plan as the optimal adjustment plan and output the optimal adjustment plan. Adjust the corresponding warehouse according to the optimal adjustment plan. If the total cost required is greater than the preset budget threshold, continue to perform cross operations on the sub-adjustment plans until the total cost required is less than or equal to the preset budget threshold, and then stop the cross operations.

[0091] The system intuitively represents the directional allocation relationships among multiple warehouses using matrix encoding and employs a hybrid initial population generation strategy. This approach ensures diversity in the solution space exploration while leveraging heuristic information to accelerate convergence. By constructing a fitness function with multiple constraints and an evaluation model with a penalty term, the system ensures that the final optimal adjustment scheme achieves the best balance between total cost and operational feasibility (satisfying all hard constraints).

[0092] The second adjustment strategy for the corresponding items includes: parsing the pre-sale quantity and the corresponding user's delivery location in the order, marking the delivery location, obtaining the coverage area of ​​each warehouse, counting the quantity of delivery locations within the warehouse's coverage area, identifying the sales quantity of the corresponding item in the warehouse, comparing the sales quantity with the warehouse's inventory quantity, if the inventory quantity is greater than the sales quantity and the difference is greater than a preset quantity, then subtract the safety stock quantity from the warehouse's inventory quantity to obtain the item's transfer quantity, and mark the warehouse as a pending transfer warehouse; if the difference is less than or equal to the preset quantity, no operation is performed; if the inventory quantity is less than the sales quantity and the difference is greater than the preset quantity, then subtract the inventory quantity from the warehouse's sales quantity to obtain the item's receiving quantity, and mark the warehouse as a pending receipt warehouse; based on the pending transfer warehouse and the transfer quantity, generating a logistics order for the item, and sending the corresponding item's transfer quantity to the pending receipt warehouse, so that the inventory quantity of the item in the pending receipt warehouse is equal to the sales quantity in the corresponding coverage area of ​​the pending receipt warehouse. By intelligently matching order receiving locations with warehouse coverage areas, inventory is aligned with actual regional sales. It automatically identifies warehouses with excess inventory (pending allocation) and warehouses with inventory shortages (pending receipt), generating precise logistics transfer orders. This ensures that each warehouse's inventory covers the established sales demand within its service area. This effectively avoids structural inventory imbalances caused by overall ample supply but localized stockouts, significantly improving inventory turnover.

[0093] The third adjustment strategy for the corresponding item includes:

[0094] Implementation of Adjustment Strategy Two is applied to each warehouse. Then, the shortage quantity of corresponding items after allocation is determined. If the shortage quantity is greater than or equal to the shortage threshold one, a severe shortage instruction is generated. The item name, warehouse location, and severe shortage instruction are sent to the smart terminal of the corresponding manufacturer. Upon receiving the severe shortage instruction, the manufacturer immediately generates a corresponding production order and delivers the items corresponding to the shortage quantity to the warehouse. If the shortage quantity is greater than or equal to the shortage threshold two but less than the shortage threshold one, a moderate shortage instruction is generated. The item name, warehouse location, and moderate shortage instruction are sent to the smart terminal of the corresponding manufacturer. Upon receiving the moderate shortage instruction, the manufacturer produces the corresponding shortage quantity of items within a preset time period (one week or half a month) and delivers them to the warehouse within the preset time period. When the shortage quantity is less than the shortage threshold two, a minor shortage instruction is generated and sent to the manufacturer's smart terminal; no replenishment is performed. Based on the severity of the shortage, the system automatically issues differentiated instructions (immediate production / scheduled production / temporary non-processing) to manufacturers, specifying the items, warehouse locations, and urgency levels. This ensures that fluctuations in end-of-supply demand can drive upstream production plans in real time and accurately, avoiding overproduction and achieving optimal cost control through dynamic supply-demand balance.

[0095] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A sensor-based method for dynamic demand forecasting and supply chain adjustment, characterized in that, include: Collect demand behavior data and logistics data through the Internet of Things and sensors; The demand behavior data is analyzed to output the prediction results of dynamic demand forecasting; Based on the prediction results, the logistics data is analyzed to output supply chain adjustment strategies; wherein the supply chain adjustment strategies include adjustment strategy one, adjustment strategy two, and adjustment strategy three.

2. The sensor-based dynamic demand forecasting and supply chain adjustment method according to claim 1, characterized in that, The analysis of the aforementioned demand behavior data specifically includes: The browsing data is parsed to obtain the item details browsing time, positive review viewing time, negative review viewing time, and other comment viewing time, and then standardized. Obtain the detail view values ​​of all purchased items for a user and take the average value to obtain the average purchase view value. Compare and analyze the detail view values ​​with the average purchase view value and output the detail view coefficient. If the details browsing coefficient is less than the minimum coefficient threshold, the system will redirect to the shopping cart data analysis, and the item browsing index will be set to the default value. If the details browsing coefficient is greater than or equal to the minimum coefficient threshold, then obtain the weights corresponding to the positive review viewing value, negative review viewing value, and other comment viewing value, calculate the sum of the details browsing coefficient and the comment viewing coefficient, output the item browsing index, and jump to the add to cart data analysis; By analyzing the data from items added to the shopping cart, an "add to cart index" for each item is output. A demand behavior vector is constructed based on item browsing index, shopping cart addition index, and monthly spending amount. Calculate the dot product of the demand behavior vector and the behavior weight vector, and output the demand value of the item; When the demand value is greater than or equal to the demand threshold, users are marked as demanding users, and the number of demanding users is counted to obtain the predicted demand quantity of the item, which is recorded as the prediction result.

3. The sensor-based dynamic demand forecasting and supply chain adjustment method according to claim 2, characterized in that, The data analysis of adding items to the shopping cart includes: Parse the shopping cart data to obtain the number of days items have been added to the cart, and obtain the number of items added to the cart by the user and the average number of days since the items were purchased. Compare the number of days an item is added to the cart to the average number of days it is purchased: If the number of days an item has been added to the cart is less than the average number of days it has been purchased, then the number of days the item has been added to the cart is subtracted from the average number of days it has been purchased to get the remaining number of days. The remaining number of days is then multiplied by the corresponding remaining coefficient to output the cart addition base. If the number of days an item has been added to the cart is greater than or equal to the average number of days it has been purchased, then the number of days the item has been added to the cart is subtracted from the average number of days it has been purchased to get the number of days the item has been over-purchased. The reciprocal of the number of days the item has been over-purchased is recorded as the cart addition base. The "Add to Cart Index" is calculated by dividing the base number of items added to the cart by the total number of items added to the cart.

4. The sensor-based dynamic demand forecasting and supply chain adjustment method according to claim 1, characterized in that, The analysis of the logistics data based on the prediction results specifically includes: Obtain orders for items, identify the orders to obtain the sales quantity of items, and add the sales quantity to the predicted demand quantity to obtain the pre-sale quantity of items. Compare the pre-sale quantity with the total warehouse quantity: If the total quantity in storage is greater than the pre-sale quantity, the difference between the two is calculated to obtain the over-pre-sale quantity of the item. If the over-pre-sale quantity exceeds the maximum value of quantity range one, the adjustment strategy one corresponding to the item is executed. If the excess quantity falls within quantity range one, then the corresponding adjustment strategy two for the item will be executed; If the total quantity in storage is less than or equal to the pre-sale quantity, the difference between the two is calculated to obtain the lower pre-sale quantity of the item. If the lower pre-sale quantity exceeds the lower pre-sale quantity threshold, the corresponding adjustment strategy three for the item is executed.

5. The sensor-based dynamic demand forecasting and supply chain adjustment method according to claim 4, characterized in that, Implement the adjustment strategy one corresponding to the item, including: Set constraints and construct several adjustment schemes for the item; Each adjustment scheme is coded; Targeted transfers between multiple warehouses are represented by matrix encoding; Initial population generation; Construct a fitness function and calculate the fitness value of the adjustment scheme; Randomly select several adjustment schemes, select the adjustment scheme with the highest fitness value and perform cross operation, repeat the random selection step until a certain number of adjustment schemes with the highest fitness value are selected. Perform cross operations on the selected adjustment schemes and output sub-adjustment schemes; Calculate the total cost required for each sub-adjustment plan. If the total cost required is less than or equal to the preset budget threshold, mark the sub-adjustment plan as the optimal adjustment plan, output the optimal adjustment plan and execute it.

6. The sensor-based dynamic demand forecasting and supply chain adjustment method according to claim 5, characterized in that, The fitness function is specifically: YFj=1 / (Mj+λ×P penalty ); Where Mj is the total cost required for the adjustment plan, λ is the penalty coefficient, and P penalty To constrain violations and penalties; , where su1i represents the safety stock quantity of items in warehouse i, su2i represents the adjusted stock quantity of items in warehouse i, and su3 is the excess quantity minus the maximum value of the quantity range.

7. The sensor-based dynamic demand forecasting and supply chain adjustment method according to claim 4, characterized in that, The second adjustment strategy corresponding to the executed item includes: The system analyzes the pre-sale quantity and the corresponding user's delivery location in the order, marks the delivery location, obtains the coverage area of ​​each warehouse, counts the quantity of delivery locations within the warehouse's coverage area, identifies the sales quantity of the corresponding item in the warehouse, compares the sales quantity with the warehouse's inventory quantity, if the inventory quantity is greater than the sales quantity and the difference is greater than a preset quantity, subtracts the safety stock quantity from the warehouse's inventory quantity to obtain the item transfer quantity, and marks the warehouse as a pending transfer warehouse; if the inventory quantity is less than the sales quantity and the difference is greater than a preset quantity, subtracts the inventory quantity from the warehouse's sales quantity to obtain the item receiving quantity, and marks the warehouse as a pending receipt warehouse; based on the pending transfer warehouse and the transfer quantity, a logistics order for the item is generated, and the corresponding item transfer quantity is sent to the pending receipt warehouse, so that the inventory quantity of the item in the pending receipt warehouse is equal to the sales quantity in the corresponding coverage area of ​​the pending receipt warehouse.

8. The sensor-based dynamic demand forecasting and supply chain adjustment method according to claim 4, characterized in that, The adjustment strategy three corresponding to the executed item specifically includes: Implementation of Adjustment Strategy Two is applied to each warehouse. Then, the shortage quantity of corresponding items after allocation is determined. If the shortage quantity is greater than or equal to the shortage threshold one, a severe shortage instruction is generated. The item name, warehouse location, and severe shortage instruction are sent to the smart terminal of the corresponding manufacturer. Upon receiving the severe shortage instruction, the manufacturer immediately generates a corresponding production order and delivers the items corresponding to the shortage quantity to the warehouse. If the shortage quantity is greater than or equal to the shortage threshold two but less than the shortage threshold one, a moderate shortage instruction is generated. The item name, warehouse location, and moderate shortage instruction are sent to the smart terminal of the corresponding manufacturer. Upon receiving the moderate shortage instruction, the manufacturer produces the corresponding shortage quantity of items within a preset time period (one week or half a month) and delivers them to the warehouse within the preset time period. When the shortage quantity is less than the shortage threshold two, a minor shortage instruction is generated and sent to the manufacturer's smart terminal; no replenishment is performed.

Images