Warehouse logistics dynamic inventory prediction method based on multi-modal AI learning
By constructing a dynamic inventory forecasting model for warehousing and logistics using multimodal AI learning methods, the problem of traditional inventory forecasting systems being unable to integrate multi-source data and make dynamic adjustments is solved, thus achieving efficient inventory management and multi-objective optimization.
Patent Information
- Application Number
- CN202511042222.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-28
- Publication Date
- 2025-11-18
AI Technical Summary
Traditional inventory forecasting methods rely on single historical data, cannot integrate multi-source real-time data, lack dynamic adjustment capabilities, are difficult to cope with sudden changes in demand, and are difficult to balance multiple objectives such as inventory cost, warehouse space utilization, and supply timeliness.
A multimodal AI learning approach is adopted, using LSTM network, XGBoost algorithm and Prophet algorithm to build a dynamic inventory prediction model for warehousing and logistics. It integrates temperature, humidity, vibration, structured and basic data to make short-term, medium-term and long-term predictions, and combines it with a replenishment decision system for dynamic inventory management.
It enables real-time data fusion, improves replenishment accuracy, shortens response time, establishes an end-to-end optimization platform, and enhances the comprehensiveness and accuracy of inventory management.
Smart Images
Figure CN120975706A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application relates to the field of warehouse logistics management, and in particular to a warehouse logistics dynamic inventory prediction method based on multi-modal AI learning. BACKGROUND
[0002] The existing traditional inventory prediction system based on a statistical method adopts a moving average method, an exponential smoothing method and the like, only uses historical sales data as input, and has a fixed prediction period (usually a week / month), and cannot handle sudden demand changes.
[0003] The replenishment system based on ERP integration is directly integrated with the ERP system, adopts a fixed reorder point (ROP) strategy, triggers replenishment based on a preset safety stock level, and lacks multi-objective optimization capability.
[0004] The traditional inventory prediction method relies on single historical sales data and cannot fuse multi-source real-time data such as vision, RFID and environment; the existing replenishment decision system lacks dynamic adjustment capability and is difficult to cope with sudden demand changes and supply chain fluctuations; the manual inventory method is low in efficiency and high in error rate, and cannot realize real-time inventory state monitoring; the discrete data acquisition system leads to difficulty in effective integration and real-time analysis of multi-source heterogeneous data; the static replenishment strategy cannot adapt to dynamic business scenarios such as seasonal fluctuations and promotional activities; and the traditional method is difficult to balance multi-objective optimization of inventory cost, warehouse space utilization rate and delivery timeliness. SUMMARY
[0005] The purpose of the present application is to provide a warehouse logistics dynamic inventory prediction method based on multi-modal AI learning, in order to solve the problem that the traditional method is difficult to balance multi-objective optimization of inventory cost, warehouse space utilization rate and delivery timeliness.
[0006] The above-mentioned purpose of the present application is realized by the following technical scheme: S1: acquiring multi-modal data of warehouse logistics and performing preprocessing; S2: constructing a warehouse logistics short-term prediction model; the warehouse logistics short-term prediction model comprises an LSTM network, an XGBoost algorithm and a Prophet algorithm; the multi-modal data is processed based on the LSTM network to obtain a short-term prediction result; S3: using the XGBoost algorithm to analyze the structured features of the multi-modal data to obtain a medium-term prediction result; S4: using the Prophet algorithm to identify the trend changes of the multi-modal data to obtain a long-term prediction result; S5: determining a warehouse logistics dynamic inventory prediction result within a preset time through the short-term prediction result, the medium-term prediction result and the long-term prediction result, and constructing a warehouse logistics replenishment decision.
[0007] Optionally, the step S1 comprises that the multi-modal data comprises temperature and humidity data, vibration data, structured data and basic data. The structured data comprises warehouse three-dimensional layout and shelf position. The basic data comprises commodity SKU information, supplier information, inventory state data and inventory and order information.
[0008] Optionally, the inventory state data specifically comprises: The shelf commodity features are extracted through a CNN network. The shelf commodities are detected through a YOLOv5 model to obtain commodity movement records, and the commodity state data are obtained in combination with the shelf commodity features. The temperature and humidity data are associated with the commodity state data through a constructed commodity shelf life model to output the inventory state data of the commodities.
[0009] Optionally, the preprocessing comprises: The collection frequency and field integrity of each multi-modal data are verified. The abnormal data in the multi-modal data are determined through cross-validation. The multi-modal data are subjected to outlier processing, missing value filling and data standardization.
[0010] Optionally, the step S2 comprises: The LSTM network comprises an encoding module and a decoding module. The encoding module adopts an attention mechanism and a multi-scale information fusion strategy to balance economic batch optimization demand, supplier selection demand and procurement order generation demand, and a short-term prediction result of inventory pre-allocation.
[0011] Optionally, the replenishment decision comprises replenishment quantity, replenishment time and replenishment path.
[0012] An electronic device comprises a processor, a memory, a user interface and a network interface, the memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to enable the electronic device to execute a warehouse logistics dynamic inventory prediction method based on multi-modal AI learning.
[0013] A computer readable storage medium stores instructions, when the instructions are executed, a warehouse logistics dynamic inventory prediction method based on multi-modal AI learning is executed.
[0014] The technical scheme provided by the present application has the beneficial effects that: Through real-time data fusion, improve the comprehensiveness of data acquisition; build a dynamic prediction model to improve the accuracy of replenishment; develop a real-time decision system to shorten the response time of replenishment; establish a unified platform to realize end-to-end optimization. BRIEF DESCRIPTION OF DRAWINGS
[0015] The present application will be further described below with reference to the accompanying drawings and examples, wherein: Figure 1 is a step diagram in the embodiments of the present application; Figure 2 is a schematic diagram of the structure of an electronic device in the embodiments of the present application. DETAILED DESCRIPTION
[0016] In order to have a clearer understanding of the technical features, objects and effects of the present application, the specific embodiments of the present application will be described in detail with reference to the accompanying drawings.
[0017] The embodiments of the present application provide a warehouse logistics dynamic inventory prediction method based on multi-modal AI learning.
[0018] Please refer to Figure 1 , Figure 1 is a step diagram of a warehouse logistics dynamic inventory prediction method based on multi-modal AI learning in the embodiments of the present application, comprising: S1: obtaining multi-modal data of warehouse logistics and pre-processing; S2: constructing a short-term prediction model of warehouse logistics; the short-term prediction model of warehouse logistics includes: LSTM network, XGBoost algorithm and Prophet algorithm; based on the LSTM network, the multi-modal data is processed to obtain a short-term prediction result; S3: using the XGBoost algorithm to analyze the structured features of the multi-modal data to obtain a medium-term prediction result; S4: using the Prophet algorithm to identify the trend changes of the multi-modal data to obtain a long-term prediction result; S5: determining the warehouse logistics dynamic inventory prediction result within a preset time through the short-term prediction result, the medium-term prediction result and the long-term prediction result, and constructing a replenishment decision of warehouse logistics.
[0019] The application provides an implementation as follows: based on data preprocessing requirements, the following standardization processes are performed: aggregating data according to the [warehouse-commodity] dimension to generate an inventory level view; calculating a dynamic safety stock threshold: safety stock = average daily demand x lead time + Z value x demand standard deviation; marking outliers: triggering an alarm when the inventory amount is outside the (mean ± 3σ) range; verifying demand based on sales forecasts: comparing historical forecast accuracy to calculate an adjustment coefficient: adjustment coefficient = actual sales / forecast sales (past 30-day moving average); applying seasonal adjustment: loading preset seasonal factors according to commodity categories; based on replenishment decision engine requirements, the following processes are automatically executed every 4 hours; merging inventory data, sales forecasts, and supply chain status.
[0020] Based on decision logic execution requirements, calculate net demand: net demand = forecast demand + safety stock - current inventory - in-transit quantity Based on economic order quantity requirements, consider the following factors to calculate the optimal order quantity: Net demand: future 7-day demand according to the output of the prediction model Ordering cost: including procurement labor, transportation, and management expenses Holding cost: considering warehouse space occupancy and capital occupancy cost Calculation formula: order quantity = max(net demand, √(2 x demand x ordering cost / holding cost)) Based on multi-objective optimization requirements, generate replenishment suggestions considering: inventory turnover rate indicators, warehouse space utilization, transportation cost constraints, and supplier delivery cycles.
[0021] Step S1 includes: the multi-modal data includes: temperature and humidity data, vibration data, structured data, and basic data; The structured data includes: warehouse three-dimensional layout and shelf location; The basic data includes: commodity SKU information, supplier information, inventory status data, and inventory and order information.
[0022] The inventory status data specifically includes: Extracting shelf commodity features through a CNN network; Detecting shelf commodities through a YOLOv5 model to obtain commodity movement records, and combining the shelf commodity features to obtain commodity status data; Associating temperature and humidity data with commodity status data through a constructed commodity shelf life model to output commodity inventory status data.
[0023] The preprocessing includes: Verifying the collection frequency and field integrity of each multi-modal data; Determining abnormal data in the multi-modal data through cross-validation; The multi-modal data is subjected to outlier processing, missing value filling and data standardization.
[0024] The step S2 comprises: The LSTM network comprises an encoding module and a decoding module. The encoding module adopts an attention mechanism and a multi-scale information fusion strategy to balance the economic batch optimization demand, the supplier selection demand and the procurement order generation demand, and the short-term prediction result of the inventory pre-allocation.
[0025] The replenishment decision comprises a replenishment quantity, a replenishment time and a replenishment path.
[0026] The application also discloses an electronic device. Referring to Figure 2 , Figure 2 is a structural schematic diagram of an electronic device disclosed by the embodiment of the application. The electronic device 500 can comprise at least one processor 501, at least one network interface 504, a user interface 503, a memory 505 and at least one communication bus 502.
[0027] The communication bus 502 is used to realize the connection communication between the components.
[0028] The user interface 503 can comprise a display screen, and the optional user interface 503 can further comprise a standard wired interface and a wireless interface.
[0029] The network interface 504 can optionally comprise a standard wired interface and a wireless interface (such as a WI-FI interface).
[0030] The application also discloses a computer readable storage medium, which stores a plurality of instructions, and the instructions are suitable for being loaded by a processor to execute the above-mentioned warehouse logistics dynamic inventory prediction method based on multi-modal AI learning.
[0031] The above are only exemplary embodiments of the present disclosure, and cannot limit the scope of the present disclosure. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure.
[0032] The present application is intended to cover any variations, uses, or adaptive changes of the present disclosure that follow the general principles of the present disclosure and include common knowledge or conventional technical means in the technical field not recorded in the present disclosure. The scope and spirit of the present disclosure are defined by the claims. The specification and examples are only regarded as exemplary, and the scope and spirit of the present disclosure are defined by the claims.
Claims
1. A method for dynamic inventory forecasting in warehousing and logistics based on multimodal AI learning, characterized in that, The method includes the following steps: S1: Acquire and preprocess multimodal data of warehousing and logistics; S2: Construct a short-term forecasting model for warehousing and logistics; the short-term forecasting model for warehousing and logistics includes: LSTM network, XGBoost algorithm and Prophet algorithm; based on the LSTM network, multimodal data is processed to obtain short-term forecasting results; S3: Use the XGBoost algorithm to analyze the structured features of multimodal data and obtain mid-term prediction results; S4: Use the Prophet algorithm to identify trend changes in multimodal data and obtain long-term prediction results; S5: Based on short-term, medium-term, and long-term forecasts, determine the dynamic inventory forecast for warehousing and logistics within a preset time period, and construct replenishment decisions for warehousing and logistics.
2. The method for dynamic inventory forecasting in warehousing and logistics based on multimodal AI learning as described in claim 1, characterized in that, Step S1 includes: the multimodal data includes: temperature and humidity data, vibration data, structured data, and basic data; Structured data includes: warehouse 3D layout and shelf locations; Basic data includes: product SKU information, supplier information, inventory status data, and inventory and order information.
3. The method for dynamic inventory forecasting in warehousing and logistics based on multimodal AI learning as described in claim 2, characterized in that, The inventory status data specifically includes: Extracting features of goods on the shelf using a CNN network; By detecting goods on the shelf using the YOLOv5 model, we obtain records of goods movement and combine them with the characteristics of the goods on the shelf to obtain goods status data. By constructing a product shelf-life model, temperature and humidity data are correlated with product status data to output product inventory status data.
4. The method for dynamic inventory forecasting in warehousing and logistics based on multimodal AI learning as described in claim 2, characterized in that, The preprocessing includes: Verify the acquisition frequency and field integrity of each multimodal data; Cross-validation was used to identify outliers in the multimodal data. Perform outlier handling, missing value imputation, and data standardization on multimodal data.
5. The method for dynamic inventory forecasting in warehousing and logistics based on multimodal AI learning as described in claim 1, characterized in that, Step S2 includes: The LSTM network includes: an encoding module and a decoding module; The encoding module employs an attention mechanism and a multi-scale information fusion strategy to balance the needs of economic batch optimization, supplier selection, and purchase order generation with short-term forecasts of inventory pre-allocation.
6. The method for dynamic inventory forecasting in warehousing and logistics based on multimodal AI learning as described in claim 1, characterized in that, The replenishment decision includes: replenishment quantity, replenishment time, and replenishment path.
7. An electronic device, characterized in that, The device includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any one of claims 1-6.
8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed by a computer, perform the method as described in any one of claims 1-6.