Material inventory management method and device, medium and product

Abstract

The invention discloses a material inventory management method and device, a medium and a product. The method comprises the following steps: carrying out material data acquisition based on a preset acquisition strategy, and determining acquisition time, belonging organization, material category and management material data corresponding to management materials, so as to generate candidate material data and store the candidate material data; in response to the inventory quota prediction request, determining an annual parameter, an organization parameter and a material category, screening the candidate material data according to the annual parameter, the organization parameter and the material category, and determining a target material meeting a screening condition; and determining a predicted inventory value of each target material according to the target material data corresponding to the target material based on a pre-trained hybrid prediction model, and performing material inventory management according to the predicted inventory value of each target material. According to the technical scheme, the scientific and accurate inventory prediction value can be determined by using the comprehensive material data and the prediction model, and the efficiency of material inventory management is improved.

Description

Technical Field

[0001] This invention relates to the field of big data, and in particular to a method, apparatus, medium and product for managing material inventory. Background Technology

[0002] Inventory quotas, as a core control indicator for enterprise material management, directly determine the reliability of material supply and the rationality of inventory costs. Traditional methods of setting inventory quotas generally suffer from problems such as reliance on manual experience, limited data dimensions, and rigid forecasting models.

[0003] As digital transformation progresses, enterprises have accumulated a large amount of historical data on procurement, consumption, and inventory. How to use this data to build intelligent prediction models and output scientific inventory quota ranges (rather than a single value) to improve the efficiency of material inventory management is an urgent problem to be solved. Summary of the Invention

[0004] This invention provides a method, apparatus, medium, and product for material inventory management, which can determine scientific and accurate inventory forecasts and improve the efficiency of material inventory management.

[0005] According to one aspect of the present invention, a material inventory management method is provided, comprising: Based on a preset data collection strategy, material data is collected to determine the collection time, organization, material category, and managed material data corresponding to the managed materials, so as to generate and store candidate material data. In response to the inventory quota forecasting request, determine the annual parameters, organizational parameters, and material categories, and filter the candidate material data based on the annual parameters, organizational parameters, and material categories to determine the target materials that meet the filtering criteria; Based on a pre-trained hybrid prediction model, the predicted inventory value of each target material is determined according to the target material data corresponding to the target material, and material inventory management is carried out based on the predicted inventory value of each target material.

[0006] According to another aspect of the present invention, a material inventory management device is provided, comprising: The generation module is used to collect material data based on a preset collection strategy, determine the collection time, organization, material category, and managed material data corresponding to the managed materials, so as to generate candidate material data and store it. The filtering module is used to respond to inventory quota forecasting requests, determine annual parameters, organizational parameters, and material categories, and filter candidate material data based on the annual parameters, organizational parameters, and material categories to determine target materials that meet the filtering criteria. The management module is used to determine the predicted inventory value of each target material based on the pre-trained hybrid prediction model and the target material data corresponding to the target material, and to manage the material inventory based on the predicted inventory value of each target material.

[0007] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the material inventory management method according to any embodiment of the present invention.

[0008] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the material inventory management method according to any embodiment of the present invention.

[0009] According to another aspect of the present invention, a computer program product is also provided, which includes a computer program that, when executed by a processor, implements the material inventory management method of any embodiment of the present invention.

[0010] The technical solution of this invention collects material data based on a preset collection strategy, determines the collection time, organization, material category, and managed material data corresponding to the managed materials, generates candidate material data, and stores it. In response to an inventory quota forecast request, it determines annual parameters, organizational parameters, and material categories, and filters the candidate material data according to these parameters to determine target materials that meet the filtering criteria. Based on a pre-trained hybrid forecasting model, it determines the predicted inventory value of each target material according to the target material data corresponding to the target materials, and manages material inventory based on the predicted inventory values ​​of each target material. By utilizing comprehensive material data and a forecasting model, scientifically accurate inventory forecasts can be determined, improving the efficiency of material inventory management.

[0011] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart of a material inventory management method provided in an embodiment of the present invention; Figure 2 This is a structural block diagram of a material inventory management device provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of the electronic device provided in an embodiment of the present invention. Detailed Implementation

[0014] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0015] It should be noted that the terms "first," "second," "target," "candidate," and "alternative," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the invention described herein can be practiced in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus. The acquisition, storage, use, and processing of data in the technical solutions of this application comply with relevant laws and regulations.

[0016] It should be noted that traditional inventory quota setting methods generally suffer from the following problems: First, data coverage is incomplete, focusing only on historical inventory quantities and simple consumption records, without integrating supply chain collaborative data (such as supplier capacity and logistics delivery timeliness), external environmental data (such as raw material price indices and industry demand), and internal dynamic data (such as production process adjustments and equipment maintenance plans), resulting in a lack of a holistic perspective in quota setting. Second, forecasting methods are simplistic, often employing static statistical models or single algorithms, failing to capture the nonlinear correlations and temporal evolution characteristics between data. Prediction results are mostly fixed values, making it difficult to cope with dynamic scenarios such as market fluctuations and sudden demand, easily leading to inventory backlog or stockout risks. Third, dynamic adaptability is weak; once quotas are determined, they are used for a long time, lacking a real-time adjustment mechanism to match business changes (such as capacity expansion, product iteration, and supply chain restructuring). When significant changes occur in the internal and external environment, the deviation between quotas and actual demand will continue to widen, seriously affecting inventory management efficiency. To address the aforementioned issues, this invention provides a dynamic range prediction scheme for enterprise inventory quotas based on multi-source time-series data fusion. Through multi-dimensional data integration, in-depth mining of time-series features, construction of hybrid intelligent models, and closed-loop feedback optimization, it outputs a dynamic inventory quota range that includes "safety lower limit - optimal value - cost upper limit". This solves the technical pain points of traditional methods, such as insufficient data dimensions, limited prediction accuracy, and poor dynamic adaptability. The specific implementation method will be described in detail in subsequent embodiments.

[0017] Example 1 Figure 1 This is a flowchart of a material inventory management method provided by an embodiment of the present invention. This embodiment is applicable to situations where inventory forecasting is performed on different materials based on material data and a hybrid forecasting model to achieve better inventory management. This method can be executed by a material inventory management device, which can be implemented in hardware and / or software. The material inventory management device can be configured in an electronic device, such as an electronic device equipped with a material inventory management system. Figure 1 As shown, this material inventory management method includes: S101. Collect material data based on the preset collection strategy, determine the collection time, organization, material category and management material data corresponding to the managed material, so as to generate candidate material data and store it.

[0018] The preset data acquisition strategy can be a dual acquisition strategy based on real-time and offline data acquisition. Managed materials refer to all materials included in the materials inventory management system. Managed material data refers to data representing the procurement, consumption, and inventory status of managed materials. The affiliated organization refers to the organization to which the managed materials belong. The affiliated organization can include organizational type and organizational level. Specific organizational types can be manufacturing, R&D, sales, or service sectors, and organizational levels can be group headquarters, subsidiaries, branches, or business units. Material categories can be preset at different levels, such as primary, secondary, and tertiary material categories. These can be pre-classified based on the importance of the materials or their usage frequency; this invention does not impose any limitations on this.

[0019] Optionally, material data is collected based on a preset collection strategy to determine the collection time, organization, material category, and managed material data corresponding to the managed materials, so as to generate and store candidate material data. This includes: collecting data based on a real-time and offline dual collection strategy to determine the collection time, organization, and material category corresponding to the managed materials, and determining the managed material data corresponding to the managed materials; performing anomaly detection, missing data processing, standardization, and fusion processing on the managed material data to obtain candidate material data, and storing the candidate material data based on a strategy of database sharding, table partitioning, and partitioned storage.

[0020] For example, based on a real-time and offline dual-collection strategy, a preset real-time computing framework combined with a timed scheduling mechanism can be used to periodically (e.g., at 2 AM daily) initiate full data synchronization (covering all business data from the previous day to avoid missing real-time collection), with a synchronization duration of 1.5 hours. For incremental data supplementation, a timed trigger mode can be used, with a synchronization cycle of 1 hour / time and a collection window duration of 5 minutes each time, capturing only newly added and changed data within the cycle, significantly reducing data transmission pressure and ensuring data timeliness (end-to-end latency ≤ 10 minutes), thus achieving real-time collection. At the same time, integrated tools can be used for offline batch data extraction. For historical archived data and non-real-time updated external static data (e.g., industry benchmark data, supplier qualification files), a full offline synchronization can be performed periodically (e.g., every Sunday at midnight), forming a "real-time + offline" dual-collection mode that balances timeliness and data integrity. To avoid data interruption during the acquisition process, a fault tolerance mechanism is introduced: the acquisition status is saved every 5 minutes during real-time acquisition, and acquisition can be resumed after abnormal interruption; offline acquisition adopts the breakpoint resume function, and a segmented transmission strategy is adopted for very large files (such as single files > 10GB) to ensure that data acquisition is not lost or duplicated.

[0021] Optionally, managed material data may include at least one of the following: procurement data, consumption data, inventory data, production data, external data, and extended data for special materials. Candidate material data refers to standardized material data obtained after preprocessing the managed material data.

[0022] Procurement data may include at least one of the following: order number, material code, purchase quantity, unit price, supplier information (code, name, qualification level), order date, estimated delivery date, actual delivery date, mode of transport (road / rail / air), transport carrier, payment method, quantity accepted and reasons for non-acceptance. Consumption data may include at least one of the following: material requisition organization, requisition department, material code, quantity consumed, consumption amount, purpose of consumption (production / repair / project / spare), corresponding work order number, consumption time (accurate to the minute), equipment code, operator, and approver. Inventory data may include at least one of the following: warehouse code, storage location code, material code, current inventory quantity, inventory amount (calculated using the moving weighted average method), inventory status (in use / stagnant / scrapped / pending inspection), batch number received, production date, shelf life, remaining shelf life, storage environment parameters (temperature / humidity, for sensitive materials), inventory turnover rate (monthly / quarterly / annual), last inventory count date, and discrepancies. Production data may include at least one of the following: planned production output, actual output, equipment runtime, work order progress, capacity utilization rate, production process adjustment records, product list change records, and process time.

[0023] For example, external data refers to multi-dimensional external data collected through standardized application programming interfaces (APIs). These APIs can use encrypted transmission and have call frequency limits set (≤10 times / minute) to avoid triggering third-party platform rate limiting. Specifically, this includes: industry raw material price indices (such as industry association-specific price indices), supplier capacity utilization rates (updated monthly and directly synchronized from supplier systems), logistics and transportation timeliness indices (obtained through logistics company APIs and statistically analyzed by transportation mode), market demand prosperity indices (such as industry demand forecast report data), and macroeconomic indicators, enriching the input dimensions of the hybrid forecasting model and improving its adaptability to changes in the external environment.

[0024] Specialized materials extended data refers to additional data collected specifically for specialized materials (such as high-precision components, customized consumables, shelf-life sensitive materials, and hazardous chemicals). For example, for high-precision components, extended data can include supplier quality pass rates (e.g., the number of batches that have passed acceptance in the past 3 years / total batches), equipment matching accuracy requirements (tolerance range), availability of alternative materials (determined by the quantity and procurement cycle of alternative materials), and inspection and testing cycles. For shelf-life sensitive materials (such as chemical raw materials and fresh consumables), extended data can include production date, shelf life, current remaining shelf life, storage environment parameters (real-time temperature / humidity / light protection), and deterioration warning thresholds. For hazardous chemicals, extended data can include storage safety levels, transportation restrictions, disposal costs, and emergency reserve requirements, ensuring the relevance and rationality of inventory quota forecasts for specialized materials.

[0025] Optionally, real-time data interaction can be performed using interfaces within a service architecture, with automatic retrying after timeout (up to 3 times, with retry intervals of 2 seconds, 5 seconds, and 10 seconds respectively). Critical business data (such as purchase order confirmation data and inventory change data) is transmitted with high reliability through message queues; a partitioning strategy (partitioning by material code hash) is adopted to improve data consumption efficiency, while setting a message expiration time (7 days) to avoid storage redundancy.

[0026] Optionally, external data can be collected through a standardized application programming interface (API), employing data encryption transmission and verification mechanisms. After each batch of data transmission is completed, the data verification codes of the local machine and the third-party platform are compared. If they are inconsistent, re-collection is triggered to ensure data security and integrity.

[0027] For example, data can be collected from the preset enterprise material management system, procurement management system, production execution system, financial system, and equipment management system to obtain the managed material data.

[0028] Optionally, based on the candidate material data and using a strategy of sharding and partitioning storage, at least one of the following tables can be generated: a material basic information table, a standardized procurement data table, a standardized consumption data table, a standardized inventory data table, and an external related data table, thereby realizing sharding and partitioning storage of candidate material data.

[0029] For example, the basic information table of materials can be used to store static information such as material code, material name, specifications, category (e.g., level 1 / 2 / 3), importance level (e.g., core / important / general / auxiliary), alternative material code, supplier list (primary supplier / alternate supplier), unit of measurement, storage requirements, inspection standards, and whether it is a special material. The primary key is the material code, and it is stored in a single table (with a small data volume, about 100,000 records), and supports regular updates (e.g., once a month).

[0030] For example, a standardized procurement data table can include fields such as order number, material code, purchase quantity, unit price (including tax), amount before tax, amount including tax, tax rate, supplier code, supplier name, order date, estimated arrival date, actual arrival date, transportation method, carrier, acceptance result, quantity of non-conforming items, payment status, and cleaning status. The primary key is order number + material code. It is stored in tables by year and partitioned by quarter, and supports quick queries by order date, material code, and supplier code.

[0031] For example, a standardized consumption data table can cover fields such as material requisition organization, material requisition department, material code, material name, consumption quantity, consumption amount, consumption date, work order number, consumption purpose (production / repair / project / spare), equipment code, operator, requisition approver, and cleaning status. The primary key is the material requisition number + material code. It is stored in separate tables by year and partitioned by month to adapt to high-frequency consumption data query needs.

[0032] For example, a standardized inventory data table can include fields such as warehouse code, storage location code, material code, inventory quantity, inventory value (calculated using moving weighted average price), inventory status (in use / stagnant / scrapped / pending inspection), batch number, production date, shelf life, remaining shelf life, storage environment parameters (temperature / humidity), inventory turnover rate (monthly / quarterly / annual), last inventory count time, inventory count discrepancy, and cleaning status. The primary key is material code + warehouse code + batch number. The table is divided by warehouse code and partitioned by month, supporting real-time updates and dynamic inventory queries.

[0033] For example, the external related data table can be used to store fields such as material code, timestamp, raw material price index, supplier capacity utilization rate, logistics timeliness index, market prosperity index, raw material cost volatility, and adjustment flag. The primary key is material code + timestamp, and the table is divided by year and partitioned by date. It supports related queries by time range and material code.

[0034] Optionally, anomaly detection processing is performed on the managed material data to determine whether there are any anomalies corresponding to the anomaly types in the managed material data. If so, the managed material data with anomalies is removed. The anomaly types are quantity anomalies, price anomalies, time anomalies, or status anomalies.

[0035] For example, a combination of "statistical thresholding + isolated forest algorithm" can be used to identify abnormal data, taking into account both statistical regularity and data distribution characteristics, thereby improving the accuracy of anomaly detection. The statistical thresholding method is used to identify normal data anomalies that conform to a normal distribution, while the isolated forest algorithm is used to identify non-linear and non-normally distributed data anomalies. The results of the two methods are combined to ensure that no abnormal data is missed.

[0036] For example, for quantity anomalies, a box plot can be used to determine the reasonable range of managed material data, such as Q1-1.5IQR to Q3+1.5IQR, where Q1 is the first quartile of the managed material data, Q3 is the third quartile, and IQR is the interquartile range of the managed material data. Purchase / consumption data exceeding this range is considered abnormal. For purchase data, additional verification can be performed using dimensions such as historical purchase batches, production plan requirements, and supplier maximum capacity to avoid misjudging reasonable large-amount purchases (such as annual centralized purchases) as abnormal. Records with negative inventory quantities or discrepancies between inventory value and quantity (such as value being 0 but quantity being non-zero, value being negative, or the product of quantity and value deviating by more than ±5%) are directly marked as abnormal. For batch-managed materials, if the total inventory quantity of the same batch deviates from the quantity received by more than ±1%, the batch quantity is considered abnormal.

[0037] For example, in response to price anomalies, the historical price data of the managed materials involved in the managed materials data can be used, combined with preset price fluctuation thresholds (such as ±30% for regular materials, ±50% for scarce materials, and ±40% for seasonal materials), to determine the abnormal prices in the managed materials data that exceed the fluctuation range, and then remove the abnormal data corresponding to the abnormal prices from the managed materials data.

[0038] For example, in response to time anomalies, records with at least one of the following can be identified based on managed material data: the actual arrival date of the purchase order is earlier than the order date; the consumption time is later than the current system time; the inventory receipt time is earlier than the material production time; there is a logical conflict between the inventory receipt time and the production time (such as the receipt time being later than the corresponding work order consumption time); and the shelf life has expired but the inventory status is still "in use". For materials purchased across time zones and regions, the time is uniformly converted to the company headquarters time zone to avoid misjudgments of time anomalies caused by time zone differences.

[0039] For example, for abnormal status, at least one of the following (1)-(4) can be judged as abnormal status: (1) The order status is “cancelled” or “returned” but there are still inbound / consumption records; (2) The inventory status is “scrapped” or “outbound” but the inventory quantity is not 0; (3) The supplier status is “disabled” but there are still new purchase orders; (4) The material status is “obsolete” but there are still consumption records.

[0040] Optionally, the correlation of status changes can be verified. For example, when the inventory status changes from in use to stagnant, it is necessary to determine whether there is a corresponding stagnant material identification report; otherwise, it is marked as an abnormal status.

[0041] Optionally, missing data handling for managed materials data includes: determining whether there are any missing data in the managed materials data; if so, determining the missing data type; the missing data type is missing inventory data, missing procurement data, missing consumption data, missing external data, or missing key fields; if the missing data type is missing key fields, then the managed materials data is removed; if the missing data type is other missing data types besides missing key fields, then the managed materials data is repaired.

[0042] Optionally, a scenario-based missing data completion strategy can be adopted, selecting an appropriate completion method based on the data type, the missing data type, and the business scenario to ensure the rationality and accuracy of the completed data and avoid data deviation caused by a single completion method.

[0043] For example, when dealing with missing inventory data, since the inventory quantity is a discrete state variable, if the missing data is an individual value in a continuous time series, the forward imputation method or the last observation carry-over method can be used to fill it (that is, assuming that the inventory has not changed, the inventory quantity at the most recent known time point is directly used); if the missing data is the value of multiple consecutive time points, the extrapolation method based on consumption deduction is used, that is, using the previous known inventory point, combined with the production consumption record and the warehousing record in that period, to perform rolling extrapolation to fill it.

[0044] For example, to address missing procurement / consumption data, a similarity-based material completion method can be used. Based on characteristics such as material category, specifications, purpose, consuming department, and procurement cycle, the similarity to the missing material data is calculated. The most similar pre-defined number of material data (e.g., 5 categories) are selected and a weighted average is applied to complete the data according to similarity weights (higher similarity results in higher weights, with a total weight of 1). If there is no similar material data, a business rule completion method is used, which completes the data based on historical average procurement / consumption volume and production plan requirements.

[0045] For example, to address missing external data, a strategy of rolling average + trend correction can be used to fill in the missing data. For instance, the rolling average of a preset period (e.g., 7 days) before the missing data is taken as the base value, and the base value is corrected by combining the industry trend index and the change rate of relevant indicators during the same period to obtain the final filled value. If the missing data exceeds 3 consecutive days, it can be filled in by manually querying data from third-party platforms to ensure data integrity.

[0046] For example, missing key fields may refer to the absence of important field information such as material code, supplier code, and consumption time. If it cannot be completed by the algorithm, it can be marked as invalid data, removed and the reason for the absence recorded. At the same time, it is pushed to the business department for supplementation and improvement, and then reintroduced into the data processing flow.

[0047] For example, standardization can be carried out according to the principles of unified dimensions, unified format, and unified naming to eliminate differences in dimensions and format conflicts. For instance, the date format can be unified as "YYYY-MM-DD HH:MM:SS", the monetary unit can be unified as "yuan", and numerical data can be uniformly retained to two decimal places.

[0048] Optionally, a fusion strategy of "primary key association + hierarchical mapping" can be adopted. For internal business data, such as procurement data, consumption data, inventory data, and production data, the "material code + timestamp" is used as the primary key for association. For external data (such as industry raw material price indices), since they do not have single-material granularity, a "material category - industry index mapping table" is constructed. During fusion, the third-level category to which the material code belongs can be parsed first, and then the corresponding external data can be associated through "material category + timestamp". For example, all "steel" material codes are associated with the "steel industry index" at the same point in time, thereby sinking macro features down to the micro-material dimension and constructing a standardized wide data table with 28 fields. The fields cover five major categories: basic material information, procurement data, consumption data, inventory data, and external data. Redundant fields such as cleaning status (cleaned / to be cleaned / abnormal), data source, and cleaning time are also added to facilitate data traceability and problem investigation.

[0049] S102. In response to the inventory quota forecasting request, determine the annual parameters, organizational parameters, and material categories, and filter the candidate material data according to the annual parameters, organizational parameters, and material categories to determine the target materials that meet the screening conditions.

[0050] The "Inventory Quota Forecast Request" is a request to forecast the inventory status of a specified target material. The "Annual Parameter" refers to the year of the target forecast, such as the past year or the past two years. The "Organizational Parameter" refers to the organizational type and / or organizational level of the target forecast. The "Material Category" refers to the category to which the target forecast material belongs. The "Target Material" refers to the material in the candidate material data that meets the screening criteria. The screening criteria can be a match between the annual parameter, organizational parameter, and material category.

[0051] Optionally, if the annual parameter is the past n years, where n is a natural number greater than or equal to 1, the corresponding data can be filtered by organizational parameters + material category, and standardized data from the past n years can be selected as the target material. At the same time, time periods with insufficient data integrity (such as a data missing rate of >10% in a certain month) can be removed to ensure the quality of the data source. For special material data with a small amount of data (such as annual consumption times <12 times), the data filtering range can be expanded, such as to the past n+2 years, and similar material data of the same category can be added as an auxiliary to avoid feature construction bias.

[0052] Optionally, after determining the annual parameters, organizational parameters, and material categories, the process also includes: verifying the annual parameters, organizational parameters, and material categories; if the parameter verification is successful, then filtering the candidate material data based on the annual parameters, organizational parameters, and material categories to determine the target materials that meet the filtering criteria.

[0053] For example, the validity of the annual parameter can be verified using regular expressions (it must be a 4-digit valid number and within the historical data coverage range, such as the past 8 years); the validity of the organization type can be verified using the organization information table (supporting division by business segment: manufacturing segment, sales segment, R&D segment, service segment; and division by organizational level: group headquarters, subsidiary, branch, business unit); and the validity of the material category parameter can be verified using the material category table (if it needs to be a three-level category code). If the parameter is determined to be invalid, the system will automatically prompt an error message and block the process execution.

[0054] S103. Based on the pre-trained hybrid prediction model, determine the predicted inventory value of each target material according to the target material data corresponding to the target material, and manage the material inventory according to the predicted inventory value of each target material.

[0055] Among them, the pre-trained hybrid prediction model can be, for example, a hybrid prediction model that combines a bidirectional long short-term memory network, an attention mechanism, and a lightweight gradient boosting tree. By capturing the bidirectional dependencies of time-series data through the bidirectional long short-term memory network, strengthening the weights of key time-series features through the attention mechanism, and fitting the nonlinear relationship between features and labels through the lightweight gradient boosting tree, the three can work together to achieve high-precision inventory prediction.

[0056] Optionally, based on a pre-trained hybrid prediction model, the predicted inventory value of each target material is determined according to the target material data corresponding to the target material, including: performing feature extraction and feature optimization processing on the target material data corresponding to the target material to determine the target material features; and determining the predicted inventory value of each target material based on the target material features according to the pre-trained hybrid prediction model.

[0057] The target material characteristics include at least one of the following: time-series trend characteristics, business-related characteristics, risk control characteristics, and cost optimization characteristics. Time-series trend characteristics may include at least one of the following: annual consumption growth rate, quarterly consumption volatility coefficient, monthly consumption series autocorrelation, and inventory change rate.

[0058] Optionally, the annual consumption growth rate can be calculated as (total consumption in the current year - total consumption in the previous year) / total consumption in the previous year × 100%. If the total consumption in the previous year is 0, it is calculated as 10% of the total consumption in the current year (to avoid a denominator of 0), reflecting the annual trend of material consumption. The quarterly consumption fluctuation coefficient can be the standard deviation / mean of the quarterly consumption data. The larger the coefficient, the more drastic the fluctuation, reflecting the quarterly stability of material consumption. The autocorrelation coefficient with a lag of 1-12 periods can be calculated using a preset time period, such as the consumption data of the past 36 months. The maximum value is taken as the characteristic group of the autocorrelation of the monthly consumption sequence, reflecting the temporal correlation of the monthly consumption data. The inventory change rate can be the average of the inventory difference between two adjacent months. A positive value indicates an increase in inventory, and a negative value indicates a decrease in inventory, reflecting the dynamic trend of inventory change.

[0059] Business-related characteristics can include at least one of the following: production plan matching degree, work order-related consumption ratio, supplier cooperation period, and material substitution difficulty coefficient. Specifically, the production plan matching degree can be the ratio of total material consumption to the corresponding planned production output. The closer the ratio is to 1, the higher the matching degree, reflecting the synergy between material consumption and production plan. The work order-related consumption ratio can be the consumption quantity directly related to the production work order / total consumption quantity × 100%, reflecting the production orientation of material consumption. The supplier cooperation period can be the difference between the current time and the supplier's first cooperation time (calculated in years), reflecting the stability of supplier cooperation. The material substitution difficulty coefficient can be preset based on the availability of substitute materials (whether there are substitute materials), substitution cost (cost difference rate between substitute materials and original materials), and substitution cycle (substitute material procurement cycle), ranging from 0 to 1. The larger the coefficient, the higher the substitution difficulty (the coefficient is 1 when there are no substitute materials, and the coefficient is below 0.3 when the substitution cost is low and the cycle is short).

[0060] Risk control characteristics include at least one of the following: supplier delivery delay rate, logistics and transportation fluctuation coefficient, and impact of material shortages. Specifically, the supplier delivery delay rate can be a preset duration, such as the number of delayed orders in the past 12 months / total number of orders × 100%. Delayed delivery is defined as the actual delivery date being later than the expected delivery date by more than a preset number of days, such as 3 days, reflecting the timeliness of supplier delivery. The logistics and transportation fluctuation coefficient can be a preset duration, such as the standard deviation / mean of transportation timeliness in the past 3 years. Transportation timeliness is calculated separately for different transportation modes, reflecting the stability of logistics and transportation. The impact of material shortages can be a preset duration, such as the production stagnation time caused by material shortages in the past 3 years / total production time × 100%, reflecting the degree of impact of material shortages on production.

[0061] Cost optimization features may include inventory holding cost volatility and / or purchase quantity discount coefficient. Specifically, inventory holding cost volatility can be a preset time period, such as the standard deviation / mean of inventory holding costs (including storage fees, capital occupation fees, and loss fees) over the past 3 years, reflecting the stability of inventory holding costs. Purchase quantity discount coefficient can be a tiered discount rate conversion value based on purchase quantity, with a value of 0.9 when the batch discount rate is ≥30%, 0.7 when it is 20%-30%, 0.5 when it is 10%-20%, and 0.3 when it is <10%, reflecting the cost advantage of purchase quantity.

[0062] Optionally, feature optimization processing may include at least one of feature standardization, feature dimensionality reduction, and label extraction. Specifically, the Min-Max standardization method is used to map all feature values ​​to the [0,1] interval, i.e., feature standardization is performed. For features with outliers, outliers can be removed first before calculating the minimum and maximum values ​​to avoid the influence of outliers on the standardization results.

[0063] Optionally, a cascaded strategy combining filtering and transformation methods can be used for feature dimensionality reduction. First, the nonlinear correlation between the original features and the label (historical best inventory value) is calculated based on mutual information, eliminating invalid features with mutual information values ​​below 0.05 (noise filtering). Second, principal component analysis is performed on the filtered relevant features, retaining principal components with a cumulative variance contribution rate ≥90% to construct the feature matrix. This strategy can both eliminate data noise irrelevant to the target through mutual information and eliminate multicollinearity among features, resulting in a dimensionally simplified and highly information-fidelity-preserving input matrix.

[0064] Optionally, the hybrid pre-model can be iteratively trained based on historical material data to obtain a trained hybrid prediction model. Each historical material data is the input to the model, and the corresponding output is the optimal inventory value corresponding to the historical material data. For example, within the historical time window, all time points where the inventory satisfaction rate reaches the target value (e.g., ≥98%) can be selected, and the weighted sum of the normalized holding cost and normalized inventory at these time points can be calculated. The actual inventory quantity corresponding to the minimum weighted sum is selected as the optimal theoretical inventory value for that period, that is, the optimal inventory value corresponding to the historical material data, which represents the ideal inventory quantity that neither causes shortages nor wastes funds under the same external conditions in history.

[0065] For example, for historical material data, a chronological partitioning method can be used to divide the training set, validation set, and test set in a ratio of 6:2:2. This strictly follows the chronological order of the time-series data, avoiding the leakage of time-series information caused by random partitioning (i.e., using future data to train predictions of past data, leading to a decrease in the model's generalization ability). For instance, the training set can use feature matrix X and labels y (arranged in chronological order) generated from historical material data of the past four years for model parameter fitting; the validation set can use historical material data from the fifth year for hyperparameter tuning and early stopping control to avoid overfitting; the test set can use data from the sixth year for evaluating the model's generalization ability. The test set data does not participate in any model training or tuning process, ensuring the objectivity and accuracy of the evaluation results. Simultaneously, the dataset is reconstructed in time series, and the feature matrix X is reconstructed into three-dimensional input data (number of samples × time step × number of features) according to the time step (12 months) to adapt to the input requirements of the bidirectional long short-term memory network model. For scenarios with a small number of samples (such as the number of special materials samples <1000), time series data augmentation technology (time axis shifting, noise addition) is used to expand the sample size and improve the model training effect, thereby training the hybrid prediction model.

[0066] For example, the hyperparameters of the hybrid prediction model can be determined by using empirical initialization combined with Bayesian optimization to ensure optimal model performance.

[0067] Optionally, for the bidirectional long short-term memory network module, the core is used to capture the forward and backward dependencies of temporal data. The number of hidden layers is set to 2, the number of neurons in each layer is 32 (the dimension after bidirectional output superposition is 64), and the input dimension is 8 (consistent with the number of features in the feature matrix X). The activation function is the hyperbolic tangent function, and the recursive activation function is the sigmoid function. The temporary deactivation rate is set to 0.2 (input layer) and 0.25 (hidden layer). The time step is set to 12, and the output strategy of the last time step is used for output, with an output dimension of 64.

[0068] Optionally, for the attention mechanism, it is embedded in a bidirectional long short-term memory network to calculate the attention weights of the features at each time step. An additive attention mechanism is adopted, with the input being the hidden layer output (dimension 12×64) of the bidirectional long short-term memory network. The input is mapped to a 32-dimensional attention score through a fully connected layer. After normalization, the hidden layer output is weighted and summed to obtain the temporal feature vector (dimension 64) with fused attention.

[0069] Optionally, a lightweight gradient boosting tree, serving as the final decision layer in the hybrid model, is used to fit the nonlinear relationship between the fused features (including the temporal feature vector F extracted by the bidirectional long short-term memory network and the original feature matrix X) and the label y. The number of trees is set to 200, the maximum tree depth to 6, and the learning rate to 0.03; the minimum number of sample splits is 6, and the minimum number of leaf node samples is 2; the regularization parameter lambda=0.05; the objective function is mean squared error, and feature parallel optimization is enabled to improve training speed.

[0070] Optionally, the TensorFlow framework can be used to train the bidirectional long short-term memory network and attention mechanism, while the Light GBM framework can be used to train the gradient boosting tree. A two-stage stacking strategy is employed. Specifically, the reconstructed 3D time-series data can be input into the bidirectional long short-term memory network with the attention mechanism embedded, using mean squared error as the loss function and the Adam optimizer for training. After training, this network is retained as a feature extractor to extract the fused time-series feature vector for each historical material data point. (Dimension 64), and simultaneously output the base model predictions. Furthermore, the extracted feature vectors Compared with the original feature matrix The features are concatenated to construct an enhanced feature matrix. (Dimension 72). Enhance the feature matrix. The lightweight gradient boosting tree model is trained using these inputs to fit the true labels. Lightweight gradient boosting trees can capture the non-linear correlation between deep temporal features and the original features using a gradient-based one-sided sampling algorithm, outputting meta-model predictions. Finally, based on the formula The predicted values ​​from the base model and the meta-model are weighted and fused to obtain the trained predicted inventory value. And iterative training is performed to train the hybrid prediction model.

[0071] Optionally, the model's performance can be evaluated based on metrics such as mean squared error, mean absolute error, coefficient of determination, and dynamic interval coverage rate, allowing for iterative optimization of the hybrid forecasting model. The dynamic interval coverage rate can be determined by statistically analyzing the percentage of samples in the test set whose actual inventory values ​​fall within the expected dynamic quota range, with a target coverage rate of ≥90%.

[0072] Optionally, after determining the predicted inventory value of each target material for inventory management, the evaluation index value of the target material within a preset evaluation period can be determined; the evaluation index value includes at least one of the following: interval coverage rate, shortage occurrence rate, backlog occurrence rate, cost control rate, and prediction accuracy deviation rate; based on the evaluation index value, when it is detected that the model optimization and / or strategy optimization conditions are met, optimization operations are performed. Among them, the interval coverage rate refers to the percentage of months in which the actual inventory falls within the quota range (target value ≥ 85%). The higher the coverage rate, the stronger the rationality of the quota range. The shortage occurrence rate is the percentage of months in which the actual inventory is lower than the safety lower limit (target value ≤ 5%). The lower the occurrence rate, the stronger the supply guarantee capability. The backlog occurrence rate is the percentage of months in which the actual inventory is higher than the cost upper limit (target value ≤ 5%). The lower the occurrence rate, the stronger the cost control capability. The cost control rate is the ratio of the actual inventory holding cost to the budgeted cost (target value ≤ 100%), reflecting the cost control effect. The prediction accuracy deviation rate can be calculated by (actual inventory optimal value - model prediction optimal value) / model prediction optimal value × 100%. The target deviation rate is ≤ 8%, reflecting the model prediction accuracy.

[0073] Optionally, when at least one of the following conditions (1)-(4) is met, it can be determined that the conditions for model optimization and / or policy optimization are met, and an optimization operation is performed.

[0074] (1) The coverage rate of the interval is less than 80% for two consecutive months, or the coverage rate of a single month is less than 75%.

[0075] (2) The shortage or backlog rate is higher than 8% for one consecutive month, or the quarterly average rate is higher than 6%.

[0076] (3) Significant changes in the business of the enterprise (such as capacity expansion ≥30%, addition of core suppliers / elimination of major suppliers, addition of material categories ≥20%, sudden changes in the external market environment (such as raw material price increase ≥50%), and significant adjustments to the production process leading to changes in material demand ≥25%).

[0077] (4) The prediction accuracy deviation rate exceeds ±10% for two consecutive months, or two or more of the model evaluation indicators fail to meet the standards.

[0078] Optionally, iterative optimization can be carried out from four perspectives: feature optimization, data update, model parameter tuning, and constraint coefficient adjustment. Specifically, for feature optimization, features related to business changes can be added (such as adding "capacity expansion coefficient" when capacity expands, adding "supplier change frequency" when suppliers change, and adding "price fluctuation warning coefficient" when market prices change suddenly); the weights of existing features can be adjusted (such as increasing the weight of supply risk-related features by 30% when the supply risk occurrence rate is too high; adjusting the weight of cost-related features by 20% when costs exceed the budget); and redundant features (features with mutual information values ​​below 0.08) can be eliminated to ensure the adaptability of features to business needs.

[0079] For data updates, the training set can be expanded with the latest year's multi-source data, and the statistical benchmarks for feature engineering (such as mean, standard deviation, maximum and minimum values) can be updated. Outdated data (such as historical data older than 8 years and old data before business changes) can be removed, and new data after business changes can be added (such as delivery data of new suppliers and consumption data corresponding to new capacity) to ensure the timeliness and relevance of training data.

[0080] For model parameter tuning, Bayesian optimization algorithms can be used to re-optimize the hyperparameters of the hybrid model, adjusting the number of hidden layer neurons in the bidirectional long short-term memory network, the mapping dimension of the attention mechanism, and the learning rate and number of lightweight gradient boosting trees. For datasets with changes in business requirements, retrain the model and use a new test set to evaluate model performance to ensure that the optimized model meets the target metrics. If the model performance improvement is not significant, consider adjusting the model architecture (such as adding hidden layers or changing the type of attention mechanism).

[0081] Regarding the adjustment of constraint coefficients, the benchmark values ​​of the safety assurance coefficient γ and the cost control coefficient δ can be adjusted according to the assessment results. For example, if the shortage rate is too high, the benchmark value of γ is increased by 0.03; if the backlog rate is too high, the benchmark value of δ is increased by 0.02; if the cost control rate is below 90% and the supply guarantee is sufficient, the benchmark value of γ is decreased by 0.02, thus balancing the goals of supply guarantee and cost control.

[0082] For example, after the model iteration and optimization are completed, the latest month's actual data is used for verification. The model can only be deployed and launched after the verification indicators meet the standards. After launch, a one-month trial operation period is set, during which the old and new models are run simultaneously to compare the prediction results and application effects, ensuring the stability and reliability of the optimized model. After the trial operation is successful, the old model is officially replaced, and the parameters and evaluation reports of the old model are archived for easy traceability. Through the above closed-loop iteration, the model can be dynamically adapted to changes in business, continuously improving the accuracy and rationality of inventory quota range prediction.

[0083] Optionally, material inventory management is carried out based on the predicted inventory value of each target material, including: determining the strategy based on the predicted inventory value of each target material and the preset limit value, and determining the corresponding inventory range for each target material; and determining the procurement plan and inventory scheduling scheme for each target material based on the corresponding inventory range for each target material, so as to carry out material inventory management.

[0084] Optionally, after determining the inventory range corresponding to each target material, the method further includes: monitoring the actual inventory of the target materials based on the inventory range corresponding to each target material; and / or visually displaying the target material characteristics and inventory range of each target material.

[0085] For example, the characteristics and inventory ranges of each target material can be visualized based on interval display, trend display, and detail display. Specifically, the annual / quarterly / monthly inventory quota range can be displayed in the form of lower limit - optimal value - upper limit value, and can be filtered and viewed by organization, material category, material importance level, and time range. The interval bar chart is used for visualization, with the optimal value marked in red and the lower and upper limits marked in blue, to intuitively present the quota range range.

[0086] For example, the trend display can show the comparison trend between actual inventory data and the predicted quota range over the past 3 years through a line chart, clearly presenting the deviation between actual inventory and the quota range; it can show the impact of key features (such as supply risk, holding costs, and importance level) on the quota range through a bar chart, and mark the weight of the features; it supports the functions of zooming in, zooming out and downloading trend charts to facilitate data analysis.

[0087] For example, when a user selects a quota range for a certain material, they can view detailed information such as feature details, model prediction process (predicted values ​​of each module, fusion process, deviation correction results), and range calculation basis (coefficient values, index calculation process). It supports exporting prediction reports in different formats, which include prediction results, feature analysis, range calculation process, model evaluation indicators, etc., to meet business approval and archiving needs.

[0088] Optionally, based on the predicted inventory value of each target material, a strategy is determined based on preset limit values ​​to determine the corresponding inventory range for each target material, including: determining the lower limit of inventory based on the predicted inventory value, preset safety contraction coefficient, supply stability coefficient, and material importance coefficient; determining the upper limit of inventory based on the predicted inventory value, preset cost expansion coefficient, batch advantage index, and holding cost index; and determining the corresponding inventory range for each target material based on the corresponding upper and lower inventory limits.

[0089] For example, based on the predicted inventory value, the preset safety contraction coefficient, the supply stability coefficient, and the material importance coefficient, it can be based on the formula Determine the lower limit of inventory. ,in, It is a forecast of inventory value. This is a preset safety shrinkage coefficient, with a baseline value of 0.15. The smaller the coefficient, the closer the lower limit is to the optimal value, and the larger the safety margin. It is the supply stability coefficient, with a value of 0-1. The larger the value, the more stable the supply (i.e., the lower the risk, and the lower limit can be adjusted downward). This is the importance coefficient of the resource. Core resources are set to 1.0, and general resources are set to 0.5. When a resource is extremely important or its supply is extremely unstable, the value in parentheses approaches 0, making... Approaching upwards (i.e., high water level warning to ensure production); conversely, if the materials are not important and the supply is stable, a lower safety limit is allowed.

[0090] For example, based on the predicted inventory value, the preset cost expansion coefficient, the batch advantage index, and the holding cost index, it can be based on the formula Determine the upper limit of inventory. ,in, It is a forecast of inventory value. This is the preset cost expansion coefficient, with a baseline value of 0.2; It's a bulk advantage index; the larger the purchase volume, the greater the discount. The higher the price, the higher the upper limit can be allowed to be adjusted to obtain a discount; It is a holding cost index; the higher the holding cost, the higher the holding cost. The larger the value, the more restrictive the increase in the upper limit; that is, when the purchase discount is large and the holding cost is low, the upper limit of inventory is allowed to increase; otherwise, the upper limit is strictly controlled.

[0091] The technical solution of this invention collects material data based on a preset collection strategy, determines the collection time, organization, material category, and managed material data corresponding to the managed materials, generates candidate material data, and stores it. In response to an inventory quota forecast request, it determines annual parameters, organizational parameters, and material categories, and filters the candidate material data according to these parameters to determine target materials that meet the filtering criteria. Based on a pre-trained hybrid forecasting model, it determines the predicted inventory value of each target material according to the target material data corresponding to the target materials, and manages material inventory based on the predicted inventory values ​​of each target material. By utilizing comprehensive material data and a forecasting model, scientifically accurate inventory forecasts can be determined, improving the efficiency of material inventory management.

[0092] Example 2 Figure 2This is a structural block diagram of a material inventory management device provided in an embodiment of the present invention. This embodiment is applicable to situations where inventory forecasting is performed on different materials based on material data and a hybrid forecasting model to achieve better inventory management. The material inventory management device provided by the present invention can execute the material inventory management method provided in any embodiment of the present invention, possessing the corresponding functional modules and beneficial effects of the method execution. This material inventory management device can be implemented in hardware and / or software and configured in an electronic device with material inventory management functions, such as… Figure 3 As shown, the material inventory management device may specifically include: The generation module 201 is used to collect material data based on a preset collection strategy, determine the collection time, organization, material category and management material data corresponding to the managed material, so as to generate candidate material data and store it. The filtering module 202 is used to respond to the inventory quota forecasting request, determine the annual parameters, organizational parameters and material categories, and filter the candidate material data according to the annual parameters, organizational parameters and material categories to determine the target materials that meet the filtering conditions. The management module 203 is used to determine the predicted inventory value of each target material based on the pre-trained hybrid prediction model and the target material data corresponding to the target material, and to manage the material inventory based on the predicted inventory value of each target material.

[0093] The technical solution of this invention collects material data based on a preset collection strategy, determines the collection time, organization, material category, and managed material data corresponding to the managed materials, generates candidate material data, and stores it. In response to an inventory quota forecast request, it determines annual parameters, organizational parameters, and material categories, and filters the candidate material data according to these parameters to determine target materials that meet the filtering criteria. Based on a pre-trained hybrid forecasting model, it determines the predicted inventory value of each target material according to the target material data corresponding to the target materials, and manages material inventory based on the predicted inventory values ​​of each target material. By utilizing comprehensive material data and a forecasting model, scientifically accurate inventory forecasts can be determined, improving the efficiency of material inventory management.

[0094] Furthermore, the generation module 201 is specifically used for: Data is collected based on a dual real-time and offline acquisition strategy to determine the acquisition time, organization, and material category of the managed materials, and to determine the managed material data corresponding to the managed materials; the managed material data includes at least one of the following: procurement data, consumption data, inventory data, production data, external data, and special material extended data; The management of material data is processed through anomaly detection, missing data handling, standardization, and fusion to obtain candidate material data, which is then stored based on a strategy of database sharding, table partitioning, and partitioned storage.

[0095] Furthermore, the generation module 201 is specifically used for: Determine if there are any missing data in the managed materials data. If so, determine the missing data type. Missing data types include missing inventory data, missing procurement data, missing consumption data, missing external data, or missing key fields. If the missing data type is a critical field missing, then the managed material data will be removed; If the missing data type is any other than the missing key field, then the missing data of the managed materials will be repaired.

[0096] Furthermore, the above-mentioned device is also used for: Perform parameter verification on annual parameters, organizational parameters, and material categories; If the parameter verification is successful, the candidate material data will be filtered based on the annual parameters, organizational parameters, and material category to determine the target materials that meet the filtering criteria.

[0097] Furthermore, the management module 203 is specifically used for: Feature extraction and optimization processes are performed on the target material data corresponding to the target material to determine the target material characteristics; the target material characteristics include at least one of the following: time-series trend characteristics, business correlation characteristics, risk control characteristics, and cost optimization characteristics; Based on a pre-trained hybrid prediction model, the predicted inventory value of each target material is determined according to its characteristics.

[0098] Furthermore, the management module 203 is specifically used for: Based on the predicted inventory value of each target material, a strategy is determined based on preset limit values ​​to determine the corresponding inventory range for each target material. Based on the corresponding inventory range of each target material, determine the procurement plan and inventory scheduling scheme for each target material in order to manage material inventory.

[0099] Furthermore, the management module 203 is specifically used for: The lower limit of inventory is determined based on the predicted inventory value, the preset safety contraction coefficient, the supply stability coefficient, and the material importance coefficient. The upper limit of inventory is determined based on the predicted inventory value, the preset cost expansion coefficient, the batch advantage index, and the holding cost index. Based on the upper and lower limits of the inventory corresponding to each target material, determine the corresponding inventory range for each target material.

[0100] Example 3 Figure 3 This is a schematic diagram of the structure of the electronic device provided in an embodiment of the present invention. Figure 3 A schematic diagram of an electronic device 10, which can be used to implement embodiments of the present invention, is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0101] like Figure 3 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory 12 or a random access memory 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the read-only memory 12 or loaded from storage unit 18 into the random access memory 13. The random access memory 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, read-only memory 12, and random access memory 13 are interconnected via a bus 14. An input / output interface 15 is also connected to the bus 14.

[0102] Multiple components in electronic device 10 are connected to input / output 15, including: input unit 16, such as a keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as a disk, optical disk, etc.; and communication unit 19, such as a network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0103] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, central processing units, graphics processing units, various special-purpose artificial intelligence computing chips, various processors running machine learning model algorithms, digital signal processors, and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as material inventory management methods.

[0104] In some embodiments, the inventory management method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via read-only memory 12 and / or communication unit 19. When the computer program is loaded into random access memory 13 and executed by processor 11, one or more steps of the inventory management method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the inventory management method by any other suitable means (e.g., by means of firmware).

[0105] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays, application-specific integrated circuits (ASICs), application-specific standard products (ASICs), system-on-a-chip (SoCs), complex programmable logic devices (PLCs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0106] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0107] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory, optical fibers, portable compact disk read-only memory, optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0108] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a cathode ray tube or liquid crystal display) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (e.g., voice input, speech input, or tactile input).

[0109] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0110] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a host product within the cloud computing service system to address the shortcomings of traditional physical hosts and virtual reality services, such as high management difficulty and weak business scalability.

[0111] In one embodiment, the present invention further includes a computer program product, which includes a computer program that, when executed by a processor, implements the material inventory management method of any embodiment of the present invention.

[0112] In the implementation of a computer program product, computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages ​​as well as conventional procedural programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including local area networks (LANs) or wide area networks (WANs), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0113] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0114] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for managing material inventory, characterized in that, include: Based on a preset data collection strategy, material data is collected to determine the collection time, organization, material category, and managed material data corresponding to the managed materials, so as to generate and store candidate material data. In response to the inventory quota forecasting request, determine the annual parameters, organizational parameters, and material categories, and filter the candidate material data based on the annual parameters, organizational parameters, and material categories to determine the target materials that meet the filtering criteria; Based on a pre-trained hybrid prediction model, the predicted inventory value of each target material is determined according to the target material data corresponding to the target material, and material inventory management is carried out based on the predicted inventory value of each target material.

2. The method according to claim 1, characterized in that, Based on a preset data collection strategy, material data is collected to determine the collection time, organization, material category, and managed material data for each material, in order to generate and store candidate material data, including: Data is collected based on a dual real-time and offline acquisition strategy to determine the acquisition time, organization, and material category of the managed materials, and to determine the managed material data corresponding to the managed materials; the managed material data includes at least one of the following: procurement data, consumption data, inventory data, production data, external data, and special material extended data; The management of material data is processed through anomaly detection, missing data handling, standardization, and fusion to obtain candidate material data, which is then stored based on a strategy of database sharding, table partitioning, and partitioned storage.

3. The method according to claim 2, characterized in that, Missing data handling for managed materials includes: Determine if there are any missing data in the managed materials data. If so, determine the missing data type. Missing data types include missing inventory data, missing procurement data, missing consumption data, missing external data, or missing key fields. If the missing data type is a critical field missing, then the managed material data will be removed; If the missing data type is any other than the missing key field, then the missing data of the managed materials will be repaired.

4. The method according to claim 1, characterized in that, After determining the annual parameters, organizational parameters, and material categories, the following also includes: Perform parameter verification on annual parameters, organizational parameters, and material categories; If the parameter verification is successful, the candidate material data will be filtered based on the annual parameters, organizational parameters, and material category to determine the target materials that meet the filtering criteria.

5. The method according to claim 1, characterized in that, Based on a pre-trained hybrid prediction model, the predicted inventory value of each target material is determined according to the target material data, including: Feature extraction and optimization processes are performed on the target material data corresponding to the target material to determine the target material characteristics; the target material characteristics include at least one of the following: time-series trend characteristics, business correlation characteristics, risk control characteristics, and cost optimization characteristics; Based on a pre-trained hybrid prediction model, the predicted inventory value of each target material is determined according to its characteristics.

6. The method according to claim 1, characterized in that, Material inventory management is conducted based on the projected inventory values ​​of each target material, including: Based on the predicted inventory value of each target material, a strategy is determined based on preset limit values ​​to determine the corresponding inventory range for each target material. Based on the corresponding inventory range of each target material, determine the procurement plan and inventory scheduling scheme for each target material in order to manage material inventory.

7. The method according to claim 6, characterized in that, Based on the predicted inventory value of each target material, a strategy is determined based on preset limit values ​​to determine the corresponding inventory range for each target material, including: The lower limit of inventory is determined based on the predicted inventory value, the preset safety contraction coefficient, the supply stability coefficient, and the material importance coefficient. The upper limit of inventory is determined based on the predicted inventory value, the preset cost expansion coefficient, the batch advantage index, and the holding cost index. Based on the upper and lower limits of the inventory corresponding to each target material, determine the corresponding inventory range for each target material.

8. A material inventory management device, characterized in that, include: The generation module is used to collect material data based on a preset collection strategy, determine the collection time, organization, material category, and managed material data corresponding to the managed materials, so as to generate candidate material data and store it. The filtering module is used to respond to inventory quota forecasting requests, determine annual parameters, organizational parameters, and material categories, and filter candidate material data based on the annual parameters, organizational parameters, and material categories to determine target materials that meet the filtering criteria. The management module is used to determine the predicted inventory value of each target material based on the pre-trained hybrid prediction model and the target material data corresponding to the target material, and to manage the material inventory based on the predicted inventory value of each target material.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the material inventory management method according to any one of claims 1-7.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the material inventory management method according to any one of claims 1-7.

Images