CN-121997288-A - Inventory prediction method, system, equipment and medium based on multi-source information

CN121997288ACN 121997288 ACN121997288 ACN 121997288ACN-121997288-A

Abstract

The invention provides an inventory prediction method, system, equipment and medium based on multi-source information, and belongs to the technical field of intelligent warehouse management. The method comprises the steps of collecting multi-source business data such as investment plans, historical inventory, warehouse entry records and the like, carrying out data management by adopting a distributed storage architecture with separated storage and calculation, determining a target material set by preprocessing the historical business data and classifying materials, extracting key influencing factors based on multidimensional feature analysis, constructing and training an inventory time sequence prediction model, receiving a prediction request through a micro-service API, carrying out prediction calculation, and finally carrying out visual display on a prediction result. By combining multi-source data fusion and an intelligent prediction model, accurate prediction and dynamic optimization of inventory are realized, inventory management efficiency is effectively improved, and inventory backlog and inventory shortage risk are reduced.

Inventors

ZHANG JUN
WANG RUIXIN
XU XINYUAN
WANG JIAMIAN
LI PENGFEI
LI JIANGHUA
MA YAN
WANG YOUJIE
WANG LI
ZHANG HONGYUAN

Assignees

国网山东省电力公司物资公司
山东鲁软数字科技有限公司

Dates

Publication Date: 20260508
Application Date: 20251201

Claims (10)

1. A method for inventory forecasting based on multi-source information, comprising: acquiring historical service data through a preset data interface, and storing the historical service data in a distributed storage system adopting a memory and calculation separation architecture; reading the historical service data from the distributed storage system, preprocessing the historical service data to obtain regular historical data, and determining a target material set from the regular historical data based on a material classification system; extracting key influence factors through multidimensional feature analysis based on the target material set and corresponding regular historical data; Based on the key influence factors and the regular historical data, constructing and training an inventory time sequence prediction model by adopting a time sequence prediction algorithm; receiving a prediction request through a micro-service API, and calling an inventory timing prediction model to execute prediction calculation based on the prediction request; And returning the prediction calculation result to the front-end interface for visual display.
2. The method for predicting inventory based on multi-source information according to claim 1, wherein the steps of collecting historical service data through a preset data interface and storing the historical service data in a distributed storage system using a memory-computation separation architecture include: Collecting historical service data from an enterprise resource planning system and a warehouse management system through a preset standard data interface based on RESTful API and a JDBC data connector, wherein the historical service data comprises investment plan data, historical inventory data and material warehouse-in and warehouse-out data; The Hadoop HDFS is adopted as a distributed storage system, and the namespaces and metadata of the file system, the mapping relation of files to data blocks and the position information of the data blocks are uniformly managed through NameNode nodes; dividing the collected historical service data into a plurality of data blocks according to a preset partitioning rule, generating a unique identifier for each data block, and recording the mapping relation between the partitioning identifiers and the storage positions in a metadata server; copying the data blocks into a plurality of data copies according to the preset copy number, and storing the data copies in different storage server nodes in a distributed mode to form a storage resource pool; Based on the mapping relation in the metadata server, the data access request is routed to the corresponding storage server node through the load balancing server, and meanwhile, the load of each node is dynamically monitored and the data distribution is adjusted.
3. The multi-source information based inventory prediction method according to claim 2, wherein the steps of reading the historical business data from the distributed storage system, preprocessing the historical business data to obtain regular historical data, and determining the target material set from the regular historical data based on the material classification system include: based on preset query conditions, historical service data is read from a distributed storage system through a data access interface, and integrity check and format verification are carried out on the read data; sequentially executing missing value processing, abnormal value detection and correction and data format standardization processing on the history service data passing verification to generate regular history data with unified space-time dimension; Classifying materials in the regular historical data according to the levels of major classes, medium classes and minor classes based on a preset material classification system, adopting a K-means clustering algorithm, identifying material groups with similar inventory characteristics by taking inventory turnover rate, average inventory quantity and seasonal fluctuation coefficient of the materials as clustering characteristic vectors, and determining a target material set from the material groups according to a preset clustering characteristic vector threshold; The main class comprises power distribution equipment, cables and switch equipment, the middle class comprises a box-type transformer station, a ring main unit, power cables and control cables, and the minor class comprises the voltage class and specification model of the middle class.
4. The multi-source information based inventory prediction method according to claim 3, wherein the extracting key impact factors through multi-dimensional feature analysis based on the target material set and the corresponding structured historical data comprises: Extracting investment plan data, historical inventory data and material warehouse-in and warehouse-out data related to a target material set from regular historical data, wherein the investment plan data comprises investment amount and execution time, the historical inventory data comprises inventory quantity and inventory amount, and the material warehouse-in and warehouse-out data comprises warehouse-in and warehouse-out time and quantity; Generating an investment plan duty ratio by calculating the ratio of the current investment amount to the total investment amount based on the extracted data, generating a production supply lead period by calculating the difference between the creation time and the warehousing time of the purchase order, and generating an inventory turnover ratio by calculating the ratio of the warehouse cost to the average inventory cost as a characteristic factor; and calculating the correlation coefficient between each characteristic factor and the inventory variation by adopting a Pearson correlation coefficient analysis method, and screening out the characteristic factors with the absolute values of the correlation coefficients larger than a preset threshold value as key influence factors.
5. The multi-source information based inventory prediction method according to claim 4, wherein the constructing and training an inventory timing prediction model using a time series prediction algorithm based on the key impact factors and the structured historical data comprises: Taking inventory amount in the regular historical data as a target time sequence, taking a key influence factor as an exogenous variable sequence, and aligning and reorganizing according to a preset time granularity to form a model training data set; Selecting an ARIMA model or an LSTM neural network model according to the data characteristics to construct an inventory timing prediction model; when the ARIMA model is selected, determining an autoregressive order p, a differential order d and a moving average order q of the model through an autocorrelation graph and a partial autocorrelation graph; the model training data set is divided into a training set and a verification set according to time sequence, the training set is used for carrying out parameter training on the inventory time sequence prediction model, if the inventory time sequence prediction model adopts an ARIMA model, a maximum likelihood estimation method is adopted for carrying out parameter estimation, if the inventory time sequence prediction model adopts an LSTM model, a back propagation algorithm is adopted and the mean square error of a predicted value and a true value is adopted as a loss function for optimization; and selecting the inventory time sequence prediction model parameter configuration with the minimum root mean square error on the verification set as a final model, and storing a model structure and parameters.
6. The multi-source information based inventory prediction method according to claim 5, wherein the receiving a prediction request through the micro service API, invoking an inventory timing prediction model based on the prediction request to perform a prediction calculation, comprises: receiving a prediction request through a SpringCloud framework-based micro-service API gateway, wherein the prediction request comprises JSON format data of prediction date, material coding and factory information; carrying out parameter analysis and data validity verification on the prediction request to ensure that the prediction date is in an effective range, the material codes exist in the target material set and the factory information has access rights; Inquiring corresponding regular historical data and key influence factors from a distributed storage system according to the request parameters, and carrying out standardized processing on the inquired data according to the input requirements of the inventory time sequence prediction model to generate model input data comprising time sequence window data and corresponding key influence factor values; And loading an inventory time sequence prediction model, inputting the model input data into the model for forward calculation, and generating prediction result data comprising the current inventory amount, the estimated storage amount, the estimated consumption amount and the estimated current inventory amount.
7. The method for predicting inventory based on multi-source information of claim 6, wherein the step of returning the prediction calculation result to the front-end interface for visual display comprises: packaging the prediction result data according to a preset JSON data format; returning the packaged prediction result data to a front-end interface through a micro-service API; On the front-end interface, based on a Vue.js frame and ElementUI component library, visually rendering the packaged prediction result through a ECharts chart library to generate a comprehensive display interface comprising an inventory trend chart, an amount comparison chart and a detail data table; and providing interactive functions of data drill-down, time range screening and multidimensional comparison in the comprehensive display interface, and supporting a user to trigger a new prediction request through interface operation.
8. A multi-source information-based inventory prediction system employing the multi-source information-based inventory prediction method according to any one of claims 1 to 7; The system comprises: The data acquisition and storage module is used for acquiring historical service data through a preset data interface and storing the historical service data in a distributed storage system adopting a memory and calculation separation architecture; The data preprocessing and target determining module is used for reading the historical service data from the distributed storage system, preprocessing the historical service data to obtain regular historical data, and determining a target material set from the regular historical data based on a material classification system; The feature analysis and factor extraction module is used for extracting key influence factors through multidimensional feature analysis based on the target material set and the corresponding regular historical data; The model construction and training module is used for constructing and training an inventory time sequence prediction model by adopting a time sequence prediction algorithm based on the key influence factors and the regular historical data; the prediction calculation module is used for receiving a prediction request through the micro-service API, calling an inventory timing prediction model based on the prediction request and executing prediction calculation; and the prediction result feedback and display module is used for returning the prediction calculation result to the front-end interface for visual display.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the multi-source information based inventory prediction method of any one of claims 1 to 7 when the program is executed.
10. A storage medium having stored thereon a computer program, which when executed by a processor performs the steps of the multi-source information based inventory prediction method according to any one of claims 1 to 7.

Description

Inventory prediction method, system, equipment and medium based on multi-source information Technical Field The invention belongs to the technical field of intelligent warehouse management, and particularly relates to a multi-source information-based inventory prediction method, a multi-source information-based inventory prediction system, multi-source information-based inventory prediction equipment and a multi-source information-based inventory prediction medium. Background In current enterprise inventory management practices, traditional inventory forecasting methods rely primarily on historical experience of the manager or decision making based on simple statistical models. The method is too dependent on subjective judgment, lacks systematic analysis of multidimensional influence factors, and is difficult to accurately capture dynamic fluctuation and seasonal variation rules of market demands. Because key factors such as investment plans, supply chain periods and the like are not fully considered, larger deviation is often caused between a predicted result and actual demands, and further the contradiction phenomenon of stock backlog and stock shortage coexistence is caused, so that the storage cost is increased, and the normal running of production and operation activities is influenced. In particular, in the management of large items and critical materials, traditional inventory management modes expose significant limitations. Because the manufacturing period of some key materials is longer, if purchasing decision is made only according to experience, the purchasing decision cannot be effectively matched with the implementation progress of the project. Such supply and demand disjoint may cause that the materials cannot be delivered on time, directly delay project construction period, and affect the key production progress of enterprises. In order to avoid the risk of shortage, an excessive pre-reserve strategy is adopted, and a series of derivative problems such as overhigh mobile funds occupation, increased storage management complexity, low inventory turnover efficiency and the like are caused. Some existing informationized inventory management systems realize the electronization of data records to a certain extent, but the core prediction model still adopts a relatively simple statistical method, and lacks of deep fusion and intelligent analysis on multi-source business data such as investment plans, supply periods, inventory turnover and the like. These systems have difficulty in building accurate time sequence prediction models, cannot effectively adapt to complex and changeable market environments, and cause insufficient flexibility and accuracy of inventory strategies. Disclosure of Invention Aiming at the problems, the invention aims to provide an inventory prediction method, a system, equipment and a medium based on multi-source information, which realize accurate prediction, dynamic optimization and efficient decision of inventory management by constructing a time sequence prediction model of multi-source data fusion and combining a distributed storage and micro-service architecture. The invention aims to achieve the aim, and the aim is achieved by the following technical scheme: in a first aspect, an embodiment of the present application provides a method for inventory prediction based on multi-source information, including: acquiring historical service data through a preset data interface, and storing the historical service data in a distributed storage system adopting a memory and calculation separation architecture; reading the historical service data from the distributed storage system, preprocessing the historical service data to obtain regular historical data, and determining a target material set from the regular historical data based on a material classification system; extracting key influence factors through multidimensional feature analysis based on the target material set and corresponding regular historical data; Based on the key influence factors and the regular historical data, constructing and training an inventory time sequence prediction model by adopting a time sequence prediction algorithm; receiving a prediction request through a micro-service API, and calling an inventory timing prediction model to execute prediction calculation based on the prediction request; And returning the prediction calculation result to the front-end interface for visual display. In an optional embodiment, the collecting the historical service data through the preset data interface and storing the historical service data in a distributed storage system adopting a memory-computation separation architecture includes: Collecting historical service data from an enterprise resource planning system and a warehouse management system through a preset standard data interface based on RESTful API and a JDBC data connector, wherein the historical service data comprises investment plan data, historical inve