CN117807114B - Logistics information intelligent retrieval method, system, equipment and storage medium - Google Patents
Logistics information intelligent retrieval method, system, equipment and storage medium Download PDFInfo
- Publication number
- CN117807114B CN117807114B CN202410236355.4A CN202410236355A CN117807114B CN 117807114 B CN117807114 B CN 117807114B CN 202410236355 A CN202410236355 A CN 202410236355A CN 117807114 B CN117807114 B CN 117807114B
- Authority
- CN
- China
- Prior art keywords
- data retrieval
- retrieval
- node
- data
- target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 65
- 238000003860 storage Methods 0.000 title claims abstract description 15
- 239000011159 matrix material Substances 0.000 claims abstract description 117
- 238000005457 optimization Methods 0.000 claims abstract description 64
- 238000004458 analytical method Methods 0.000 claims abstract description 58
- 238000007621 cluster analysis Methods 0.000 claims abstract description 34
- 230000004927 fusion Effects 0.000 claims abstract description 18
- 238000004422 calculation algorithm Methods 0.000 claims description 33
- 238000004364 calculation method Methods 0.000 claims description 27
- 238000012545 processing Methods 0.000 claims description 27
- 230000006870 function Effects 0.000 claims description 24
- 238000009826 distribution Methods 0.000 claims description 23
- 230000007246 mechanism Effects 0.000 claims description 20
- 238000013528 artificial neural network Methods 0.000 claims description 11
- 238000006243 chemical reaction Methods 0.000 claims description 11
- 230000002068 genetic effect Effects 0.000 claims description 8
- 238000013507 mapping Methods 0.000 claims description 6
- 238000000513 principal component analysis Methods 0.000 claims description 6
- 238000012300 Sequence Analysis Methods 0.000 claims description 5
- 238000000354 decomposition reaction Methods 0.000 claims description 5
- 238000012544 monitoring process Methods 0.000 claims description 5
- 238000007781 pre-processing Methods 0.000 claims description 5
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 238000013016 damping Methods 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 230000008569 process Effects 0.000 description 22
- 238000010606 normalization Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 238000012731 temporal analysis Methods 0.000 description 5
- 238000000700 time series analysis Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 4
- 238000000605 extraction Methods 0.000 description 4
- 230000035772 mutation Effects 0.000 description 4
- 230000007613 environmental effect Effects 0.000 description 3
- 230000010365 information processing Effects 0.000 description 3
- 238000007726 management method Methods 0.000 description 3
- 239000007787 solid Substances 0.000 description 3
- 238000012916 structural analysis Methods 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 230000000737 periodic effect Effects 0.000 description 2
- 238000010845 search algorithm Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 230000001364 causal effect Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001934 delay Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000007499 fusion processing Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000001932 seasonal effect Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000013068 supply chain management Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application relates to the technical field of information retrieval and discloses a method, a system, equipment and a storage medium for intelligent retrieval of logistics information. The method comprises the following steps: performing multi-level retrieval node analysis on the target logistics data center to obtain a data center retrieval node, a plurality of upper layer data retrieval nodes and a plurality of lower layer data retrieval nodes; creating a first data retrieval super-network model and performing network cluster analysis to obtain a second data retrieval super-network model; carrying out logistics data retrieval feature analysis to obtain a logistics data retrieval feature set, and carrying out feature coding and matrix fusion to generate a logistics data retrieval feature matrix; dynamic prediction of the data retrieval node is carried out through a variable structure dynamic Bayesian network model, and a target dynamic prediction result is obtained; generating a target model optimization strategy and carrying out model self-adaptive updating on the second data retrieval super-network model to obtain the target data retrieval super-network model.
Description
Technical Field
The application relates to the technical field of information retrieval, in particular to a method, a system, equipment and a storage medium for intelligent retrieval of logistics information.
Background
The logistics industry is an important component of global economy, with efficiency and accuracy critical to the overall supply chain management. With the rapid development of electronic commerce and the increasing diversity of consumer demands, challenges and pressures in the logistics industry are also increasing. Modern logistics not only requires processing of massive amounts of data, but also requires accurate decisions to be made in a short time. Therefore, the intelligent and automatic logistics information retrieval method is a key technology for improving logistics efficiency, reducing operation cost and improving customer satisfaction.
However, conventional logistics information systems often face problems such as data islands, information delays, low retrieval efficiency, and the like. The data island causes that information cannot be integrated effectively, so that the decision process lacks data support, and logistics efficiency and service quality are seriously affected. In addition, as the complexity of the logistics network increases, it has been difficult for conventional data processing methods to meet high requirements for real-time and accuracy. The existence of these problems not only increases the logistical costs, but also reduces the customer's service experience.
Disclosure of Invention
The application provides a method, a system, equipment and a storage medium for intelligently retrieving logistics information.
In a first aspect, the present application provides a method for intelligently retrieving logistics information, where the method for intelligently retrieving logistics information includes:
Performing multi-level retrieval node analysis on the target logistics data center to obtain a data center retrieval node, a plurality of upper layer data retrieval nodes and a plurality of lower layer data retrieval nodes;
Creating a first data retrieval super-network model according to the data center retrieval node, the plurality of upper layer data retrieval nodes and the plurality of lower layer data retrieval nodes, and performing network cluster analysis on the first data retrieval super-network model to obtain a second data retrieval super-network model;
Based on the second data retrieval super-network model, carrying out logistics data retrieval feature analysis on a plurality of target data sources in the target logistics data center respectively to obtain a logistics data retrieval feature set of each target data source;
respectively carrying out feature coding and matrix fusion on the logistics data retrieval feature set to generate a corresponding logistics data retrieval feature matrix;
Inputting the logistics data retrieval feature matrix into a preset variable structure dynamic Bayesian network model to conduct data retrieval node dynamic prediction, and obtaining a target dynamic prediction result;
and generating a corresponding target model optimization strategy according to the target dynamic prediction result, and carrying out model self-adaptive updating on the second data retrieval super network model according to the target model optimization strategy to obtain a target data retrieval super network model.
In a second aspect, the present application provides a smart physical distribution information retrieval system, comprising:
The analysis module is used for carrying out multi-level retrieval node analysis on the target logistics data center to obtain a data center retrieval node, a plurality of upper layer data retrieval nodes and a plurality of lower layer data retrieval nodes;
The creation module is used for creating a first data retrieval super-network model according to the data center retrieval node, the plurality of upper layer data retrieval nodes and the plurality of lower layer data retrieval nodes, and performing network cluster analysis on the first data retrieval super-network model to obtain a second data retrieval super-network model;
the processing module is used for respectively carrying out logistics data retrieval characteristic analysis on a plurality of target data sources in the target logistics data center based on the second data retrieval super-network model to obtain a logistics data retrieval characteristic set of each target data source;
The coding module is used for respectively carrying out feature coding and matrix fusion on the logistics data retrieval feature set to generate a corresponding logistics data retrieval feature matrix;
The prediction module is used for inputting the logistics data retrieval feature matrix into a preset variable structure dynamic Bayesian network model to conduct dynamic prediction of the data retrieval nodes, so as to obtain a target dynamic prediction result;
and the updating module is used for generating a corresponding target model optimization strategy according to the target dynamic prediction result, and carrying out model self-adaptive updating on the second data retrieval super network model according to the target model optimization strategy to obtain a target data retrieval super network model.
A third aspect of the present application provides a computer apparatus comprising: a memory and at least one processor, the memory having instructions stored therein; the at least one processor invokes the instructions in the memory to cause the computer device to perform the logistic information wisdom retrieval method described above.
A fourth aspect of the present application provides a computer readable storage medium having instructions stored therein which, when run on a computer, cause the computer to perform the above-described logistic information wisdom retrieval method.
According to the technical scheme provided by the application, the core nodes and the hierarchical relationship thereof of the data center can be accurately identified and analyzed by carrying out multi-level retrieval node analysis on the logistics data center. This in-depth structural analysis helps to better understand the flow and processing mechanisms of the logistics data, thereby providing a solid foundation for efficient data retrieval and information extraction. The first data retrieval super network model and the network cluster analysis are utilized to construct and optimize a second data retrieval super network model. The method can reflect complex logistics data relation, and find out key functional areas or information processing clusters of the data center through network cluster analysis, so that accuracy and efficiency of data retrieval are improved. By carrying out logistics data retrieval feature analysis on each target data source and carrying out feature coding and matrix fusion, key information in logistics data can be fully mined and utilized. The feature fusion not only improves the expression capability of the data, but also improves the robustness and accuracy of the retrieval method by integrating different types of features. The feature matrix is input into a preset variable structure dynamic Bayesian network model, so that dynamic prediction of the logistics data retrieval node can be realized. The prediction not only considers historical data, but also can adapt to environmental changes through a dynamic model, and provides more accurate and real-time prediction results. And generating a target model optimization strategy according to the dynamic prediction result, and carrying out self-adaptive updating on the data retrieval super-network model. The self-optimizing mechanism ensures that the system can continuously provide high-quality retrieval service along with the time and the environment change, and the application improves the retrieval accuracy and efficiency of the logistics information.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings may be obtained based on these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a diagram illustrating an embodiment of a smart retrieval method for stream information according to an embodiment of the present application;
FIG. 2 is a diagram of an embodiment of a smart retrieval system for information about a flow of material.
Detailed Description
The embodiment of the application provides a method, a system, equipment and a storage medium for intelligently retrieving logistics information. The terms "first," "second," "third," "fourth" and the like in the description and in the claims and in the above drawings, if any, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments described herein may be implemented in other sequences than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed or inherent to such process, method, article, or apparatus.
For easy understanding, the following describes a specific flow of an embodiment of the present application, referring to fig. 1, and one embodiment of a method for intelligently retrieving information about a physical stream in an embodiment of the present application includes:
Step 101, carrying out multi-level retrieval node analysis on a target logistics data center to obtain a data center retrieval node, a plurality of upper layer data retrieval nodes and a plurality of lower layer data retrieval nodes;
it can be understood that the execution subject of the present application may be a smart retrieval system for information about a stream, and may also be a terminal or a server, which is not limited herein. The embodiment of the application is described by taking a server as an execution main body as an example.
Specifically, the multi-level retrieval node identification is performed on the target logistics data center by analyzing the data structure of the data center and the data flow mode in the data center. In the identification process, the data center retrieval node and a plurality of initial data retrieval nodes are identified by analyzing the data flow direction and the data type. And carrying out node degree centrality calculation on each initial data retrieval node, and measuring the importance and influence of each node in the whole network. The calculation of the centrality of the node takes into account the degree of the node, i.e. the number of connections of one node to other nodes, and the total number of nodes in the whole network, so as to evaluate the centrality of each node. A proximity centrality calculation is performed that reflects the proximity of each node to other nodes in the network. And measuring the position and influence of the nodes in the whole network by calculating the shortest path length among the nodes. The proximity to the centrality indicates the efficiency and importance of the nodes in data transmission and information flow. And calculating the betting center of each initial data retrieval node. The calculation of the betting centrality takes into account the shortest paths in the network and analyzes the number of these paths through a particular node. This index can reflect the bridging effect of a node in the network, i.e. the effect that the node plays in the network connecting the different parts. According to the node degree centrality, the proximity centrality and the betweenness centrality, the initial data retrieval nodes are hierarchically divided to obtain an initial hierarchical result of each node, and the hierarchical position of the node depends on the importance and the effect of the node in the network. And according to the initial layering result, performing PageRank value calculation on each initial data retrieval node. The PageRank algorithm is a method of evaluating the importance of a page or node through a link structure in a network, which considers the number and quality of other nodes pointing to a node. The nodes can be further classified through the PageRank value obtained through calculation, and finally a plurality of upper layer data retrieval nodes and a plurality of lower layer data retrieval nodes are determined.
102, Creating a first data retrieval super-network model according to a data center retrieval node, a plurality of upper layer data retrieval nodes and a plurality of lower layer data retrieval nodes, and performing network cluster analysis on the first data retrieval super-network model to obtain a second data retrieval super-network model;
Specifically, the relation analysis is carried out on the data center retrieval node and the plurality of upper layer data retrieval nodes and the plurality of lower layer data retrieval nodes through a preset graph neural network algorithm. The graph neural network algorithm can effectively process and analyze complex relationships among the nodes, and potential relationships among the nodes are identified by learning the connection modes among the nodes. By analysis, two sets of relationships were obtained: the first group is a relationship between a data center retrieval node and an upper layer data retrieval node, and the second group is a relationship between a data center retrieval node and a lower layer data retrieval node. Based on the analysis result, a first relationship edge between the data center retrieval node and the upper node and a second relationship edge between the data center retrieval node and the lower node are respectively created according to the first group and the second group of retrieval node relationships. The construction of the relation edges is based on the interaction and data flow characteristics among the nodes, so that the super network model can truly reflect the actual structure and operation mode of the logistics data center. And connecting the first relation edge and the second relation edge with a super network model based on a preset initial node edge weight, so as to generate a corresponding first data retrieval super network model. The initial node edge weight is set based on importance of nodes and the degree of closeness of relationships among the nodes, so that the super network model is ensured to have certain data reflecting capacity and logic property during initial construction. And performing network cluster analysis on the first data retrieval super-network model, and further revealing and optimizing node structures and relations in the network through a clustering algorithm. The network cluster analysis result can show the connection modes and the characteristics of different nodes and node groups, and provides important basis for subsequent model optimization. And performing weight optimization on the first data retrieval super-network model based on the target node edge weight generated by the network cluster analysis result. And adjusting the importance of the relation between the nodes according to the insight of the cluster analysis, thereby optimizing the structure and the performance of the whole super network model. Through weight optimization, the finally obtained second data retrieval super-network model can reflect the operation characteristics and the data flow mode of the logistics data center more accurately and efficiently.
Step 103, based on the second data retrieval super-network model, carrying out logistics data retrieval feature analysis on a plurality of target data sources in the target logistics data center respectively to obtain a logistics data retrieval feature set of each target data source;
Specifically, the physical distribution information retrieval monitoring is carried out on a plurality of target data sources in the target physical distribution data center based on the second data retrieval super-network model, and initial physical distribution information retrieval log data of each target data source are collected. Detailed log information for each data source is obtained by tracking and recording operations of the data center, including query requests, data processing, and data output, in real time. And carrying out data preprocessing on the initial logistics information retrieval log data of each target data source, and improving the quality and usability of the data, wherein the data preprocessing comprises the steps of data cleaning, data conversion, data normalization and the like. Through this process, extraneous or erroneous data is removed, the data format is converted to be suitable for analysis, and the data is normalized to eliminate differences between different data sources. The target logistics information retrieval log data obtained after the processing is more suitable for further analysis. And carrying out time series analysis on the target logistics information retrieval log data of each target data source. The time series analysis is to analyze the trend of the data with time to reveal the characteristics and rules of the data, and calculate a plurality of time series characteristic values of each target data source by using a time series analysis function. The function calculates correlation characteristics of the data at different time lags by comparing the relationship between the data values at different time points and the overall data mean. Such analysis may reveal periodicity, trending, and randomness of the data, helping to understand the operational mode of the logistics data center. Feature normalization processing and feature set conversion are performed on a plurality of time-series feature values of each target data source. Feature normalization is to eliminate the dimensional influence between different feature values so that the feature values are comparable at the same scale. Feature set conversion is to convert these feature values into a format that can be efficiently handled by the data retrieval model. And finally obtaining a logistics data retrieval feature set of each target data source, wherein the feature set contains key information about the data source.
104, Respectively carrying out feature coding and matrix fusion on the logistics data retrieval feature sets to generate corresponding logistics data retrieval feature matrixes;
Specifically, feature encoding is performed on a collection of physical distribution data retrieval features, converting various features of physical distribution data into a format suitable for calculation and analysis, generally involving converting category data into digital form, and normalizing the dimensions of the different features. And obtaining a coded data retrieval characteristic set of each target data source after coding. And (3) performing linear matrix conversion on the data retrieval feature set, and converting the feature set into a linear matrix form so as to be suitable for complex mathematical operation. And performing matrix operation on the coded data retrieval characteristic linear matrix of each target data source by using a kernel function in a kernel principal component analysis algorithm. The kernel principal component analysis (KERNEL PCA) is a nonlinear dimension reduction technique, and the kernel function is used for mapping data to a higher-dimension space, so that a more complex structure and relation of the data can be revealed in the new space, and an initial data retrieval characteristic kernel matrix of each target data source is obtained. To further refine these features, an initial data retrieval feature kernel matrix for each target data source is centered, and the data in the matrix is adjusted to zero mean. After the centering processing, the characteristic value decomposition is carried out on the target data retrieval characteristic core matrix of each target data source, so as to obtain the characteristic value and the characteristic vector corresponding to each target data source. These eigenvalues and eigenvectors are key to understanding the main variables of the data, and they reveal the most important patterns and trends in the data. And selecting a main component corresponding to each target data source according to the size of the obtained characteristic value, and mapping the logistics data retrieval characteristic set to a new characteristic space to obtain a dimension-reduction data retrieval characteristic matrix corresponding to each target data source. The dataset is simplified by reducing the number of features while retaining as much of the most important information as possible. And carrying out matrix fusion on the reduced-dimension data retrieval feature matrix corresponding to each target data source to generate a final logistics data retrieval feature matrix.
Step 105, inputting the logistics data retrieval feature matrix into a preset variable structure dynamic Bayesian network model to conduct dynamic prediction of data retrieval nodes, and obtaining a target dynamic prediction result;
Specifically, the logistics data retrieval feature matrix is input into a preset variable structure dynamic Bayesian network model to conduct data retrieval node dynamic prediction, and future trend and change of logistics information are accurately predicted through the model. And performing feature cluster analysis on the feature matrix of the stream data retrieval through a K nearest neighbor algorithm layer. By analyzing the similarity of the individual data points in the feature matrix, similar features are clustered together to generate a target data retrieval feature matrix, which helps reveal potential patterns and structures in the data. And carrying out attention mechanism weighted analysis on the target data retrieval feature matrix through a core attention mechanism layer. The kernel attention mechanism layer can identify and assign higher weights to those features that are more important to the predicted outcome, thereby generating an attention data retrieval feature matrix. And inputting the attention data retrieval feature matrix into a preset three-layer Bayesian network. A connection between the attention data retrieval feature matrix and each layer in the three-layer bayesian network is established. Specifically, a state characteristic node connection is established between the first layer Bayesian network and the attention data retrieval characteristic matrix, a state compensation node connection is established between the second layer Bayesian network and the attention data retrieval characteristic matrix, and a dynamic prediction node connection is established between the third layer Bayesian network and the attention data retrieval characteristic matrix. The establishment of these connections enables the bayesian network to analyze and process the data in the feature matrix from different levels and angles. Forward deducing the attention data retrieval feature matrix through a three-layer Bayesian network, and calculating probability distribution of each dynamic prediction node, each state compensation node and each state feature node. In this process, the bayesian network estimates the probability distribution of different nodes by taking into account various causal relationships and interdependencies. The result of the forward inference is a target dynamic prediction result, i.e. a prediction of the future trend of the logistics data.
And 106, generating a corresponding target model optimization strategy according to the target dynamic prediction result, and carrying out model self-adaptive updating on the second data retrieval super network model according to the target model optimization strategy to obtain the target data retrieval super network model.
Specifically, according to the target dynamic prediction result, model optimization strategy analysis is carried out on the second data retrieval super network model, and according to the prediction result, defects and improvement space in the model are identified, so that an initial model optimization strategy is obtained. This includes analyzing structural, parameter settings, and prediction accuracy aspects of the model, improving performance and accuracy of the model in future applications. And carrying out global scheme initialization on the target logistics data center according to the initial model optimization strategy by a preset genetic algorithm to generate a plurality of candidate model optimization strategies. The genetic algorithm is a search algorithm for simulating the biological evolution process, and the optimal solution is found through the processes of iteration, selection, crossover, mutation and the like of the population. The genetic algorithm finds a more efficient model adjustment scheme through evaluation and optimization of multiple candidate model optimization strategies. And respectively calculating the fitness value of each candidate model optimization strategy, and carrying out group division and optimization solution on the candidate strategies according to the fitness value. The calculation of fitness values is based on the expected contribution of each candidate strategy to the model improvement effect, with high fitness values meaning that the strategy is more likely to improve the overall performance of the model. Through division and optimization solving of strategy groups, the most appropriate model optimization strategy is ensured to be selected. And carrying out self-adaptive updating on the second data retrieval super network model according to the target model optimization strategy, thereby obtaining the target data retrieval super network model. The self-adaptive updating refers to adjusting the structure, parameters or learning mode of the model according to an optimization strategy so as to better adapt to the data characteristics and the prediction requirements. The method ensures that the super network model can keep high efficiency and accuracy when processing complex logistics data retrieval tasks, and simultaneously enhances the adaptability of the model to future data changes.
According to the embodiment of the application, the core nodes and the hierarchical relationship thereof of the data center can be accurately identified and analyzed by carrying out multi-level retrieval node analysis on the logistics data center. This in-depth structural analysis helps to better understand the flow and processing mechanisms of the logistics data, thereby providing a solid foundation for efficient data retrieval and information extraction. The first data retrieval super network model and the network cluster analysis are utilized to construct and optimize a second data retrieval super network model. The method can reflect complex logistics data relation, and find out key functional areas or information processing clusters of the data center through network cluster analysis, so that accuracy and efficiency of data retrieval are improved. By carrying out logistics data retrieval feature analysis on each target data source and carrying out feature coding and matrix fusion, key information in logistics data can be fully mined and utilized. The feature fusion not only improves the expression capability of the data, but also improves the robustness and accuracy of the retrieval method by integrating different types of features. The feature matrix is input into a preset variable structure dynamic Bayesian network model, so that dynamic prediction of the logistics data retrieval node can be realized. The prediction not only considers historical data, but also can adapt to environmental changes through a dynamic model, and provides more accurate and real-time prediction results. And generating a target model optimization strategy according to the dynamic prediction result, and carrying out self-adaptive updating on the data retrieval super-network model. The self-optimizing mechanism ensures that the system can continuously provide high-quality retrieval service along with the time and the environment change, and the application improves the retrieval accuracy and efficiency of the logistics information.
In a specific embodiment, the process of executing step 101 may specifically include the following steps:
(1) Carrying out multi-level retrieval node identification on the target logistics data center to obtain a data center retrieval node and a plurality of initial data retrieval nodes;
(2) Respectively carrying out node degree centrality calculation on a plurality of initial data retrieval nodes to obtain the node degree centrality of each initial data retrieval node, wherein the node degree centrality function is as follows: ，/> Node degree centrality representing initial data retrieval node v,/> A degree indicating the initial data retrieval node v, n indicating the total number of nodes of the plurality of initial data retrieval nodes;
(3) Respectively carrying out proximity centrality calculation on a plurality of initial data retrieval nodes to obtain the proximity centrality of each initial data retrieval node, wherein the proximity centrality function is as follows: ，/> representing the proximity centrality of the initial data retrieval node v,/> Representing the shortest path length between the initial data retrieval node v and the initial data retrieval node u, n representing the total number of nodes of the plurality of initial data retrieval nodes;
(4) Performing betweenness centrality calculation on the plurality of initial data retrieval nodes respectively to obtain betweenness centrality of each initial data retrieval node, wherein betweenness centrality functions are as follows: ，/> representing the betweenness centrality of the initial data retrieval node v,/> Representing the number of all shortest paths between the initial data retrieval node s and the initial data retrieval node t,/>Representing the number of shortest paths between the initial data retrieval node s and the initial data retrieval node t passing through the initial data retrieval node v;
(5) Performing node hierarchical division on a plurality of initial data retrieval nodes according to node degree centrality, approximate centrality and medium centrality to obtain an initial hierarchical result of each initial data retrieval node;
(6) Performing PageRank calculation on a plurality of initial data retrieval nodes according to the initial layering result to obtain a PageRank value of each initial data retrieval node, wherein the PageRank function is as follows: ，/> PageRank value representing initial data retrieval node v, B v represents all node sets pointing to initial data retrieval node v,/> Representing the number of outer chains of the initial data retrieval node u,/>PageRank value representing initial data retrieval node u, d representing damping coefficient, set to 0.85;
(7) And classifying the nodes of the plurality of initial data retrieval nodes according to the PageRank value to obtain a plurality of upper layer data retrieval nodes and a plurality of lower layer data retrieval nodes.
Specifically, the data flow, the processing flow and the data structure of the target logistics data center are analyzed, and the retrieval node and a plurality of initial data retrieval nodes of the data center are identified. For example, a logistics data center may include a plurality of data processing points such as order processing, goods tracking, customer feedback, etc., each of which may be considered an initial data retrieval node. And carrying out node degree centrality calculation on the initial data retrieval nodes, and evaluating the influence of each node in the whole network. Node centrality calculation scales the importance of one node by considering its number of direct connections with other nodes. For example, if one data retrieval node is directly connected to multiple other nodes, its role in the overall data stream is more pronounced. In this way, key nodes in the processing of logistics data, such as nodes that process a large amount of orders or information, are identified. And (5) performing proximity centrality calculation, and evaluating the proximity degree of the data retrieval node and other nodes in the network. The proximity centrality is determined by analyzing the shortest path between nodes, the shorter the distance between nodes, the higher its proximity centrality. For example, a logistics data node may be said to play a critical role in data transmission and communication if it is able to access other parts of the network quickly. And performing the betting centrality calculation to determine the intermediation of each node in the network. The betting center is measured by considering the importance of a node in the network as a path to connect with other nodes. For example, if a logistics data node often appears on the shortest path between different nodes, it plays an important role in the information flow throughout the network. The initial data retrieval nodes are hierarchically partitioned based on node centrality, near centrality, and medium centrality. For example, some key nodes are divided into upper layer data retrieval nodes, while other auxiliary or secondary nodes are divided into lower layer data retrieval nodes. PageRank calculation is performed, and the importance of each node in the whole network is evaluated. PageRank evaluates the importance of a node by considering how many other nodes it is referenced or connected to. The PageRank high nodes may be nodes that exchange data with other nodes frequently, such as a core's order processing system or a centralized customer service center. The PageRank value of each node is obtained through calculation, and the importance of each node in the network is further defined. And classifying the initial data retrieval nodes according to the PageRank value, and further determining which nodes belong to the upper layer data retrieval nodes and which nodes belong to the lower layer data retrieval nodes. Such classification not only helps to optimize data retrieval and processing flows, but also makes management and monitoring of the data center more efficient and targeted.
In a specific embodiment, the process of executing step 102 may specifically include the following steps:
(1) Performing relationship analysis on the data center retrieval node and the plurality of upper layer data retrieval nodes through a preset graph neural network algorithm to obtain a plurality of first retrieval node relationships, and performing relationship analysis on the data center retrieval node and the plurality of lower layer data retrieval nodes through the graph neural network algorithm to obtain a plurality of second retrieval node relationships;
(2) Creating a plurality of first relationship edges between the data center retrieval node and the plurality of upper layer data retrieval nodes according to the plurality of first retrieval node relationships, and creating a plurality of second relationship edges between the data center retrieval node and the plurality of lower layer data retrieval nodes according to the plurality of second retrieval node relationships;
(3) Performing super-network model connection on a plurality of first relation edges and a plurality of second relation edges based on preset initial node edge weights, and generating a corresponding first data retrieval super-network model;
(4) Performing network cluster analysis on the first data retrieval super-network model to generate a network cluster analysis result, and generating corresponding target node edge weights according to the network cluster analysis result;
(5) And performing the optimization of the super network model weight on the first data retrieval super network model based on the target node edge weight to obtain a second data retrieval super network model.
Specifically, the relationship between the data center retrieval node and the plurality of upper layer data retrieval nodes is identified and understood by carrying out relationship analysis through a preset graph neural network algorithm. The graph neural network algorithm can effectively reveal hidden relations and dependencies among nodes by learning the connection modes among the nodes. For example, a physical distribution data center may include a plurality of upper layer data retrieval nodes for order processing, inventory management, and customer service, where the relationship between these nodes and the data center retrieval nodes is critical to the efficient operation of the overall data stream. And performing relationship analysis on the data center retrieval node and the plurality of lower layer data retrieval nodes by using a graph neural network algorithm to further obtain a second retrieval node relationship. This helps understand how data center retrieval nodes affect underlying nodes such as logistics tracking, cargo distribution, customer feedback, and the like. And obtaining a plurality of first retrieval node relations and a plurality of second retrieval node relations through analysis. And creating a first relation edge between the data center retrieval node and the upper layer data retrieval node and a second relation edge between the data center retrieval node and the lower layer data retrieval node according to the analysis result. These relationship edges ensure that the super network model can accurately reflect the actual operation and data flow conditions of the logistics data center. For example, if the interaction between the order processing node and the data center retrieval node is frequent, then in the super network model, the relationship edges between the two nodes will be given a higher weight. And based on the preset initial node edge weight, performing the super-network model connection on the first relation edge and the second relation edge, and generating a corresponding first data retrieval super-network model. Each relationship edge is given an initial weight that reflects the strength and importance of the relationship between the different nodes. And performing network cluster analysis on the first data retrieval super-network model, and further optimizing the network structure and improving the efficiency of the model. Network cluster analysis helps identify key node groups and potential data flow paths in a network by separating nodes in the network into different groups to reveal structural features within the network and patterns of relationships between nodes. And generating target node edge weights based on the network cluster analysis result, and performing weight optimization on the first data retrieval super-network model according to the target node edge weights. And adjusting the weight of the relation edge according to the insight of the network clustering, so as to optimize the structure and the performance of the whole network. For example, if the cluster analysis shows that a certain group of nodes plays a central role in the data stream, the relationship edge weights between these nodes will be increased to reflect their importance in the network.
In a specific embodiment, the process of executing step 103 may specifically include the following steps:
(1) Based on the second data retrieval super network model, respectively carrying out logistics information retrieval monitoring on a plurality of target data sources in the target logistics data center to obtain initial logistics information retrieval log data of each target data source;
(2) Respectively carrying out data preprocessing on the initial logistics information retrieval log data of each target data source to obtain target logistics information retrieval log data of each target data source;
(3) Performing time sequence analysis on the target logistics information retrieval log data of each target data source respectively to obtain a plurality of time sequence characteristic values of each target data source, wherein the time sequence analysis function is as follows: ，/> time sequence characteristic value of target logistics information retrieval log data at lag k is represented, and is/are Value of target logistics information retrieval log data at time t,/>Represents the average value of the target logistics information retrieval log data, t represents the moment, n represents the total length of the target logistics information retrieval log data, and/>A value representing the target logistics information retrieval log data at time t+k;
(4) And carrying out feature standardization processing and feature set conversion on a plurality of time sequence feature values of each target data source to obtain a logistics data retrieval feature set of each target data source.
Specifically, the physical distribution information retrieval monitoring is carried out on a plurality of target data sources in the target physical distribution data center based on the second data retrieval super network model, and initial physical distribution information retrieval log data of each target data source are collected and recorded. Various operations of the logistics data center are tracked in real time, including order processing, cargo transportation, inventory management, and the like. And carrying out data preprocessing on the initial logistics information retrieval log data, improving the quality of the data and ensuring the accuracy of subsequent analysis. Including data cleansing (removing erroneous or irrelevant data), data conversion (converting the data into a format suitable for analysis), and data normalization (ensuring consistency of the data). For example, for order data in a logistics center, invalid or duplicate order records need to be removed and converted into a unified data format for analysis. And carrying out time sequence analysis on the target logistics information retrieval log data of each target data source, and identifying the time-varying characteristics and trends of the data. Time series analysis reveals the dynamic characteristics of data by calculating statistical characteristics of the data at different time points, such as mean, variance, autocorrelation, etc. The time series analysis function can help identify periodic variations, trends, and seasonal patterns in the data. For example, peak and valley periods, as well as periodic fluctuations, may be found by analyzing order volume data for a certain flow center. Feature normalization processing and feature set conversion are performed on a plurality of time-series feature values of each target data source. Feature normalization is to eliminate the influence of dimensions between different features so that feature values are comparable under the same standard. Feature set transformation is the transformation of these feature values into a format suitable for model analysis and prediction. For example, when processing transit time data for a logistics center, it may be desirable to convert the time data into a normalized fractional form in order to more efficiently use the data in the model.
In a specific embodiment, the process of executing step 104 may specifically include the following steps:
(1) Respectively carrying out feature coding on the logistics data retrieval feature sets to obtain coded data retrieval feature sets of each target data source;
(2) Performing linear matrix conversion on the coded data retrieval feature set of each target data source to obtain a coded data retrieval feature linear matrix of each target data source;
(3) Performing matrix operation on the coding data retrieval characteristic linear matrix of each target data source through a kernel function in a kernel principal component analysis algorithm to obtain an initial data retrieval characteristic kernel matrix of each target data source;
(4) Respectively carrying out centering treatment on the initial data retrieval feature kernel matrix of each target data source to obtain a target data retrieval feature kernel matrix of each target data source, and respectively carrying out feature value decomposition on the target data retrieval feature kernel matrix of each target data source to obtain a feature value and a feature vector corresponding to each target data source;
(5) Selecting a main component corresponding to each target data source according to the size of the characteristic value, and mapping the logistics data retrieval characteristic set to a new characteristic space according to the main component corresponding to each target data source to obtain a reduced-dimension data retrieval characteristic matrix corresponding to each target data source;
(6) And carrying out matrix fusion on the reduced-dimension data retrieval feature matrix corresponding to each target data source to generate a corresponding logistics data retrieval feature matrix.
Specifically, feature coding is performed on the feature sets of the logistics data retrieval respectively, and various features of the logistics data are converted into a format suitable for calculation and analysis. For example, the characteristics may include order volume, time of transportation, inventory level, etc. These features may be classified, numerical or time-sequential, and they need to be encoded into a uniform numerical format for subsequent data processing and analysis. And performing linear matrix conversion on the coded data retrieval feature set of each target data source. The encoded feature data is converted into a linear matrix form, which enables it to be adapted to more complex mathematical and statistical analyses. For example, order data, shipping time, and inventory levels are converted into a unified matrix format, where each row represents a data point and each column represents a feature. And performing matrix operation on the coding data retrieval characteristic linear matrix of each target data source through a kernel function in the kernel principal component analysis algorithm. Nuclear principal component analysis (KERNEL PCA) is a nonlinear dimension reduction technique that reveals complex structures and relationships in data by mapping the data in a high-dimensional space. The original linear feature space is converted into a higher dimensional feature space by the operation of a kernel function, thereby revealing the potential nonlinear relationship in the data. And respectively carrying out centering treatment on the initial data retrieval feature kernel matrix of each target data source. Centering refers to adjusting the data such that the mean value of each feature is zero, which helps to remove the bias of the data, ensuring a fair comparison between different features. For example, for data containing different volume level order volumes, the centralization process can ensure that orders of different sizes are comparable in analysis. Then, the feature value decomposition is performed on the target data retrieval feature kernel matrix of each target data source. The kernel matrix is decomposed into eigenvalues and eigenvectors, which represent the most important sources of variance in the data. For example, by eigenvalue decomposition, it can be identified which features play a major role in interpreting order quantity changes. And selecting a main component corresponding to each target data source according to the size of the characteristic value, mapping the logistics data retrieval characteristic set to a new characteristic space, and obtaining a dimension-reduction data retrieval characteristic matrix corresponding to each target data source, thereby simplifying the data set, reducing the computational complexity and simultaneously retaining the most important information as far as possible. And carrying out matrix fusion on the reduced-dimension data retrieval feature matrix corresponding to each target data source to generate a corresponding logistics data retrieval feature matrix. The fusion process not only integrates information from different data sources, but also ensures consistency and integrity of the data in analysis and model training.
In a specific embodiment, the process of executing step 105 may specifically include the following steps:
(1) Inputting the logistics data retrieval feature matrix into a preset variable structure dynamic Bayesian network model, wherein the variable structure dynamic Bayesian network model comprises: a K nearest neighbor algorithm layer, a nuclear attention mechanism layer and a three-layer Bayesian network;
(2) Performing feature cluster analysis on the feature matrix of the stream data retrieval by using a K nearest neighbor algorithm layer to generate a target data retrieval feature matrix;
(3) Carrying out attention mechanism weighted analysis on the target data retrieval feature matrix through a nuclear attention mechanism layer to obtain an attention data retrieval feature matrix;
(4) Inputting the attention data retrieval feature matrix into a preset three-layer Bayesian network, establishing state feature node connection between the attention data retrieval feature matrix and a first layer Bayesian network in the three-layer Bayesian network, establishing state compensation node connection between the attention data retrieval feature matrix and a second layer Bayesian network in the three-layer Bayesian network, and respectively establishing dynamic prediction node connection between the attention data retrieval feature matrix and a third layer Bayesian network in the three-layer Bayesian network;
(5) Forward deducing the attention data retrieval feature matrix through a three-layer Bayesian network, calculating probability distribution of each dynamic prediction node, each state compensation node and each state feature node, and outputting a target dynamic prediction result.
Specifically, the physical distribution data retrieval feature matrix is input into a preset variable structure dynamic Bayesian network model, wherein the model comprises a K nearest neighbor algorithm layer, a nuclear attention mechanism layer and a three-layer Bayesian network. And the K nearest neighbor algorithm layer performs feature cluster analysis on the feature matrix of the stream data retrieval, groups data points according to the similarity of features, and therefore identifies main modes and trends in the data. For example, data from a logistic data center may show that certain types of orders are often occurring together over a particular period of time, and that the K-nearest neighbor algorithm may cluster these orders together by analyzing their characteristics. And carrying out attention mechanism weighted analysis on the target data retrieval feature matrix through a core attention mechanism layer. The role of the kernel attention mechanism layer is to give more important features higher weight when analyzing the data, which helps the model focus on information that is more critical to the prediction task. For example, if a strong correlation is found between certain inventory level features and order volume, this layer may increase the weight of those features to ensure that the model can focus on these critical information when predicting. And inputting the data retrieval feature matrix into a preset three-layer Bayesian network. And establishing connection between the attention data retrieval feature matrix and each layer in the three-layer Bayesian network. Specifically, established between the first layer Bayesian network and the attention data retrieval feature matrix are state feature node connections, which help capture basic features and states of the data; the state compensation node connection is established between the second-layer Bayesian network and the attention data retrieval feature matrix, and can help correct or compensate deviation in the model; and the third layer Bayesian network is connected with the attention data retrieval feature matrix by dynamic prediction nodes for predicting future trend and mode. Forward inference is made on the attention data retrieval feature matrix through the three-layer bayesian network. And calculating probability distribution of each dynamic prediction node, each state compensation node and each state characteristic node through a network, and outputting a target dynamic prediction result based on the distribution. For example, by analyzing the time series characteristics of the logistics data, the network can predict future order volume fluctuations, inventory demand changes, or logistics delay risks.
In a specific embodiment, the process of executing step 106 may specifically include the following steps:
(1) According to the target dynamic prediction result, performing model optimization strategy analysis on the second data retrieval super-network model to obtain an initial model optimization strategy;
(2) Carrying out global scheme initialization on the target logistics data center according to an initial model optimization strategy by a preset genetic algorithm to generate a plurality of candidate model optimization strategies;
(3) Calculating the fitness value of each candidate model optimization strategy respectively, and carrying out strategy group division and optimization solution on a plurality of candidate model optimization strategies according to the fitness value to generate corresponding target model optimization strategies;
(4) And carrying out model self-adaptive updating on the second data retrieval super network model according to the target model optimization strategy to obtain the target data retrieval super network model.
Specifically, according to the target dynamic prediction result, model optimization strategy analysis is carried out on the second data retrieval super network model. For example, if the predictions show that certain types of order predictions are less accurate, this may indicate that the model requires finer feature extraction or more complex network structure in processing such order data. And (5) obtaining an initial model optimization strategy through analysis. And initializing a global scheme of the target object flow data center through a preset genetic algorithm to generate a plurality of candidate model optimization strategies. Genetic algorithm is a search algorithm imitating biological evolution process, and the solution of the problem is optimized by simulating mechanisms such as natural selection, genetics, mutation and the like. For example, a search may be performed through a genetic algorithm across multiple model structures and parameter configurations to find the best model optimization combination. In this process, each candidate model optimization strategy is considered an "individual" that will undergo selection, crossover and mutation to generate a new generation strategy with better performance. And calculating a fitness value of each candidate model optimization strategy, wherein the fitness value is an index for measuring the effect of each strategy in practical application. Each strategy's performance in the simulated environment is evaluated, for example, by comparing the prediction accuracy, response speed, and resource consumption of the different strategies to determine their fitness. According to the fitness value, the candidate strategies are divided into different groups, and optimization solution is carried out. The strategy with high fitness is selected for further interleaving and mutation to be expected to generate a more excellent model optimization strategy. And carrying out self-adaptive updating on the second data retrieval super network model according to the target model optimization strategy to obtain the target data retrieval super network model. And adjusting the structure, parameters or learning mode of the model according to the selected strategy. For example, if it is found that adding a network layer or changing an activation function can improve the predictive performance of the model, these adjustments will be applied to the existing model. Such adaptive updating ensures that the model adapts not only to the current data characteristics, but also predicts new trends and patterns that may occur in the future.
The method for intelligent retrieval of physical distribution information in the embodiment of the present application is described above, and the system for intelligent retrieval of physical distribution information in the embodiment of the present application is described below, referring to fig. 2, an embodiment of the system for intelligent retrieval of physical distribution information in the embodiment of the present application includes:
the analysis module 201 is configured to perform multi-level search node analysis on the target logistics data center, so as to obtain a data center search node, a plurality of upper layer data search nodes and a plurality of lower layer data search nodes;
The creation module 202 is configured to create a first data retrieval super-network model according to the data center retrieval node, the plurality of upper layer data retrieval nodes, and the plurality of lower layer data retrieval nodes, and perform network cluster analysis on the first data retrieval super-network model to obtain a second data retrieval super-network model;
the processing module 203 is configured to perform a logistic data retrieval feature analysis on the plurality of target data sources in the target logistic data center based on the second data retrieval super-network model, so as to obtain a logistic data retrieval feature set of each target data source;
the encoding module 204 is configured to perform feature encoding and matrix fusion on the feature set of the physical distribution data retrieval, so as to generate a corresponding physical distribution data retrieval feature matrix;
the prediction module 205 is configured to input the logistic data retrieval feature matrix into a preset variable structure dynamic bayesian network model to perform dynamic prediction on the data retrieval node, so as to obtain a target dynamic prediction result;
and the updating module 206 is configured to generate a corresponding target model optimization strategy according to the target dynamic prediction result, and perform model adaptive updating on the second data retrieval super network model according to the target model optimization strategy to obtain a target data retrieval super network model.
Through the cooperation of the components, the core nodes and the hierarchical relationship thereof of the data center can be accurately identified and analyzed through multi-level retrieval node analysis on the logistics data center. This in-depth structural analysis helps to better understand the flow and processing mechanisms of the logistics data, thereby providing a solid foundation for efficient data retrieval and information extraction. The first data retrieval super network model and the network cluster analysis are utilized to construct and optimize a second data retrieval super network model. The method can reflect complex logistics data relation, and find out key functional areas or information processing clusters of the data center through network cluster analysis, so that accuracy and efficiency of data retrieval are improved. By carrying out logistics data retrieval feature analysis on each target data source and carrying out feature coding and matrix fusion, key information in logistics data can be fully mined and utilized. The feature fusion not only improves the expression capability of the data, but also improves the robustness and accuracy of the retrieval method by integrating different types of features. The feature matrix is input into a preset variable structure dynamic Bayesian network model, so that dynamic prediction of the logistics data retrieval node can be realized. The prediction not only considers historical data, but also can adapt to environmental changes through a dynamic model, and provides more accurate and real-time prediction results. And generating a target model optimization strategy according to the dynamic prediction result, and carrying out self-adaptive updating on the data retrieval super-network model. The self-optimizing mechanism ensures that the system can continuously provide high-quality retrieval service along with the time and the environment change, and the application improves the retrieval accuracy and efficiency of the logistics information.
The application also provides a computer device, which comprises a memory and a processor, wherein the memory stores computer readable instructions, and the computer readable instructions, when executed by the processor, cause the processor to execute the steps of the intelligent logistics information retrieval method in the above embodiments.
The present application also provides a computer readable storage medium, which may be a non-volatile computer readable storage medium, or a volatile computer readable storage medium, where instructions are stored in the computer readable storage medium, when the instructions are executed on a computer, cause the computer to perform the steps of the logistic information wisdom retrieval method.
It will be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, systems and units may refer to the corresponding processes in the foregoing method embodiments, which are not repeated herein.
The integrated units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present application may be embodied essentially or in part or all of the technical solution or in part in the form of a software product stored in a storage medium, including instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a read-only memory (ROM), a random access memory (random access memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
The above embodiments are only for illustrating the technical solution of the present application, and not for limiting the same; although the application has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present application.
Claims (8)
1. The intelligent logistics information retrieval method is characterized by comprising the following steps of:
Performing multi-level retrieval node analysis on the target logistics data center to obtain a data center retrieval node, a plurality of upper layer data retrieval nodes and a plurality of lower layer data retrieval nodes; the method specifically comprises the following steps: carrying out multi-level retrieval node identification on the target logistics data center to obtain a data center retrieval node and a plurality of initial data retrieval nodes; respectively carrying out node degree centrality calculation on the plurality of initial data retrieval nodes to obtain node degree centrality of each initial data retrieval node, wherein the node degree centrality function is as follows: ，/> Represents the node degree centrality of the initial data retrieval node v, A degree indicating the initial data retrieval node v, n indicating the total number of nodes of the plurality of initial data retrieval nodes; and respectively carrying out proximity centrality calculation on the plurality of initial data retrieval nodes to obtain the proximity centrality of each initial data retrieval node, wherein the proximity centrality function is as follows: /(I)，/>Representing the proximity centrality of the initial data retrieval node v,/>Representing the shortest path length between the initial data retrieval node v and the initial data retrieval node u, n representing the total number of nodes of the plurality of initial data retrieval nodes; performing betweenness centrality calculation on the plurality of initial data retrieval nodes respectively to obtain betweenness centrality of each initial data retrieval node, wherein betweenness centrality functions are as follows:，/> representing the betweenness centrality of the initial data retrieval node v,/> Representing the number of all shortest paths between the initial data retrieval node s and the initial data retrieval node t,/>Representing the number of shortest paths between the initial data retrieval node s and the initial data retrieval node t passing through the initial data retrieval node v; performing node hierarchy division on the plurality of initial data retrieval nodes according to the node degree centrality, the proximity centrality and the betweenness centrality to obtain an initial hierarchical result of each initial data retrieval node; performing PageRank calculation on the plurality of initial data retrieval nodes according to the initial layering result to obtain a PageRank value of each initial data retrieval node, wherein the PageRank function is as follows: /(I)，/>PageRank value representing initial data retrieval node v, B v represents all node sets pointing to initial data retrieval node v,/>Representing the number of outer chains of the initial data retrieval node u,/>PageRank value representing initial data retrieval node u, d representing damping coefficient, set to 0.85; according to the PageRank value, node classification is carried out on the plurality of initial data retrieval nodes, so that a plurality of upper layer data retrieval nodes and a plurality of lower layer data retrieval nodes are obtained;
Creating a first data retrieval super-network model according to the data center retrieval node, the plurality of upper layer data retrieval nodes and the plurality of lower layer data retrieval nodes, and performing network cluster analysis on the first data retrieval super-network model to obtain a second data retrieval super-network model; the method specifically comprises the following steps: performing relationship analysis on the data center retrieval node and the plurality of upper layer data retrieval nodes through a preset graph neural network algorithm to obtain a plurality of first retrieval node relationships, and performing relationship analysis on the data center retrieval node and the plurality of lower layer data retrieval nodes through the graph neural network algorithm to obtain a plurality of second retrieval node relationships; creating a plurality of first relationship edges between the data center retrieval node and the plurality of upper layer data retrieval nodes according to the plurality of first retrieval node relationships, and creating a plurality of second relationship edges between the data center retrieval node and the plurality of lower layer data retrieval nodes according to the plurality of second retrieval node relationships; performing super-network model connection on the plurality of first relation edges and the plurality of second relation edges based on preset initial node edge weights, and generating a corresponding first data retrieval super-network model; performing network cluster analysis on the first data retrieval super-network model to generate a network cluster analysis result, and generating a corresponding target node edge weight according to the network cluster analysis result; performing super-network model weight optimization on the first data retrieval super-network model based on the target node edge weight to obtain a second data retrieval super-network model;
Based on the second data retrieval super-network model, carrying out logistics data retrieval feature analysis on a plurality of target data sources in the target logistics data center respectively to obtain a logistics data retrieval feature set of each target data source;
respectively carrying out feature coding and matrix fusion on the logistics data retrieval feature set to generate a corresponding logistics data retrieval feature matrix;
Inputting the logistics data retrieval feature matrix into a preset variable structure dynamic Bayesian network model to conduct data retrieval node dynamic prediction, and obtaining a target dynamic prediction result;
and generating a corresponding target model optimization strategy according to the target dynamic prediction result, and carrying out model self-adaptive updating on the second data retrieval super network model according to the target model optimization strategy to obtain a target data retrieval super network model.
2. The method for intelligently retrieving logistics information according to claim 1, wherein the performing the feature analysis of the logistics data retrieval on the plurality of target data sources in the target logistics data center based on the second data retrieval super-network model to obtain the feature set of the logistics data retrieval of each target data source comprises:
Based on the second data retrieval super network model, respectively carrying out logistics information retrieval monitoring on a plurality of target data sources in the target logistics data center to obtain initial logistics information retrieval log data of each target data source;
Respectively carrying out data preprocessing on the initial logistics information retrieval log data of each target data source to obtain target logistics information retrieval log data of each target data source;
Performing time sequence analysis on the target logistics information retrieval log data of each target data source respectively to obtain a plurality of time sequence characteristic values of each target data source, wherein the time sequence analysis function is as follows: ，/> time sequence characteristic value of target logistics information retrieval log data at lag k is represented, and is/are Value of target logistics information retrieval log data at time t,/>Represents the average value of the target logistics information retrieval log data, t represents the moment, n represents the total length of the target logistics information retrieval log data, and/>A value representing the target logistics information retrieval log data at time t+k;
And carrying out feature standardization processing and feature set conversion on a plurality of time sequence feature values of each target data source to obtain a logistics data retrieval feature set of each target data source.
3. The method for intelligently retrieving logistics information according to claim 1, wherein the feature encoding and matrix fusion are performed on the feature sets of the logistics data retrieval respectively to generate corresponding feature matrices of the logistics data retrieval, and the method comprises the following steps:
respectively carrying out feature coding on the logistics data retrieval feature sets to obtain coded data retrieval feature sets of each target data source;
Performing linear matrix conversion on the coded data retrieval feature set of each target data source to obtain a coded data retrieval feature linear matrix of each target data source;
Performing matrix operation on the coding data retrieval characteristic linear matrix of each target data source through a kernel function in a kernel principal component analysis algorithm to obtain an initial data retrieval characteristic kernel matrix of each target data source;
Respectively carrying out centering treatment on the initial data retrieval feature kernel matrix of each target data source to obtain a target data retrieval feature kernel matrix of each target data source, and respectively carrying out feature value decomposition on the target data retrieval feature kernel matrix of each target data source to obtain a feature value and a feature vector corresponding to each target data source;
selecting a main component corresponding to each target data source according to the size of the characteristic value, and mapping the logistics data retrieval characteristic set to a new characteristic space according to the main component corresponding to each target data source to obtain a dimensionality reduction data retrieval characteristic matrix corresponding to each target data source;
And carrying out matrix fusion on the reduced-dimension data retrieval feature matrix corresponding to each target data source to generate a corresponding logistics data retrieval feature matrix.
4. The method for intelligently retrieving logistics information according to claim 1, wherein inputting the feature matrix of the logistics data retrieval into a preset variable structure dynamic bayesian network model for dynamic prediction of data retrieval nodes to obtain a target dynamic prediction result comprises:
inputting the logistics data retrieval feature matrix into a preset variable structure dynamic Bayesian network model, wherein the variable structure dynamic Bayesian network model comprises: a K nearest neighbor algorithm layer, a nuclear attention mechanism layer and a three-layer Bayesian network;
performing feature cluster analysis on the logistics data retrieval feature matrix through the K nearest neighbor algorithm layer to generate a target data retrieval feature matrix;
performing attention mechanism weighted analysis on the target data retrieval feature matrix through the kernel attention mechanism layer to obtain an attention data retrieval feature matrix;
Inputting the attention data retrieval feature matrix into a preset three-layer Bayesian network, establishing state feature node connection between the attention data retrieval feature matrix and a first-layer Bayesian network in the three-layer Bayesian network, establishing state compensation node connection between the attention data retrieval feature matrix and a second-layer Bayesian network in the three-layer Bayesian network, and respectively establishing dynamic prediction node connection between the attention data retrieval feature matrix and a third-layer Bayesian network in the three-layer Bayesian network;
and forward deducing the attention data retrieval feature matrix through the three-layer Bayesian network, calculating probability distribution of each dynamic prediction node, each state compensation node and each state feature node, and outputting a target dynamic prediction result.
5. The method for intelligently retrieving logistics information according to claim 1, wherein the generating a corresponding target model optimization strategy according to the target dynamic prediction result, and performing model adaptive update on the second data retrieval super network model according to the target model optimization strategy, to obtain a target data retrieval super network model, comprises:
according to the target dynamic prediction result, performing model optimization strategy analysis on the second data retrieval super-network model to obtain an initial model optimization strategy;
Carrying out global scheme initialization on the target logistics data center according to the initial model optimization strategy through a preset genetic algorithm to generate a plurality of candidate model optimization strategies;
calculating the fitness value of each candidate model optimization strategy respectively, and carrying out strategy group division and optimization solution on the plurality of candidate model optimization strategies according to the fitness value to generate corresponding target model optimization strategies;
and carrying out model self-adaptive updating on the second data retrieval super network model according to the target model optimization strategy to obtain a target data retrieval super network model.
6. The utility model provides a commodity circulation information wisdom retrieval system which characterized in that, commodity circulation information wisdom retrieval system includes:
The analysis module is used for carrying out multi-level retrieval node analysis on the target logistics data center to obtain a data center retrieval node, a plurality of upper layer data retrieval nodes and a plurality of lower layer data retrieval nodes; the method specifically comprises the following steps: carrying out multi-level retrieval node identification on the target logistics data center to obtain a data center retrieval node and a plurality of initial data retrieval nodes; respectively carrying out node degree centrality calculation on the plurality of initial data retrieval nodes to obtain node degree centrality of each initial data retrieval node, wherein the node degree centrality function is as follows: ，/> Node degree centrality representing initial data retrieval node v,/> A degree indicating the initial data retrieval node v, n indicating the total number of nodes of the plurality of initial data retrieval nodes; and respectively carrying out proximity centrality calculation on the plurality of initial data retrieval nodes to obtain the proximity centrality of each initial data retrieval node, wherein the proximity centrality function is as follows: /(I)，/>Representing the proximity centrality of the initial data retrieval node v,/>Representing the shortest path length between the initial data retrieval node v and the initial data retrieval node u, n representing the total number of nodes of the plurality of initial data retrieval nodes; performing betweenness centrality calculation on the plurality of initial data retrieval nodes respectively to obtain betweenness centrality of each initial data retrieval node, wherein betweenness centrality functions are as follows:，/> representing the betweenness centrality of the initial data retrieval node v,/> Representing the number of all shortest paths between the initial data retrieval node s and the initial data retrieval node t,/>Representing the number of shortest paths between the initial data retrieval node s and the initial data retrieval node t passing through the initial data retrieval node v; performing node hierarchy division on the plurality of initial data retrieval nodes according to the node degree centrality, the proximity centrality and the betweenness centrality to obtain an initial hierarchical result of each initial data retrieval node; performing PageRank calculation on the plurality of initial data retrieval nodes according to the initial layering result to obtain a PageRank value of each initial data retrieval node, wherein the PageRank function is as follows: /(I)，/>PageRank value representing initial data retrieval node v, B v represents all node sets pointing to initial data retrieval node v,/>Representing the number of outer chains of the initial data retrieval node u,/>PageRank value representing initial data retrieval node u, d representing damping coefficient, set to 0.85; according to the PageRank value, node classification is carried out on the plurality of initial data retrieval nodes, so that a plurality of upper layer data retrieval nodes and a plurality of lower layer data retrieval nodes are obtained;
The creation module is used for creating a first data retrieval super-network model according to the data center retrieval node, the plurality of upper layer data retrieval nodes and the plurality of lower layer data retrieval nodes, and performing network cluster analysis on the first data retrieval super-network model to obtain a second data retrieval super-network model; the method specifically comprises the following steps: performing relationship analysis on the data center retrieval node and the plurality of upper layer data retrieval nodes through a preset graph neural network algorithm to obtain a plurality of first retrieval node relationships, and performing relationship analysis on the data center retrieval node and the plurality of lower layer data retrieval nodes through the graph neural network algorithm to obtain a plurality of second retrieval node relationships; creating a plurality of first relationship edges between the data center retrieval node and the plurality of upper layer data retrieval nodes according to the plurality of first retrieval node relationships, and creating a plurality of second relationship edges between the data center retrieval node and the plurality of lower layer data retrieval nodes according to the plurality of second retrieval node relationships; performing super-network model connection on the plurality of first relation edges and the plurality of second relation edges based on preset initial node edge weights, and generating a corresponding first data retrieval super-network model; performing network cluster analysis on the first data retrieval super-network model to generate a network cluster analysis result, and generating a corresponding target node edge weight according to the network cluster analysis result; performing super-network model weight optimization on the first data retrieval super-network model based on the target node edge weight to obtain a second data retrieval super-network model;
the processing module is used for respectively carrying out logistics data retrieval characteristic analysis on a plurality of target data sources in the target logistics data center based on the second data retrieval super-network model to obtain a logistics data retrieval characteristic set of each target data source;
The coding module is used for respectively carrying out feature coding and matrix fusion on the logistics data retrieval feature set to generate a corresponding logistics data retrieval feature matrix;
The prediction module is used for inputting the logistics data retrieval feature matrix into a preset variable structure dynamic Bayesian network model to conduct dynamic prediction of the data retrieval nodes, so as to obtain a target dynamic prediction result;
and the updating module is used for generating a corresponding target model optimization strategy according to the target dynamic prediction result, and carrying out model self-adaptive updating on the second data retrieval super network model according to the target model optimization strategy to obtain a target data retrieval super network model.
7. A computer device, the computer device comprising: a memory and at least one processor, the memory having instructions stored therein;
the at least one processor invokes the instructions in the memory to cause the computer device to perform the logistic information wisdom retrieval method of any one of claims 1-5.
8. A computer readable storage medium having instructions stored thereon, wherein the instructions when executed by a processor implement the logistic information wisdom retrieval method of any one of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410236355.4A CN117807114B (en) | 2024-03-01 | 2024-03-01 | Logistics information intelligent retrieval method, system, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410236355.4A CN117807114B (en) | 2024-03-01 | 2024-03-01 | Logistics information intelligent retrieval method, system, equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117807114A CN117807114A (en) | 2024-04-02 |
| CN117807114B true CN117807114B (en) | 2024-05-07 |
Family
ID=90434970
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410236355.4A Active CN117807114B (en) | 2024-03-01 | 2024-03-01 | Logistics information intelligent retrieval method, system, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117807114B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113761124A (en) * | 2021-05-25 | 2021-12-07 | 腾讯科技（深圳）有限公司 | Training method of text coding model, information retrieval method and equipment |
| CN114880360A (en) * | 2022-07-06 | 2022-08-09 | 杭州费尔斯通科技有限公司 | Data retrieval method and device based on Bayesian optimization |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8037043B2 (en) * | 2008-09-09 | 2011-10-11 | Microsoft Corporation | Information retrieval system |
-
2024
- 2024-03-01 CN CN202410236355.4A patent/CN117807114B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113761124A (en) * | 2021-05-25 | 2021-12-07 | 腾讯科技（深圳）有限公司 | Training method of text coding model, information retrieval method and equipment |
| CN114880360A (en) * | 2022-07-06 | 2022-08-09 | 杭州费尔斯通科技有限公司 | Data retrieval method and device based on Bayesian optimization |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117807114A (en) | 2024-04-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Soui et al. | Rule-based credit risk assessment model using multi-objective evolutionary algorithms | |
| Wang et al. | Rule induction for forecasting method selection: Meta-learning the characteristics of univariate time series | |
| Chang et al. | Trend discovery in financial time series data using a case based fuzzy decision tree | |
| CN111324642A (en) | Model algorithm type selection and evaluation method for power grid big data analysis | |
| Raju et al. | An approach for demand forecasting in steel industries using ensemble learning | |
| Mustafa et al. | A new fuzzy grach model to forecast stock market technical analysis | |
| CN117807114B (en) | Logistics information intelligent retrieval method, system, equipment and storage medium | |
| Riesener et al. | Identification of evaluation criteria for algorithms used within the context of product development | |
| Ardyanta et al. | A prediction of stock price movements using support vector machines in Indonesia | |
| Mehlstäubl et al. | Data mining in product portfolio and variety management–literature review on use cases and research potentials | |
| Agusta et al. | Technical analysis based automatic trading prediction system for stock exchange using support vector machine | |
| Huang et al. | An enhanced approach to optimizing the stock portfolio selection based on Modified Markowitz MV Method | |
| Yoon et al. | Methodology for Improving the Performance of Demand Forecasting Through Machine Learning | |
| CN117670066B (en) | Questor management method, system, equipment and storage medium based on intelligent decision | |
| CN117495109B (en) | Power stealing user identification system based on neural network | |
| Alghieth | Review of an ANFIS Methodology-Based Stock Market Prediction System | |
| CN117539520B (en) | Firmware self-adaptive upgrading method, system and equipment | |
| US11769100B2 (en) | Leading indicators and monitor business KPIs and metrics for preemptive action | |
| Huang et al. | Application of enhanced cluster validity index function to automatic stock portfolio selection system | |
| CN117787512A (en) | Enterprise judicial management method, system, equipment and storage medium | |
| Goldani | Evaluating Feature Selection Methods for Macro-Economic Forecasting, Applied for Inflation Indicator of Iran | |
| US20210319259A1 (en) | Method and apparatus for extracting a pattern of time series data | |
| Lyvia | Application of Data Mining to Predict Stock Price Movements in MNC Bank Using K-Nears Neighbor | |
| Park et al. | 11‐1: AI‐based Interaction Commonality Analysis | |
| Kavitha et al. | GA Based Stochastic Optimization For Stock Price Forecasting using Fuzzy Time series Hidden Markov Model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |