CN116501876B - Big data tracking method and AI system for cloud collaborative digital service - Google Patents

Big data tracking method and AI system for cloud collaborative digital service Download PDF

Info

Publication number
CN116501876B
CN116501876B CN202310500410.1A CN202310500410A CN116501876B CN 116501876 B CN116501876 B CN 116501876B CN 202310500410 A CN202310500410 A CN 202310500410A CN 116501876 B CN116501876 B CN 116501876B
Authority
CN
China
Prior art keywords
group
graph
knowledge
network
knowledge graph
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
Application number
CN202310500410.1A
Other languages
Chinese (zh)
Other versions
CN116501876A (en
Inventor
王刚
李琼
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Gtcom Technology Shaanxi Co ltd
Original Assignee
Gtcom Technology Shaanxi Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Gtcom Technology Shaanxi Co ltd filed Critical Gtcom Technology Shaanxi Co ltd
Priority to CN202310500410.1A priority Critical patent/CN116501876B/en
Publication of CN116501876A publication Critical patent/CN116501876A/en
Application granted granted Critical
Publication of CN116501876B publication Critical patent/CN116501876B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/35—Clustering; Classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367—Ontology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/04—Inference or reasoning models
    • G06N5/041—Abduction
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/50—Network services
    • H04L67/535—Tracking the activity of the user
    • 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

Abstract

The invention provides a big data tracking method and an AI system for cloud collaborative digital services, and relates to the technical field of artificial intelligence. In the invention, the first group knowledge graph and the second group knowledge graph are respectively subjected to characteristic mining operation to form a group identification characterization vector of the first group knowledge graph and a group non-identification characterization vector of the second group knowledge graph; performing joint feature mining operation on the first group knowledge graph and the second group knowledge graph to output joint mining characterization vectors; vector aggregation operation is carried out on the group identification characterization vector, the group non-identification characterization vector and the joint mining characterization vector so as to form an aggregated group characterization vector; performing feature mining reverse operation on the aggregated group characterization vector to output a third group knowledge graph; and carrying out group anomaly analysis operation on the object group to be tracked based on the third group knowledge graph. Based on the above, the reliability of anomaly tracking can be improved.

Description

Big data tracking method and AI system for cloud collaborative digital service
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a big data tracking method and an AI system for cloud collaborative digital services.
Background
The cloud cooperation provides the digital service, so that the service providing efficiency can be improved to a certain extent. In the process of providing the digital service, a large amount of behavior data (network behavior data) can be formed for the user (service object), and by analyzing the behavior data, valuable information can be obtained, for example, by analyzing, tracking of behavior abnormality can be realized, however, in the prior art, there is a problem of low reliability.
Disclosure of Invention
In view of the above, the present invention aims to provide a big data tracking method and an AI system for cloud collaborative digital services, so as to improve reliability of anomaly tracking.
In order to achieve the above purpose, the embodiment of the present invention adopts the following technical scheme:
a big data tracking method for cloud collaborative digital services, the big data tracking method comprising:
extracting a first group knowledge graph and a second group knowledge graph, wherein the first group knowledge graph is formed by performing mapping processing on current behavior data of a plurality of objects to be tracked, which are included in a group of objects to be tracked, and the second group knowledge graph is formed by performing mapping processing on historical behavior data of a plurality of objects to be tracked, which are included in the group of objects to be tracked, and the current behavior data and the historical behavior data are in a data form including text data;
Performing feature mining operation on the first group knowledge graph and the second group knowledge graph respectively to form a group identification characterization vector of the first group knowledge graph and a group non-identification characterization vector of the second group knowledge graph, wherein the group identification characterization vector is used for characterizing group-specific information of the group of the objects to be tracked, and the group non-identification characterization vector is used for characterizing other feature information except for the group-specific information of the group of the objects to be tracked;
performing joint feature mining operation on the first group knowledge graph and the second group knowledge graph to output joint mining characterization vectors corresponding to the first group knowledge graph and the second group knowledge graph in the whole;
vector aggregation operation is carried out on the group identification characterization vector, the group non-identification characterization vector and the joint mining characterization vector so as to form a corresponding aggregation group characterization vector;
performing feature mining reverse operation on the aggregate group representation vectors to output corresponding third group knowledge patterns, wherein a correlation is formed between the third group knowledge patterns and group identification representation vectors of the first group knowledge patterns, and a correlation is formed between the third group knowledge patterns and group non-identification representation vectors of the second group knowledge patterns;
And carrying out group abnormality analysis operation on the object group to be tracked based on the third group knowledge graph to obtain a group abnormality analysis result corresponding to the object group to be tracked so as to realize behavior abnormality tracking of the object group to be tracked.
In some preferred embodiments, in the above big data tracking method for cloud collaborative digitization service, the step of extracting the first group knowledge graph and the second group knowledge graph includes:
constructing a first object knowledge graph and a second object knowledge graph, wherein the first object knowledge graph is formed by performing mapping processing on current behavior data of a plurality of management objects, the second object knowledge graph is formed by performing mapping processing on historical behavior data of the plurality of management objects, each graph member in the first object knowledge graph corresponds to one management object in the plurality of management objects, attribute data of the graph member at least comprises corresponding current behavior data, each graph member in the second object knowledge graph corresponds to one management object in the plurality of management objects, and attribute data of the graph member at least comprises corresponding historical behavior data;
Respectively carrying out group determination operation on the first object knowledge graph and the second object knowledge graph to determine group parts in the first object knowledge graph and the second object knowledge graph;
extracting a corresponding first group knowledge graph based on the group part determined in the first object knowledge graph, and extracting a corresponding second group knowledge graph based on the group part determined in the second object knowledge graph;
the big data tracking method for the cloud collaborative digital service further comprises the following steps:
based on the third group knowledge graph, carrying out replacement operation on the group part in the second object knowledge graph to form a corresponding third object knowledge graph, wherein a correlation is formed between the third object knowledge graph and a group identification characterization vector of the group part in the first object knowledge graph, and a correlation is formed between the third object knowledge graph and a group non-identification characterization vector of the group part in the second object knowledge graph;
and performing object anomaly analysis operation on the plurality of management objects based on the third object knowledge graph to obtain object anomaly analysis results corresponding to the plurality of management objects so as to track behavior anomalies of the plurality of management objects.
In some preferred embodiments, in the above big data tracking method for cloud collaborative digitization service, the step of performing feature mining operation on the first population knowledge graph and the second population knowledge graph to form a population identification token vector of the first population knowledge graph and a population non-identification token vector of the second population knowledge graph includes:
performing feature mining operation on the first group knowledge graph by utilizing a feature mining first network to form a group identification characterization vector of the first group knowledge graph, wherein the feature mining first network performs network optimization operation on the basis of first type of exemplary data to form the group identification characterization vector;
and performing feature mining operation on the second group knowledge graph by using a feature mining second network which is different from the feature mining first network to form a group non-identification characterization vector of the second group knowledge graph, wherein the feature mining second network is formed by performing network optimization operation on the basis of second-class exemplary data and third-class exemplary data rotation, the second-class exemplary data is not configured with corresponding data tags, and the third-class exemplary data is configured with corresponding data tags.
In some preferred embodiments, in the above big data tracking method for cloud collaborative digitization service, the step of using features to mine a first network and performing feature mining operation on the first group knowledge graph to form a group identifier characterization vector of the first group knowledge graph includes:
loading first map data distribution of the first group knowledge maps to load the first map data distribution into a feature mining first network; and utilizing the characteristics to mine a first network, and performing characteristic mining operation on the first map data distribution to form a group identification characterization vector of the first group knowledge map;
the step of performing feature mining operation on the second group knowledge graph by using features different from the feature mining first network to form a group non-identification characterization vector of the second group knowledge graph, includes:
loading a second graph data distribution of the second group of knowledge graphs to a second feature mining network different from the first feature mining network; and utilizing the characteristics to excavate a second network, performing characteristic excavation operation on the second spectrum data distribution to form a group non-identification characterization vector of the second group knowledge spectrum, wherein network parameters of the characteristic excavation first network and the characteristic excavation second network are different.
In some preferred embodiments, in the above big data tracking method for cloud collaborative digitization service, the step of performing a joint feature mining operation on the first population knowledge graph and the second population knowledge graph to output joint mining characterization vectors corresponding to the first population knowledge graph and the second population knowledge graph as a whole includes:
and performing joint feature mining operation on the first group knowledge graph and the second group knowledge graph by using a joint feature mining network to output joint mining characterization vectors corresponding to the first group knowledge graph and the second group knowledge graph on the whole, and performing network optimization operation on the second type of exemplary data and the third type of exemplary data based on rotation to form the joint feature mining network and the feature mining second network.
In some preferred embodiments, in the foregoing big data tracking method for cloud collaborative digitization service, the step of performing a vector aggregation operation on the group identification token vector, the group non-identification token vector and the joint mining token vector to form a corresponding aggregated group token vector includes:
Performing vector aggregation operation on the group identification characterization vector, the group non-identification characterization vector and the joint mining characterization vector by using a vector aggregation network to form a corresponding aggregated group characterization vector;
the step of performing feature mining reverse operation on the aggregated group characterization vector to output a corresponding third group knowledge graph includes:
and performing feature mining reverse operation on the aggregated group characterization vector by using a feature mining reverse processing network to output a corresponding third group knowledge graph, wherein the vector aggregation network, the feature mining reverse processing network, the joint feature mining network and the feature mining second network are formed by performing network optimization operation on the second type of exemplary data and the third type of exemplary data based on rotation.
In some preferred embodiments, in the foregoing big data tracking method for cloud collaborative digitization service, the step of performing a vector aggregation operation on the group identification token vector, the group non-identification token vector, and the joint mining token vector by using a vector aggregation network to form a corresponding aggregated group token vector includes:
Performing cascade combination operation on the group identification characterization vector, the group non-identification characterization vector and the joint mining characterization vector by using a front-end processing unit of a vector aggregation network so as to output a corresponding cascade combination characterization vector;
and performing deep filtering operation on the cascade combination characterization vector by utilizing a back-end processing unit of the vector aggregation network to form a corresponding aggregation group characterization vector.
In some preferred embodiments, in the above big data tracking method for cloud collaborative digitization service, the feature mining first network, the feature mining second network, the joint feature mining network, the vector aggregation network, and the feature mining inverse processing network are included in a knowledge graph reconstruction model, and the step of optimizing the knowledge graph reconstruction model includes:
extracting second-class exemplary data and third-class exemplary data, wherein the third-class exemplary data comprises an exemplary first group knowledge-graph, an exemplary second group knowledge-graph and a reconstructed comparison knowledge-graph, and the second-class exemplary data comprises an exemplary third group knowledge-graph and an exemplary fourth group knowledge-graph;
Based on the third type of exemplary data, comparing and optimizing the knowledge graph reconstruction model to optimize and update network parameters of the feature mining second network, the joint feature mining network, the vector aggregation network and the feature mining reverse processing network;
based on the second class of exemplary data, carrying out self-optimization operation on the knowledge graph reconstruction model so as to carry out optimization updating on network parameters of the feature mining second network, the joint feature mining network, the vector aggregation network and the feature mining reverse processing network;
and carrying out the self-optimizing operation and the contrast optimizing operation in a rotating way to form the network optimization of the knowledge graph reconstruction model.
In some preferred embodiments, in the foregoing big data tracking method for cloud collaborative digitization service, the step of performing a contrast optimization operation on the knowledge-graph reconstruction model based on the third type of exemplary data to perform an optimization update on network parameters of the feature mining second network, the joint feature mining network, the vector aggregation network, and the feature mining inverse processing network includes:
Extracting a knowledge graph resolution model; and reconstructing an exemplary first reconstructed population knowledge graph corresponding to the exemplary first population knowledge graph and the exemplary second population knowledge graph by using the knowledge graph reconstruction model; and, marking at least one of the exemplary first population knowledge-graph, the exemplary second population knowledge-graph, and the reconstructed comparison knowledge-graph to mark as relevant exemplary data of the knowledge-graph resolution model, and marking the exemplary first reconstructed population knowledge-graph to mark as irrelevant exemplary data of the knowledge-graph resolution model; according to the related exemplary data and the non-related exemplary data, a comparison learning cost index for optimizing the knowledge-graph resolution model and the knowledge-graph reconstruction model is calculated, and the comparison learning cost index has a correlation with the difference between the first exemplary reconstruction group knowledge-graph and the reconstruction comparison knowledge-graph; based on the comparison learning cost index, optimizing and updating network parameters of the feature mining second network, the joint feature mining network, the vector aggregation network and the feature mining reverse processing network;
And performing self-optimization operation on the knowledge graph reconstruction model based on the second class of exemplary data to perform optimization updating on network parameters of the feature mining second network, the joint feature mining network, the vector aggregation network and the feature mining reverse processing network, wherein the method comprises the following steps:
reconstructing an exemplary second reconstructed group knowledge graph corresponding to the exemplary third group knowledge graph and the exemplary fourth group knowledge graph by using the knowledge graph reconstruction model; and, marking at least one of the exemplary third population knowledge-graph and the exemplary fourth population knowledge-graph to mark as relevant exemplary data of the knowledge-graph resolution model, and marking the exemplary second reconstructed population knowledge-graph to mark as irrelevant exemplary data of the knowledge-graph resolution model; and according to the related exemplary data and the non-related exemplary data, calculating self-learning cost indexes for optimizing the knowledge-graph resolution model and the knowledge-graph reconstruction model, wherein the self-learning cost indexes have a related relationship with the differences between the second exemplary reconstructed group knowledge graph and the fourth exemplary group knowledge graph; and optimizing and updating network parameters of the feature mining second network, the joint feature mining network, the vector aggregation network and the feature mining reverse processing network based on the self-learning cost index.
The embodiment of the invention also provides an AI system, which comprises a processor and a memory, wherein the memory is used for storing a computer program, and the processor is used for executing the computer program so as to realize the big data tracking method.
The big data tracking method and the AI system for cloud collaborative digital services provided by the embodiment of the invention can firstly extract the first group knowledge graph and the second group knowledge graph; respectively performing feature mining operation on the first group knowledge graph and the second group knowledge graph to form a group identification characterization vector of the first group knowledge graph and a group non-identification characterization vector of the second group knowledge graph; performing joint feature mining operation on the first group knowledge graph and the second group knowledge graph to output joint mining characterization vectors corresponding to the first group knowledge graph and the second group knowledge graph on the whole; vector aggregation operation is carried out on the group identification characterization vector, the group non-identification characterization vector and the joint mining characterization vector so as to form a corresponding aggregation group characterization vector; performing feature mining reverse operation on the aggregated group characterization vector to output a corresponding third group knowledge graph; and carrying out group abnormality analysis operation on the object group to be tracked based on the third group knowledge graph so as to obtain a group abnormality analysis result corresponding to the object group to be tracked. Based on the above, the first group knowledge graph is not directly analyzed, but the second group knowledge graph is combined to reconstruct the third group knowledge graph, and the third group knowledge graph has a correlation with the group identification characterization vector of the first group knowledge graph and a correlation with the group non-identification characterization vector of the second group knowledge graph, so that the current behavior data and the historical behavior data can be fully combined, and therefore, the reliability of anomaly tracking can be improved to a certain extent by carrying out group anomaly analysis operation on the group of objects to be tracked according to the third group knowledge graph, and the joint feature mining operation can be carried out in the reconstruction process, so that the reliability of the mined information is higher, the reliability of anomaly tracking can be further improved, and the problem of low reliability in the prior art is solved.
In order to make the above objects, features and advantages of the present invention more comprehensible, preferred embodiments accompanied with figures are described in detail below.
Drawings
Fig. 1 is a block diagram of an AI system according to an embodiment of the present invention.
Fig. 2 is a flowchart illustrating steps included in the big data tracking method for cloud collaborative digital services according to an embodiment of the present invention.
Fig. 3 is a schematic diagram of each module included in the big data tracking device for cloud collaborative digital services according to an embodiment of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, but not all embodiments of the present invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
As shown in fig. 1, an embodiment of the present invention provides an AI system. Wherein the AI system may include a memory and a processor.
In detail, the memory and the processor are electrically connected directly or indirectly to realize transmission or interaction of data. For example, electrical connection may be made to each other via one or more communication buses or signal lines. The memory may store at least one software functional module (computer program) that may exist in the form of software or firmware. The processor may be configured to execute the executable computer program stored in the memory, so as to implement the big data tracking method for cloud collaborative digital service provided by the embodiment of the present invention.
It should be appreciated that in some embodiments, the Memory may be, but is not limited to, random access Memory (Random Access Memory, RAM), read Only Memory (ROM), programmable Read Only Memory (Programmable Read-Only Memory, PROM), erasable Read Only Memory (Erasable Programmable Read-Only Memory, EPROM), electrically erasable Read Only Memory (Electric Erasable Programmable Read-Only Memory, EEPROM), and the like.
It should be appreciated that in some embodiments, the processor may be a general purpose processor, including a central processing unit (Central Processing Unit, CPU), a network processor (Network Processor, NP), a System on Chip (SoC), etc.; but also Digital Signal Processors (DSPs), application Specific Integrated Circuits (ASICs), field Programmable Gate Arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
It should be appreciated that in some embodiments, the AI system may be a server having data processing capabilities.
With reference to fig. 2, the embodiment of the invention further provides a big data tracking method for cloud collaborative digital services, which can be applied to the AI system. The method steps defined by the flow related to the big data tracking method for cloud collaborative digital service can be realized by the AI system.
The specific flow shown in fig. 2 will be described in detail.
Step S110, extracting a first group knowledge graph and a second group knowledge graph.
In the embodiment of the invention, the AI system can extract a first group knowledge graph and a second group knowledge graph. The first group knowledge graph is formed by performing mapping processing on current behavior data of a plurality of objects to be tracked (which can be users receiving services) included in an object group to be tracked, and the second group knowledge graph is formed by performing mapping processing on historical behavior data of a plurality of objects to be tracked included in the object group to be tracked. The data form of the current behavior data and the historical behavior data comprises text data, namely the current behavior and the historical behavior of the object to be tracked are described in the text form.
And step S120, performing feature mining operation on the first group knowledge graph and the second group knowledge graph respectively to form a group identification characterization vector of the first group knowledge graph and a group non-identification characterization vector of the second group knowledge graph.
In the embodiment of the invention, the AI system may perform feature mining operations on the first population knowledge graph and the second population knowledge graph, respectively, so as to form a population identification characterization vector of the first population knowledge graph and a population non-identification characterization vector of the second population knowledge graph. The group identification characterization vector is used for characterizing group specific information of the group of the objects to be tracked, namely information enabling the plurality of objects to be tracked to form one group, or can be understood as identity information of the group and can be distinguished from other groups, and the group non-identification characterization vector is used for characterizing other characteristic information except the group specific information of the group of the objects to be tracked, namely information except the identity information of the group, such as object compactness or behavior relativity among the objects to be tracked in the group.
And step S130, performing joint feature mining operation on the first group knowledge graph and the second group knowledge graph to output joint mining characterization vectors corresponding to the first group knowledge graph and the second group knowledge graph in the whole.
In the embodiment of the invention, the AI system can perform joint feature mining operation on the first group knowledge graph and the second group knowledge graph to output joint mining characterization vectors corresponding to the first group knowledge graph and the second group knowledge graph in the whole.
Step S140, performing a vector aggregation operation on the group identification token vector, the group non-identification token vector and the joint mining token vector to form a corresponding aggregated group token vector.
In the embodiment of the invention, the AI system can perform vector aggregation operation on the group identification characterization vector, the group non-identification characterization vector and the joint mining characterization vector to form a corresponding aggregated group characterization vector.
And step S150, performing feature mining reverse operation on the aggregate group characterization vector so as to output a corresponding third group knowledge graph.
In the embodiment of the invention, the AI system can perform feature mining reverse operation on the aggregate group characterization vector so as to output a corresponding third group knowledge graph. The third group knowledge graph and the group identification characterization vector of the first group knowledge graph have a correlation relationship, and the third group knowledge graph and the group non-identification characterization vector of the second group knowledge graph have a correlation relationship.
Step S160, based on the third group knowledge graph, performing group abnormality analysis operation on the to-be-tracked object group to obtain a group abnormality analysis result corresponding to the to-be-tracked object group.
In the embodiment of the invention, the AI system can perform a group anomaly analysis operation on the object group to be tracked based on the third group knowledge graph to obtain a group anomaly analysis result corresponding to the object group to be tracked, so as to realize behavior anomaly tracking on the object group to be tracked. The group abnormality analysis result may be used to reflect whether or not the object group to be tracked has abnormal behavior, the degree of abnormality existing, or the like, or may be used to reflect the type of abnormality existing, or the like, and is not particularly limited. In addition, the group anomaly analysis operation can be realized through a corresponding neural network, the neural network can be formed through network optimization through an exemplary knowledge graph and corresponding actual group anomaly data, so that the neural network can learn the mapping relation between the knowledge graph and the group anomaly data, and therefore, the group anomaly analysis result corresponding to the third group knowledge graph can be determined based on the mapping relation, and the group anomaly analysis result corresponding to the group of the object to be tracked is obtained. And, the neural network generally includes a feature mining unit, an abnormality analysis unit (may include a softmax function), and the like.
Based on the above, the first group knowledge graph is not directly analyzed, but the second group knowledge graph is combined to reconstruct the third group knowledge graph, and the third group knowledge graph has a correlation with the group identification characterization vector of the first group knowledge graph and a correlation with the group non-identification characterization vector of the second group knowledge graph, so that the current behavior data and the historical behavior data can be fully combined, and therefore, the reliability of anomaly tracking can be improved to a certain extent by carrying out group anomaly analysis operation on the group of objects to be tracked according to the third group knowledge graph, and the joint feature mining operation can be carried out in the reconstruction process, so that the reliability of the mined information is higher, the reliability of anomaly tracking can be further improved, and the problem of low reliability in the prior art is solved.
It should be understood that, in some embodiments, the step S110 may further include the following specific embodiments:
constructing a first object knowledge graph and a second object knowledge graph, wherein the first object knowledge graph is formed by performing a graph processing on current behavior data (behavior data in a last period) of a plurality of management objects (users receiving service), the second object knowledge graph is formed by performing a graph processing on historical behavior data (behavior data in a historical period) of the plurality of management objects, each graph member corresponds to one management object in the plurality of management objects in the first object knowledge graph, the attribute data of each graph member at least comprises the corresponding current behavior data, each graph member corresponds to one management object in the plurality of management objects in the second object knowledge graph, the attribute data of each graph member at least comprises the corresponding historical behavior data, and for example, the weight of connecting wires and connecting wires between the graph members can be determined based on the correlation between the corresponding attribute data;
The first object knowledge graph and the second object knowledge graph are respectively subjected to group determination operation to determine group parts in the first object knowledge graph and the second object knowledge graph, for example, group recognition can be performed to determine corresponding group parts, the group recognition can be realized through corresponding neural networks, and each management object marked manually in advance can also be marked as an object to be tracked directly, so that the group parts formed by the object to be tracked are determined;
and extracting a corresponding first group knowledge graph (namely, a local knowledge graph of the first object knowledge graph) based on the group part determined in the first object knowledge graph, and extracting a corresponding second group knowledge graph (namely, a local knowledge graph of the second object knowledge graph) based on the group part determined in the second object knowledge graph.
It should be understood that, in some embodiments, the big data tracking method for cloud collaborative digital services may further include the following steps:
based on the third group knowledge graph, carrying out replacement operation on the group part in the second object knowledge graph to form a corresponding third object knowledge graph, wherein a correlation is formed between the third object knowledge graph and a group identification characterization vector of the group part in the first object knowledge graph, and a correlation is formed between the third object knowledge graph and a group non-identification characterization vector of the group part in the second object knowledge graph;
And performing object anomaly analysis operation (such as related description in step S160) on the plurality of management objects based on the third object knowledge graph to obtain object anomaly analysis results corresponding to the plurality of management objects, so as to track behavioral anomalies of the plurality of management objects, thereby comprehensively tracking the plurality of management objects.
It should be understood that, in some embodiments, the step S120 may further include the following specific embodiments:
performing feature mining operation on the first group knowledge graph by utilizing a feature mining first network to form a group identification characterization vector of the first group knowledge graph, wherein the feature mining first network is formed by performing network optimization operation on the basis of first type of exemplary data, and the feature mining operation at least can comprise convolution operation and the like;
the feature mining second network is used to form a group non-identification characterization vector of the second group knowledge graph, the feature mining second network is formed by performing network optimization operation based on second-class exemplary data and third-class exemplary data rotation, the second-class exemplary data is not configured with corresponding data tags, and the third-class exemplary data is configured with corresponding data tags, namely, by partially adopting the second-class exemplary data which is not configured with corresponding data tags, the problems of high cost and low efficiency caused by configuring corresponding data tags can be reduced to a certain extent.
It should be appreciated that, in some embodiments, the step of performing feature mining on the first group knowledge-graph to form a group identifier token vector of the first group knowledge-graph by using features to mine the first network may further include the following specific embodiments:
loading first map data distribution of the first group knowledge maps to load the first map data distribution into a feature mining first network; and utilizing the characteristics to mine the first network, performing characteristic mining operation on the first map data distribution to form a group identification characterization vector of the first group knowledge map, wherein in the first group knowledge map, attribute data of map members comprise corresponding current behavior data, and the number of the current behavior data corresponding to one map member can be multiple, so that the multiple current behavior data can be fused, such as the last current behavior data is screened out, and thus, the first map data distribution of the first group knowledge map can be constructed for the screened current behavior data corresponding to each map member.
It should be appreciated that, in some embodiments, the step of performing the feature mining operation on the second population knowledge-graph to form the population non-identification token vector of the second population knowledge-graph by using the feature different from the feature mining first network may further include the following specific embodiments:
Loading a second graph data distribution of the second group of knowledge graphs to a second feature mining network different from the first feature mining network; and performing feature mining operation on the second map data distribution by using the features to mine a second network, so as to form a group non-identification characterization vector of the second group knowledge map, wherein network parameters of the feature mining first network and the feature mining second network are different, so that the emphasis points of the feature mining operation of the feature mining first network and the feature mining second network are different, and in addition, the forming mode of the second map data distribution can be consistent with the forming mode of the first map data distribution.
It should be understood that, in some embodiments, the step S130 may further include the following specific embodiments:
and performing joint feature mining operation on the first group knowledge graph and the second group knowledge graph by using a joint feature mining network to output joint mining characterization vectors corresponding to the first group knowledge graph and the second group knowledge graph on the whole, and performing network optimization operation on the second type of exemplary data and the third type of exemplary data based on rotation to form the joint feature mining network and the feature mining second network. For example, the first group knowledge graph and the second group knowledge graph may be combined, for example, attribute data of respective graph members may be combined, and then, feature mining operations, such as performing convolution operation, may be performed on the combined group knowledge graph.
It should be understood that, in some embodiments, the step S140 may further include the following specific embodiments:
and carrying out vector aggregation operation on the group identification characterization vector, the group non-identification characterization vector and the joint mining characterization vector by using a vector aggregation network to form a corresponding aggregate group characterization vector, wherein the vector aggregation network can be formed by carrying out network optimization based on corresponding data.
It should be appreciated that, in some embodiments, the step of performing a vector aggregation operation on the group identification token vector, the group non-identification token vector, and the joint mining token vector to form a corresponding aggregated group token vector using a vector aggregation network may further include the following specific embodiments:
performing cascade combination operation on the group identification characterization vector, the group non-identification characterization vector and the joint mining characterization vector by using a front-end processing unit of a vector aggregation network so as to output a corresponding cascade combination characterization vector, such as { the group identification characterization vector, the group non-identification characterization vector and the joint mining characterization vector };
And performing deep filtering operation on the cascade combined representation vector by using a back-end processing unit of the vector aggregation network to form a corresponding aggregate group representation vector, wherein the deep filtering operation can be, for example, sequentially performing filtering operation through a plurality of cascaded filtering units or filtering matrixes to form the corresponding aggregate group representation vector.
It should be appreciated that, in some embodiments, the step S150 may further include the following specific embodiments:
and performing feature mining reverse operation on the aggregated group characterization vector by using a feature mining reverse processing network to output a corresponding third group knowledge graph, wherein the vector aggregation network, the feature mining reverse processing network, the joint feature mining network and the feature mining second network are formed by performing network optimization operation on the second type of exemplary data and the third type of exemplary data based on rotation. That is, the feature mining operation may process the knowledge patterns to represent them in the form of vectors, and the feature mining reverse operation may process the data in the form of vectors to reconstruct or restore the corresponding knowledge patterns, so that the principle of the feature mining operation and the feature mining reverse operation may be reciprocal in the process.
It should be appreciated that, in some embodiments, the feature mining first network, the feature mining second network, the joint feature mining network, the vector aggregation network, and the feature mining inverse processing network may be included in a knowledge-graph reconstruction model, and thus, the big data tracking for cloud collaborative digitizing services may further include a step of network optimization of the knowledge-graph reconstruction model, which may further include the following specific embodiments described below:
extracting second class of exemplary data and third class of exemplary data, wherein the third class of exemplary data comprises an exemplary first group knowledge-graph, an exemplary second group knowledge-graph and a reconstructed comparison knowledge-graph (namely, a knowledge-graph with tag property), and the second class of exemplary data comprises an exemplary third group knowledge-graph and an exemplary fourth group knowledge-graph;
based on the third type of exemplary data, performing a comparison optimization operation on the knowledge graph reconstruction model to perform optimization updating on network parameters of the feature mining second network, the joint feature mining network, the vector aggregation network and the feature mining reverse processing network, wherein the reconstruction comparison knowledge graph is provided, so that comparison optimization can be performed;
Based on the second class of exemplary data, performing self-optimization operation on the knowledge graph reconstruction model to perform optimization updating on network parameters of the feature mining second network, the joint feature mining network, the vector aggregation network and the feature mining reverse processing network, wherein the knowledge graph reconstruction model does not have reconstruction contrast knowledge graphs, so that contrast optimization cannot be performed;
and carrying out the self-optimizing operation and the contrast optimizing operation in a rotating way to form the network optimization of the knowledge graph reconstruction model, and carrying out the self-optimizing operation and the contrast optimizing operation in a rotating way to ensure the optimizing precision of the knowledge graph reconstruction model.
It should be appreciated that, in some embodiments, the step of performing a contrast optimization operation on the knowledge-graph reconstruction model to perform an optimization update on the network parameters of the feature mining second network, the joint feature mining network, the vector aggregation network, and the feature mining inverse processing network based on the third class of exemplary data may include the following specific embodiments:
extracting a knowledge graph resolution model, wherein the knowledge graph reconstruction model is used for random sampling to mark as input data, processing the input data to obtain corresponding output data, and the output data imitates the exemplary data; the input data of the knowledge graph resolution model is exemplary data or output data of the knowledge graph reconstruction model, and the input data is used for resolving the output data of the knowledge graph reconstruction model from the exemplary data;
Reconstructing an exemplary first reconstructed group knowledge graph corresponding to the exemplary first group knowledge graph and the exemplary second group knowledge graph by using the knowledge graph reconstruction model, wherein the reconstruction process can refer to the related steps;
marking at least one of the exemplary first population knowledge-graph, the exemplary second population knowledge-graph, and the reconstructed comparison knowledge-graph to mark relevant exemplary data (i.e., positive exemplary data) of the knowledge-graph resolution model, and marking the exemplary first reconstructed population knowledge-graph to mark non-relevant exemplary data (i.e., negative exemplary data) of the knowledge-graph resolution model;
according to the related exemplary data and the non-related exemplary data, a comparison learning cost index for optimizing the knowledge-graph resolution model and the knowledge-graph reconstruction model is calculated, and the comparison learning cost index has a correlation with the difference between the first exemplary reconstruction group knowledge-graph and the reconstruction comparison knowledge-graph;
based on the comparison learning cost index, optimizing and updating network parameters of the feature mining second network, the joint feature mining network, the vector aggregation network and the feature mining reverse processing network, for example, the network parameters of the feature mining second network, the joint feature mining network, the vector aggregation network and the feature mining reverse processing network can be optimized and updated along the direction of reducing the comparison learning cost index.
It should be understood that, in some embodiments, the step of calculating, according to the relevant exemplary data and the non-relevant exemplary data, a comparative learning cost index for optimizing the knowledge-graph resolution model and the knowledge-graph reconstruction model may further include the following specific embodiments:
excavating a first network by utilizing the characteristics of the knowledge graph reconstruction model, excavating a group identification characterization vector of the first group knowledge graph, and excavating a group identification characterization vector of the first reconstructed group knowledge graph;
excavating a second network by utilizing the characteristics of the knowledge graph reconstruction model, excavating a group non-identification characterization vector of the reconstructed comparison knowledge graph, and excavating a group non-identification characterization vector of the first reconstructed group knowledge graph;
according to the resolution error of the knowledge-graph resolution model (that is, the error calculated based on the result of the resolution processing, the resolution processing is performed on the relevant exemplary data and the non-relevant exemplary data), the difference between the first exemplary reconstructed group knowledge graph and the reconstructed comparison knowledge graph, the difference between the first exemplary group knowledge graph and the first exemplary reconstructed group knowledge graph of the group identification characterization vector, the difference between the second reconstructed comparison knowledge graph and the first exemplary reconstructed group knowledge graph of the group non-identification characterization vector, the comparison learning cost index of the knowledge-graph reconstruction model is determined, for example, the weighted summation calculation can be performed on the resolution error and each difference (such as distance, difference degree, etc.), and the comparison learning cost index can be obtained.
It should be appreciated that, in some embodiments, the step of performing, based on the second class of exemplary data, the self-optimization operation on the knowledge-graph reconstruction model to perform optimization updating on the network parameters of the feature mining second network, the joint feature mining network, the vector aggregation network, and the feature mining inverse processing network may include the following specific embodiments:
reconstructing an exemplary second reconstructed group knowledge spectrum corresponding to the exemplary third group knowledge spectrum and the exemplary fourth group knowledge spectrum by using the knowledge spectrum reconstruction model, wherein the reconstruction process can refer to the related steps;
marking at least one of the exemplary third population knowledge-graph and the exemplary fourth population knowledge-graph to mark relevant exemplary data as the knowledge-graph resolution model, and marking the exemplary second reconstructed population knowledge-graph to mark irrelevant exemplary data as the knowledge-graph resolution model;
according to the related exemplary data and the non-related exemplary data, self-learning cost indexes for optimizing the knowledge-graph resolution model and the knowledge-graph reconstruction model are calculated, and the self-learning cost indexes have a related relationship with the differences between the second exemplary reconstructed group knowledge graph and the fourth exemplary group knowledge graph;
Based on the self-learning cost index, optimizing and updating network parameters of the feature mining second network, the joint feature mining network, the vector aggregation network and the feature mining reverse processing network, for example, the network parameters of the feature mining second network, the joint feature mining network, the vector aggregation network and the feature mining reverse processing network can be optimized and updated along the direction of reducing the self-learning cost index.
It should be understood, that, in some embodiments, the step of calculating the self-learning cost index for optimizing the knowledge-graph resolution model and the knowledge-graph reconstruction model according to the relevant exemplary data and the non-relevant exemplary data may further include the following specific embodiments:
excavating a first network by utilizing the characteristics of the knowledge graph reconstruction model, excavating a group identification characterization vector of the third group knowledge graph, and excavating a group identification characterization vector of the second group knowledge graph;
excavating a second network by utilizing the characteristics of the knowledge graph reconstruction model, excavating a group non-identification characterization vector of the fourth group knowledge graph, and excavating a group non-identification characterization vector of the second reconstructed group knowledge graph;
Determining a comparison learning cost index of the knowledge-graph reconstruction model according to a resolution error of the knowledge-graph resolution model (the resolution error can be obtained based on related exemplary data and non-related exemplary data), a difference between the exemplary second population knowledge-graph and the exemplary fourth population knowledge-graph, a difference between the exemplary third population knowledge-graph and the exemplary second population knowledge-graph of a population identification characterization vector, and a difference between the exemplary fourth population knowledge-graph and the exemplary second population knowledge-graph of a population non-identification characterization vector, for example, weighting and summing the resolution error and each difference to obtain a corresponding comparison learning cost index.
With reference to fig. 3, the embodiment of the invention further provides a big data tracking device for cloud collaborative digital services, which can be applied to the AI system. The big data tracking device for cloud collaborative digital service may include:
the knowledge graph extraction module is used for extracting a first group knowledge graph and a second group knowledge graph, wherein the first group knowledge graph is formed by performing mapping processing on current behavior data of a plurality of objects to be tracked, which are included in the object group to be tracked, and the second group knowledge graph is formed by performing mapping processing on historical behavior data of a plurality of objects to be tracked, which are included in the object group to be tracked;
The feature mining module is used for performing feature mining operation on the first group knowledge graph and the second group knowledge graph respectively to form a group identification characterization vector of the first group knowledge graph and a group non-identification characterization vector of the second group knowledge graph, wherein the group identification characterization vector is used for characterizing group specific information of the group of the objects to be tracked, and the group non-identification characterization vector is used for characterizing other feature information except for the group specific information of the group of the objects to be tracked;
the joint feature mining module is used for performing joint feature mining operation on the first group knowledge graph and the second group knowledge graph so as to output joint mining characterization vectors corresponding to the first group knowledge graph and the second group knowledge graph in the whole;
the vector aggregation module is used for carrying out vector aggregation operation on the group identification characterization vector, the group non-identification characterization vector and the joint mining characterization vector so as to form a corresponding aggregate group characterization vector;
the feature mining reverse module is used for performing feature mining reverse operation on the aggregate group representation vectors so as to output corresponding third group knowledge patterns, wherein a correlation is formed between the third group knowledge patterns and group identification representation vectors of the first group knowledge patterns, and a correlation is formed between the third group knowledge patterns and group non-identification representation vectors of the second group knowledge patterns;
And the group anomaly analysis module is used for carrying out group anomaly analysis operation on the object group to be tracked based on the third group knowledge graph so as to obtain a group anomaly analysis result corresponding to the object group to be tracked, so as to realize behavior anomaly tracking on the object group to be tracked.
In summary, the big data tracking method and the AI system for the cloud collaborative digital service provided by the invention can firstly extract the first group knowledge graph and the second group knowledge graph; respectively performing feature mining operation on the first group knowledge graph and the second group knowledge graph to form a group identification characterization vector of the first group knowledge graph and a group non-identification characterization vector of the second group knowledge graph; performing joint feature mining operation on the first group knowledge graph and the second group knowledge graph to output joint mining characterization vectors corresponding to the first group knowledge graph and the second group knowledge graph on the whole; vector aggregation operation is carried out on the group identification characterization vector, the group non-identification characterization vector and the joint mining characterization vector so as to form a corresponding aggregation group characterization vector; performing feature mining reverse operation on the aggregated group characterization vector to output a corresponding third group knowledge graph; and carrying out group abnormality analysis operation on the object group to be tracked based on the third group knowledge graph so as to obtain a group abnormality analysis result corresponding to the object group to be tracked. Based on the above, the first group knowledge graph is not directly analyzed, but the second group knowledge graph is combined to reconstruct the third group knowledge graph, and the third group knowledge graph has a correlation with the group identification characterization vector of the first group knowledge graph and a correlation with the group non-identification characterization vector of the second group knowledge graph, so that the current behavior data and the historical behavior data can be fully combined, and therefore, the reliability of anomaly tracking can be improved to a certain extent by carrying out group anomaly analysis operation on the group of objects to be tracked according to the third group knowledge graph, and the joint feature mining operation can be carried out in the reconstruction process, so that the reliability of the mined information is higher, the reliability of anomaly tracking can be further improved, and the problem of low reliability in the prior art is solved.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (10)

1. The big data tracking method for the cloud collaborative digital service is characterized by comprising the following steps of:
extracting a first group knowledge graph and a second group knowledge graph, wherein the first group knowledge graph is formed by performing mapping processing on current behavior data of a plurality of objects to be tracked, which are included in a group of objects to be tracked, and the second group knowledge graph is formed by performing mapping processing on historical behavior data of a plurality of objects to be tracked, which are included in the group of objects to be tracked, and the current behavior data and the historical behavior data are in a data form including text data;
performing feature mining operation on the first group knowledge graph and the second group knowledge graph respectively to form a group identification characterization vector of the first group knowledge graph and a group non-identification characterization vector of the second group knowledge graph, wherein the group identification characterization vector is used for characterizing group-specific information of the group of the objects to be tracked, and the group non-identification characterization vector is used for characterizing other feature information except for the group-specific information of the group of the objects to be tracked;
Performing joint feature mining operation on the first group knowledge graph and the second group knowledge graph to output joint mining characterization vectors corresponding to the first group knowledge graph and the second group knowledge graph in the whole;
vector aggregation operation is carried out on the group identification characterization vector, the group non-identification characterization vector and the joint mining characterization vector so as to form a corresponding aggregation group characterization vector;
performing feature mining reverse operation on the aggregate group representation vectors to output corresponding third group knowledge patterns, wherein a correlation is formed between the third group knowledge patterns and group identification representation vectors of the first group knowledge patterns, and a correlation is formed between the third group knowledge patterns and group non-identification representation vectors of the second group knowledge patterns;
and carrying out group abnormality analysis operation on the object group to be tracked based on the third group knowledge graph to obtain a group abnormality analysis result corresponding to the object group to be tracked so as to realize behavior abnormality tracking of the object group to be tracked.
2. The big data tracking method for cloud collaborative digital services according to claim 1, wherein the step of extracting a first population knowledge-graph and a second population knowledge-graph comprises:
Constructing a first object knowledge graph and a second object knowledge graph, wherein the first object knowledge graph is formed by performing mapping processing on current behavior data of a plurality of management objects, the second object knowledge graph is formed by performing mapping processing on historical behavior data of the plurality of management objects, each graph member in the first object knowledge graph corresponds to one management object in the plurality of management objects, attribute data of the graph member at least comprises corresponding current behavior data, each graph member in the second object knowledge graph corresponds to one management object in the plurality of management objects, and attribute data of the graph member at least comprises corresponding historical behavior data;
respectively carrying out group determination operation on the first object knowledge graph and the second object knowledge graph to determine group parts in the first object knowledge graph and the second object knowledge graph;
extracting a corresponding first group knowledge graph based on the group part determined in the first object knowledge graph, and extracting a corresponding second group knowledge graph based on the group part determined in the second object knowledge graph;
The big data tracking method for the cloud collaborative digital service further comprises the following steps:
based on the third group knowledge graph, carrying out replacement operation on the group part in the second object knowledge graph to form a corresponding third object knowledge graph, wherein a correlation is formed between the third object knowledge graph and a group identification characterization vector of the group part in the first object knowledge graph, and a correlation is formed between the third object knowledge graph and a group non-identification characterization vector of the group part in the second object knowledge graph;
and performing object anomaly analysis operation on the plurality of management objects based on the third object knowledge graph to obtain object anomaly analysis results corresponding to the plurality of management objects so as to track behavior anomalies of the plurality of management objects.
3. The big data tracking method for cloud collaborative digital services according to claim 1, wherein the step of performing feature mining operations on the first population knowledge-graph and the second population knowledge-graph to form a population identification token vector of the first population knowledge-graph and a population non-identification token vector of the second population knowledge-graph, respectively, comprises:
Performing feature mining operation on the first group knowledge graph by utilizing a feature mining first network to form a group identification characterization vector of the first group knowledge graph, wherein the feature mining first network performs network optimization operation on the basis of first type of exemplary data to form the group identification characterization vector;
and performing feature mining operation on the second group knowledge graph by using a feature mining second network which is different from the feature mining first network to form a group non-identification characterization vector of the second group knowledge graph, wherein the feature mining second network is formed by performing network optimization operation on the basis of second-class exemplary data and third-class exemplary data rotation, the second-class exemplary data is not configured with corresponding data tags, and the third-class exemplary data is configured with corresponding data tags.
4. The big data tracking method for cloud collaborative digital services according to claim 3, wherein the step of performing feature mining operations on the first group knowledge graph to form a group identification token vector of the first group knowledge graph using a feature mining first network comprises:
loading first map data distribution of the first group knowledge maps to load the first map data distribution into a feature mining first network; and utilizing the characteristics to mine a first network, and performing characteristic mining operation on the first map data distribution to form a group identification characterization vector of the first group knowledge map;
The step of performing feature mining operation on the second group knowledge graph by using features different from the feature mining first network to form a group non-identification characterization vector of the second group knowledge graph, includes:
loading a second graph data distribution of the second group of knowledge graphs to a second feature mining network different from the first feature mining network; and utilizing the characteristics to excavate a second network, performing characteristic excavation operation on the second spectrum data distribution to form a group non-identification characterization vector of the second group knowledge spectrum, wherein network parameters of the characteristic excavation first network and the characteristic excavation second network are different.
5. The big data tracking method for cloud collaborative digitizing service according to claim 3, wherein the step of performing a joint feature mining operation on the first population knowledge-graph and the second population knowledge-graph to output joint mining token vectors corresponding to the first population knowledge-graph and the second population knowledge-graph as a whole comprises:
and performing joint feature mining operation on the first group knowledge graph and the second group knowledge graph by using a joint feature mining network to output joint mining characterization vectors corresponding to the first group knowledge graph and the second group knowledge graph on the whole, and performing network optimization operation on the second type of exemplary data and the third type of exemplary data based on rotation to form the joint feature mining network and the feature mining second network.
6. The big data tracking method for cloud collaborative digitizing service according to claim 5, wherein the step of vector aggregating the group identification token vector, the group non-identification token vector, and the joint mining token vector to form a corresponding aggregate group token vector comprises:
performing vector aggregation operation on the group identification characterization vector, the group non-identification characterization vector and the joint mining characterization vector by using a vector aggregation network to form a corresponding aggregated group characterization vector;
the step of performing feature mining reverse operation on the aggregated group characterization vector to output a corresponding third group knowledge graph includes:
and performing feature mining reverse operation on the aggregated group characterization vector by using a feature mining reverse processing network to output a corresponding third group knowledge graph, wherein the vector aggregation network, the feature mining reverse processing network, the joint feature mining network and the feature mining second network are formed by performing network optimization operation on the second type of exemplary data and the third type of exemplary data based on rotation.
7. The big data tracking method for cloud collaborative digitizing service according to claim 6, wherein the step of performing a vector aggregation operation on the group identification token vector, the group non-identification token vector, and the joint mining token vector to form a corresponding aggregated group token vector using a vector aggregation network comprises:
performing cascade combination operation on the group identification characterization vector, the group non-identification characterization vector and the joint mining characterization vector by using a front-end processing unit of a vector aggregation network so as to output a corresponding cascade combination characterization vector;
and performing deep filtering operation on the cascade combination characterization vector by utilizing a back-end processing unit of the vector aggregation network to form a corresponding aggregation group characterization vector.
8. The big data tracking method for cloud collaborative digital services according to claim 6, wherein the feature mining first network, the feature mining second network, the joint feature mining network, the vector aggregation network, and the feature mining inverse processing network are included in a knowledge-graph reconstruction model, the step of network optimization of the knowledge-graph reconstruction model comprising:
Extracting second-class exemplary data and third-class exemplary data, wherein the third-class exemplary data comprises an exemplary first group knowledge-graph, an exemplary second group knowledge-graph and a reconstructed comparison knowledge-graph, and the second-class exemplary data comprises an exemplary third group knowledge-graph and an exemplary fourth group knowledge-graph;
based on the third type of exemplary data, comparing and optimizing the knowledge graph reconstruction model to optimize and update network parameters of the feature mining second network, the joint feature mining network, the vector aggregation network and the feature mining reverse processing network;
based on the second class of exemplary data, carrying out self-optimization operation on the knowledge graph reconstruction model so as to carry out optimization updating on network parameters of the feature mining second network, the joint feature mining network, the vector aggregation network and the feature mining reverse processing network;
and carrying out the self-optimizing operation and the contrast optimizing operation in a rotating way to form the network optimization of the knowledge graph reconstruction model.
9. The big data tracking method for cloud collaborative digital services according to claim 8, wherein the step of performing a contrast optimization operation on the knowledge-graph reconstruction model based on the third class of exemplary data to perform an optimization update on network parameters of the feature mining second network, the joint feature mining network, the vector aggregation network, and the feature mining inverse processing network comprises:
Extracting a knowledge graph resolution model; and reconstructing an exemplary first reconstructed population knowledge graph corresponding to the exemplary first population knowledge graph and the exemplary second population knowledge graph by using the knowledge graph reconstruction model; and, marking at least one of the exemplary first population knowledge-graph, the exemplary second population knowledge-graph, and the reconstructed comparison knowledge-graph to mark as relevant exemplary data of the knowledge-graph resolution model, and marking the exemplary first reconstructed population knowledge-graph to mark as irrelevant exemplary data of the knowledge-graph resolution model; according to the related exemplary data and the non-related exemplary data, a comparison learning cost index for optimizing the knowledge-graph resolution model and the knowledge-graph reconstruction model is calculated, and the comparison learning cost index has a correlation with the difference between the first exemplary reconstruction group knowledge-graph and the reconstruction comparison knowledge-graph; based on the comparison learning cost index, optimizing and updating network parameters of the feature mining second network, the joint feature mining network, the vector aggregation network and the feature mining reverse processing network;
And performing self-optimization operation on the knowledge graph reconstruction model based on the second class of exemplary data to perform optimization updating on network parameters of the feature mining second network, the joint feature mining network, the vector aggregation network and the feature mining reverse processing network, wherein the method comprises the following steps:
reconstructing an exemplary second reconstructed group knowledge graph corresponding to the exemplary third group knowledge graph and the exemplary fourth group knowledge graph by using the knowledge graph reconstruction model; and, marking at least one of the exemplary third population knowledge-graph and the exemplary fourth population knowledge-graph to mark as relevant exemplary data of the knowledge-graph resolution model, and marking the exemplary second reconstructed population knowledge-graph to mark as irrelevant exemplary data of the knowledge-graph resolution model; and according to the related exemplary data and the non-related exemplary data, calculating self-learning cost indexes for optimizing the knowledge-graph resolution model and the knowledge-graph reconstruction model, wherein the self-learning cost indexes have a related relationship with the differences between the second exemplary reconstructed group knowledge graph and the fourth exemplary group knowledge graph; and optimizing and updating network parameters of the feature mining second network, the joint feature mining network, the vector aggregation network and the feature mining reverse processing network based on the self-learning cost index.
10. An AI system comprising a processor and a memory, the memory for storing a computer program, the processor for executing the computer program to implement the method of any of claims 1-9.
CN202310500410.1A 2023-05-06 2023-05-06 Big data tracking method and AI system for cloud collaborative digital service Active CN116501876B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310500410.1A CN116501876B (en) 2023-05-06 2023-05-06 Big data tracking method and AI system for cloud collaborative digital service

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310500410.1A CN116501876B (en) 2023-05-06 2023-05-06 Big data tracking method and AI system for cloud collaborative digital service

Publications (2)

Publication Number Publication Date
CN116501876A CN116501876A (en) 2023-07-28
CN116501876B true CN116501876B (en) 2023-12-08

Family

ID=87322773

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310500410.1A Active CN116501876B (en) 2023-05-06 2023-05-06 Big data tracking method and AI system for cloud collaborative digital service

Country Status (1)

Country Link
CN (1) CN116501876B (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109829089A (en) * 2018-12-12 2019-05-31 中国科学院计算技术研究所 Social network user method for detecting abnormality and system based on association map
CN112416994A (en) * 2019-08-21 2021-02-26 中移(苏州)软件技术有限公司 Information processing method, device and storage medium
CN113538137A (en) * 2021-07-29 2021-10-22 中国工商银行股份有限公司 Capital flow monitoring method and device based on double-spectrum fusion calculation
CN114942947A (en) * 2022-07-04 2022-08-26 上海市胸科医院 Follow-up visit data processing method and system based on intelligent medical treatment

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109829089A (en) * 2018-12-12 2019-05-31 中国科学院计算技术研究所 Social network user method for detecting abnormality and system based on association map
CN112416994A (en) * 2019-08-21 2021-02-26 中移(苏州)软件技术有限公司 Information processing method, device and storage medium
CN113538137A (en) * 2021-07-29 2021-10-22 中国工商银行股份有限公司 Capital flow monitoring method and device based on double-spectrum fusion calculation
CN114942947A (en) * 2022-07-04 2022-08-26 上海市胸科医院 Follow-up visit data processing method and system based on intelligent medical treatment

Also Published As

Publication number Publication date
CN116501876A (en) 2023-07-28

Similar Documents

Publication Publication Date Title
CN110210227B (en) Risk detection method, device, equipment and storage medium
CN106250461A (en) A kind of algorithm utilizing gradient lifting decision tree to carry out data mining based on Spark framework
CN116126945B (en) Sensor running state analysis method and system based on data analysis
CN116109121B (en) User demand mining method and system based on big data analysis
CN114598539B (en) Root cause positioning method and device, storage medium and electronic equipment
CN116126947B (en) Big data analysis method and system applied to enterprise management system
CN116109630B (en) Image analysis method and system based on sensor acquisition and artificial intelligence
CN113688490A (en) Network co-construction sharing processing method, device, equipment and storage medium
CN115800538A (en) Intelligent power grid operation and maintenance monitoring method and system based on artificial intelligence
CN116489038A (en) Network traffic prediction method, device, equipment and medium
CN115730659A (en) Vehicle safety analysis method and system applied to AI (Artificial Intelligence)
CN114328277A (en) Software defect prediction and quality analysis method, device, equipment and medium
CN116501876B (en) Big data tracking method and AI system for cloud collaborative digital service
CN116361567B (en) Data processing method and system applied to cloud office
CN117194219A (en) Fuzzy test case generation and selection method, device, equipment and medium
CN116582414A (en) Fault root cause positioning method, device, equipment and readable storage medium
CN113535815B (en) Business operation behavior big data mining method and system suitable for electronic commerce
CN115616408A (en) Battery thermal management data processing method and system
CN114881521A (en) Service evaluation method, device, electronic equipment and storage medium
CN116910729B (en) Nuclear body processing method and system applied to multi-organization architecture
CN112750047A (en) Behavior relation information extraction method and device, storage medium and electronic equipment
CN116662415B (en) Intelligent matching method and system based on data mining
CN116994609B (en) Data analysis method and system applied to intelligent production line
CN117236617B (en) Enterprise business management method and system
CN116644935A (en) Digital service processing method and digital service system based on artificial intelligence

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
TA01 Transfer of patent application right

Effective date of registration: 20230913

Address after: 650000 Jinniu Community, Wen'an Road, Xishan District, Kunming City, Yunnan Province

Applicant after: Wang Gang

Address before: 650000 Jinniu Road, Xishan District, Kunming, Yunnan

Applicant before: Kunming Yukang Technology Co.,Ltd.

TA01 Transfer of patent application right
TA01 Transfer of patent application right

Effective date of registration: 20231115

Address after: 712039 No. 2-1-90, 2nd Floor, No. 6 Jinxu Road, Weicheng Street Office, Qinhan New City, Xixian New District, Xi'an City, Shaanxi Province

Applicant after: GTCOM Technology (Shaanxi) Co.,Ltd.

Address before: 650000 Jinniu Community, Wen'an Road, Xishan District, Kunming City, Yunnan Province

Applicant before: Wang Gang

TA01 Transfer of patent application right
GR01 Patent grant
GR01 Patent grant