CN116436802A - Intelligent control method and system based on machine learning - Google Patents
Intelligent control method and system based on machine learning Download PDFInfo
- Publication number
- CN116436802A CN116436802A CN202310691791.6A CN202310691791A CN116436802A CN 116436802 A CN116436802 A CN 116436802A CN 202310691791 A CN202310691791 A CN 202310691791A CN 116436802 A CN116436802 A CN 116436802A
- Authority
- CN
- China
- Prior art keywords
- sub
- resource
- resource transmission
- network
- path
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000010801 machine learning Methods 0.000 title claims abstract description 37
- 230000005540 biological transmission Effects 0.000 claims abstract description 188
- 238000013508 migration Methods 0.000 claims abstract description 19
- 230000005012 migration Effects 0.000 claims abstract description 19
- 238000012549 training Methods 0.000 claims description 70
- 230000006870 function Effects 0.000 claims description 19
- 230000004044 response Effects 0.000 claims description 17
- 238000012360 testing method Methods 0.000 claims description 16
- 230000006978 adaptation Effects 0.000 claims description 11
- 238000013526 transfer learning Methods 0.000 claims description 8
- 238000001914 filtration Methods 0.000 claims description 6
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 238000012546 transfer Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 3
- 230000008569 process Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 230000002452 interceptive effect Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 1
- 238000005056 compaction Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000012905 input function Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/16—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks using machine learning or artificial intelligence
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/04—Network management architectures or arrangements
- H04L41/044—Network management architectures or arrangements comprising hierarchical management structures
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/14—Network analysis or design
- H04L41/145—Network analysis or design involving simulating, designing, planning or modelling of a network
-
- 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
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The invention provides an intelligent control method and system based on machine learning, which relate to the technical field of data processing, wherein the method comprises the following steps: according to the connection relation of a first computer cluster of a target enterprise, a first computer network is generated, a historical resource transmission path of the first computer network is collected, characteristics of the historical resource transmission path are identified, path distribution characteristics are obtained to split the first computer network, a plurality of split sub-networks are generated to carry out migration learning, a plurality of control sub-models respectively corresponding to the split sub-networks are output, and resource transmission control is carried out on the first computer cluster according to the plurality of control sub-models.
Description
Technical Field
The invention relates to the technical field of data processing, in particular to an intelligent control method and system based on machine learning.
Background
In recent years, machine learning has thoroughly changed many areas of intelligent control. Machine learning has a great influence on a machine learning community by solving the challenges which cannot be solved by the traditional mode control method, and the introduction of the machine learning greatly improves the system precision specially designed for resource control, object detection, data control and resource transmission control. A key aspect of machine learning is that the features used to interpret the data are automatically learned from training data, rather than being manually crafted by an engineer.
In the prior art, machine learning iteration is lacked in the process of transmitting the resources, so that the technical problem of low resource transmission efficiency is caused.
Disclosure of Invention
The application provides an intelligent control method and system based on machine learning, which are used for solving the technical problem of low resource transmission efficiency caused by lack of machine learning iteration in the process of transmitting resources in the prior art.
In view of the above, the present application provides an intelligent control method and system based on machine learning.
In a first aspect, the present application provides a machine learning-based intelligent control method, the method comprising: acquiring a first computer cluster of a target enterprise; generating a first computer network according to the connection relation of the first computer cluster; collecting a historical resource transmission path of the first computer network, and carrying out feature recognition on the historical resource transmission path to obtain path distribution features; splitting the first computer network according to the path distribution characteristics to generate a plurality of split sub-networks; performing migration learning on the plurality of split sub-networks, and outputting a plurality of control sub-models respectively corresponding to the plurality of split sub-networks; and controlling the resource transmission of the first computer cluster according to the plurality of control sub-models.
In a second aspect, the present application provides a machine learning based intelligent control system, the system comprising: the cluster acquisition module is used for acquiring a first computer cluster of a target enterprise; the network acquisition module is used for generating a first computer network according to the connection relation of the first computer cluster; the characteristic identification module is used for acquiring a historical resource transmission path of the first computer network, carrying out characteristic identification on the historical resource transmission path and acquiring path distribution characteristics; the splitting module is used for splitting the first computer network according to the path distribution characteristics to generate a plurality of splitting sub-networks; the migration learning module is used for performing migration learning on the plurality of split sub-networks and outputting a plurality of control sub-models respectively corresponding to the plurality of split sub-networks; and the control module is used for controlling the resource transmission of the first computer cluster according to the plurality of control submodels.
One or more technical solutions provided in the present application have at least the following technical effects or advantages:
the intelligent control method and the intelligent control system based on machine learning, which are provided by the application, relate to the technical field of data processing, solve the technical problem that in the prior art, the resource transmission efficiency is low due to lack of machine learning iteration in the process of transmitting the resource, realize rationalized intelligent control on the resource transmission based on machine learning, and further improve the resource transmission efficiency.
Drawings
FIG. 1 is a schematic flow chart of an intelligent control method based on machine learning;
FIG. 2 is a schematic diagram of a splitting process of a first computer network in an intelligent control method based on machine learning;
FIG. 3 is a schematic diagram of a scheduling control flow in the intelligent control method based on machine learning;
FIG. 4 is a schematic flow chart of a plurality of control sub-models in the intelligent control method based on machine learning;
fig. 5 is a schematic structural diagram of an intelligent control system based on machine learning.
Reference numerals illustrate: the system comprises a cluster acquisition module 1, a network acquisition module 2, a feature identification module 3, a splitting module 4, a migration learning module 5 and a control module 6.
Detailed Description
The intelligent control method and the intelligent control system based on machine learning are used for solving the technical problem that in the prior art, the resource transmission efficiency is low due to the lack of machine learning iteration in the process of transmitting the resource.
Example 1
As shown in fig. 1, an embodiment of the present application provides an intelligent control method based on machine learning, including:
step S100: acquiring a first computer cluster of a target enterprise;
specifically, the intelligent control method based on machine learning provided by the embodiment of the application is applied to an intelligent control system based on machine learning, so that in order to ensure the accuracy of controlling resource transmission in the later stage, a first computer cluster of a target enterprise is required to be extracted, the target enterprise is related to a transmission environment and transmission contents where resources to be transmitted are located, the first computer cluster refers to a plurality of computers in the target enterprise, so that at least two computers are randomly selected from the plurality of computers, and the at least two computers are used as the first computer cluster of the target enterprise, and the control when the resource transmission in the target enterprise is realized in the later stage is used as an important reference.
Step S200: generating a first computer network according to the connection relation of the first computer cluster;
specifically, on the basis of the first computer cluster extracted from the target enterprise, the connection relationship between all the computers included in the first computer cluster is determined, the connection manner between the computers may be a plurality of connection manners such as direct connection through a network cable, coaxial cable connection, hub connection, bridge connection, switch connection, router connection, and the like, and the first computer network corresponding to the first computer cluster is determined based on the connection manner between the computers in the first computer cluster, and if the computers in the first computer cluster are connected through a router, then, firstly, it is determined whether the IP of the target host receiving the data resource is in the same network segment as the current host, that is, the IP of the target host is in a sub-network mask of the previous network segment, if the IP of the target host is in the same network segment, the mac address of the other party is obtained by transmitting ARP packets, and if the mac address of the other party is not in the same network segment, the mac address of the other party is obtained by transmitting ARP packets through the router, and if the mac address of the other party is not in the same network segment, and the first computer network is used as the first network, and the communication resource is further controlled by the first computer network.
Step S300: collecting a historical resource transmission path of the first computer network, and carrying out feature recognition on the historical resource transmission path to obtain path distribution features;
specifically, in order to increase the rate of resource transmission in the target enterprise, the resource transmission path during data resource transmission needs to be collected according to the generated first computer network before the current moment, where the resource transmission path refers to a computer node through which resource data passes from a computer end sending the resource to a computer end receiving the resource in the first computer cluster, on this basis, a historical path of resource transmission is obtained, and at the same time, feature recognition is performed on the historical resource transmission path, which refers to the degree of density between all paths included in the historical resource transmission path, and illustratively, different departments will be separated in the target enterprise, and transmission and reception of resources between personnel in each department will cross more, so that the distribution feature of the resource transmission path is extracted, which refers to the characteristics of the resource transmission distribution path between different departments and in the same department, and then the distribution feature of the resource transmission path in the target enterprise is obtained, which is the control and compaction basis for implementing the internal resource transmission in the target enterprise later.
Step S400: splitting the first computer network according to the path distribution characteristics to generate a plurality of split sub-networks;
further, as shown in fig. 2, step S400 of the present application further includes:
step S410: the path distribution characteristics are obtained, wherein the path distribution characteristics comprise the number of path transmission objects, the number of path transmission target objects and the path transmission frequency;
step S420: performing density identification on each node in the first computer network according to the number of path transmission objects, the number of path transmission target objects and the path transmission frequency to obtain a density identification index;
step S430: splitting the first computer network according to the density identification index.
Further, step S430 of the present application includes:
step S431: identifying the density identification index, obtaining nodes which are larger than or equal to a preset density identification index, and determining a plurality of clustering centers;
step S432: clustering the first computer network by using the plurality of clustering centers to obtain a clustering result;
step S433: determining a split node set according to edge nodes of adjacent clusters in the clustering result;
step S434: splitting the first computer network according to the splitting node set.
Specifically, splitting the first computer network in the target enterprise based on the obtained path distribution feature refers to the number of path transmission objects, the number of path transmission target objects and the path transmission frequency contained in the path distribution feature based on the path distribution feature, the number of path transmission objects refers to the number of resource sending ends in the resource transmission process, the number of path transmission target objects refers to the number of resource receiving ends in the resource transmission process, the path transmission frequency refers to the number of transmission times of resources in the transmission path, further, performing density identification on each node contained in the first computer network according to the number of path transmission objects, the number of path transmission target objects and the path transmission frequency, determining the number of computer nodes contained in the first computer network according to the number of path transmission objects and the number of path transmission target objects, the method comprises identifying the length of paths among computer nodes in a first computer network according to path transmission frequency, wherein the shorter the path length is and the greater the number of computer nodes is, the greater the density is, the density of each node in the first computer network is identified on the basis, so as to obtain a density identification index, meanwhile, on the basis of the computer nodes with the density identification index, the first computer network is split according to different density identification indexes, namely, firstly, the density identification index is identified, the node with the density identification index greater than or equal to a preset density identification index is extracted, wherein the preset density identification index is preset by related technicians according to standard density data in big data, further, all the extracted computer nodes with the density identification index larger than or equal to the preset density identification index are marked as a plurality of clustering centers, the clustering of the computer nodes is carried out on the first computer network according to the position of each clustering center by taking the plurality of clustering centers as a reference, the splitting node set is determined according to the edge nodes of the adjacent clusters existing in the clustering result after clustering, namely the edge computer nodes contained on the splitting line of the adjacent clustering area, the edge computer nodes are marked as splitting node sets when the first computer network is split, the first computer network is split according to the positions of the obtained splitting node sets, and the split sub-networks are marked as a plurality of split sub-networks to be output, so that the control when the resource transmission in a target enterprise is carried out is realized.
Step S500: performing migration learning on the plurality of split sub-networks, and outputting a plurality of control sub-models respectively corresponding to the plurality of split sub-networks;
further, as shown in fig. 3, step S500 of the present application further includes:
step S510: determining a first sub-network according to the plurality of split sub-networks;
step S520: generating a resource transmission training sample set and a resource transmission training test sample set according to the resource transmission task type corresponding to the first sub-network, wherein each sample set comprises a resource packet size, a resource storage type, a resource storage queue, a resource transmission time and a convergence index for identifying resource delay;
step S530: and performing model training by using the resource transmission training sample set and the resource transmission training test sample set, and outputting a control sub-model corresponding to the first sub-network for scheduling and controlling the resource transmission task in the first sub-network.
Further, as shown in fig. 4, step S500 of the present application further includes:
step S540: acquiring resource transmission task types corresponding to the rest sub-networks in the plurality of split sub-networks;
step S550: comparing the resource transmission task type corresponding to the residual sub-network with the resource transmission task type corresponding to the first sub-network to obtain a first similarity;
step S560: and according to the first similarity, controlling the transfer learning weight layers of the resource transmission training sample set and the resource transmission training test sample set to carry out coefficient adjustment, and outputting a plurality of control sub-models corresponding to the plurality of split sub-networks.
Specifically, in order to ensure the efficiency of transmitting resources in the first computer network, once you need to build a plurality of control sub-models corresponding to the split sub-networks, then transfer-learn the control sub-models, build a plurality of control sub-models corresponding to the split sub-networks, and determine the first sub-network according to the split sub-networks, wherein the determined first sub-network is a sub-network randomly selected from the split sub-networks as an initial network, further, extract the resource transmission task type corresponding to the first sub-network, the resource transmission task type corresponding to the first sub-network can comprise multiple-target transmission, single-target interactive transmission, resource encryption transmission and other types, collect each data of resource transmission in the historical task on the basis of the control sub-model, generate a resource transmission training sample set and a resource transmission training sample set, and each sample set comprises a resource packet size, a resource storage type, a resource storage queue, a resource transmission time index, and a resource delay index, the packet size refers to the current transmission resource size refers to the resource type, the current transmission resource size refers to the resource type is the resource size, the resource storage queue, the resource storage time index is the resource storage time index, the current transmission delay is the current transmission resource storage type is the resource storage time index, the current transmission delay is the resource storage type is the current transmission resource storage time index, and the current transmission type is the storage resource storage time index is the storage resource storage type, and the storage time index is the storage of the current transmission type, a determined value to which the convergence index tends is identified.
Further, model training is carried out on the generated resource transmission training sample set and the generated resource transmission training test sample set, and a control sub-model corresponding to the first sub-network is output
The control sub-model corresponding to the first sub-network is a neural network model in machine learning, and can be subjected to self iterative optimization continuously, the control sub-model corresponding to the first sub-network is obtained through training of a resource transmission training sample set and a resource transmission training test sample set, wherein each group of resource transmission training data in the resource transmission training sample set comprises a resource packet size, a resource storage type, a resource storage queue, a resource transmission time, and a convergence index for identifying resource delay, and the resource transmission training test sample set is test data corresponding to the resource transmission training sample set one by one.
The control sub-model construction process corresponding to the first sub-network comprises the following steps: inputting each group of resource transmission training data in the resource transmission training sample set into the control sub-model corresponding to the first sub-network, performing output test adjustment of the control sub-model corresponding to the first sub-network through the test data corresponding to the group of resource transmission training data, and finishing the training of the control sub-model corresponding to the first sub-network when the output result of the control sub-model corresponding to the first sub-network is consistent with the test data, and finishing the current group of resource transmission training, and finishing the training of all the resource transmission training data in the resource transmission training sample set.
The control sub-model corresponding to the first sub-network after training is used for scheduling and controlling the resource transmission task in the first sub-network, which means that when the resource task time delay occurs, the task can be scheduled and controlled.
Further, to obtain multiple control sub-models corresponding to multiple split sub-networks, firstly determining the types of resource transmission tasks corresponding to the remaining sub-networks in the multiple split sub-networks, wherein the types of resource transmission tasks corresponding to the remaining sub-networks in the multiple split sub-networks refer to the types of resource transmission tasks corresponding to the remaining sub-networks in the multiple split sub-networks except for the determined types of resource transmission tasks corresponding to the first sub-network, the types of resource transmission tasks can be multi-target transmission, single-target interactive transmission, resource encryption transmission and other types, further, comparing the types of resource transmission tasks corresponding to the remaining sub-networks with the types of resource transmission tasks corresponding to the first sub-network according to the types of resource transmission tasks corresponding to the remaining sub-networks, namely, the similarity of the types of resource transmission tasks corresponding to the remaining sub-networks with the types of resource transmission tasks corresponding to the first sub-network, if the similarities of the types of resource transmission tasks corresponding to the remaining sub-networks in the multiple split sub-networks are higher, then considering that the types of resource sample sets in the remaining sub-networks are more similar to the types of resource transmission tasks corresponding to the first sub-network, so that the similarity of the types of resource transmission tasks in the remaining sub-networks can be more than the types of resource transmission tasks corresponding to the first sub-networks, thereby forming the first sub-network, and training the models can be based on the first sub-network, and the corresponding to the corresponding control sub-network can be a training model, and the corresponding to the first sub-network, and the model is based on the corresponding to the first sub-model, so as to be used as reference data for later control when resource transmission is carried out in the target enterprise.
Further, step S560 of the present application includes:
step S561: obtaining a difference sub-network with similarity smaller than preset similarity, and establishing a migration filter layer according to the difference sub-network;
step S562: filtering the sample set for transfer learning according to the transfer filter layer to obtain a transferable sample set;
step S563: training the difference sub-network according to the movable sample set and the newly added sample set, and outputting a control sub-model corresponding to the difference sub-network.
Further, step S560 of the present application includes:
step S564: taking the first similarity as an adaptation target, taking a training weight layer of the resource transmission training sample set as an input target, and establishing an adaptation function;
step S565: introducing a first loss function to analyze the first similarity to obtain first loss data;
step S566: and inputting the first loss data into the adaptive function to respond, and obtaining a first response result, wherein the first response result comprises a weight layer with the coefficient adjusted.
Specifically, a sub-network with similarity smaller than preset similarity is recorded as a difference sub-network according to the first similarity obtained by comparing the resource transmission task types corresponding to the rest sub-networks with the resource transmission task types corresponding to the first sub-network, wherein the preset similarity is preset by related technicians according to the data quantity of the resource transmission task types, when one sub-network similarity is smaller than the preset similarity, the resource transmission task type contained in the current sub-network is regarded as the unusual resource transmission task type, meanwhile, a migration filter layer is established on the basis of the obtained difference sub-network, the migration filter layer can be used for filtering a resource sample set for migration learning, reserving useful resource sample data in a transfer learning resource sample, filtering and screening ineffective resource sample data, wherein the useful resource sample data is similar to the resource sample data in a first sub-network in a differential network, the ineffective resource sample data is dissimilar to the resource sample data in the first sub-network in the differential network, thereby recording the useful resource sample data as a movable sample set, further training the differential sub-network according to the movable sample set and a newly added sample set, wherein the newly added sample set is used for supplementing the similar resource sample data to the dissimilar resource sample data in the first sub-network in the differential network, and outputting a control sub-model corresponding to the differential sub-network
The control sub-model corresponding to the difference sub-network is obtained through training of a training data set and a supervision data set, wherein each group of training data in the training data set comprises a movable sample set and a newly added sample set, and the supervision data set is supervision data corresponding to the training data set one by one.
And inputting each group of training data in the training data set into a control sub-model corresponding to the difference sub-network, performing output supervision adjustment on the control sub-model through supervision data corresponding to the group of training data, finishing the current group of training when the output result of the control sub-model is consistent with the supervision data, finishing all training data in the training data set, and finishing the training of the control sub-model.
When the similarity of the resource transmission task type corresponding to the remaining sub-network and the resource transmission task type corresponding to the first sub-network is high, the data weight in the remaining sub-network is correspondingly adjusted in the weight layer, namely the first similarity is taken as an adaptation target, the training weight layer of the resource transmission training sample set is taken as an input target, an adaptation function is established, the adaptation function is used for quantifying the importance of the resource transmission data, namely the influence degree on the migration learning, if the first similarity is higher, only the weight adjustment is carried out in the weight layer, further, the first loss function is introduced, the first similarity is analyzed on the basis, the first loss function is a function of mapping the value of the resource transmission sample set to be non-negative to represent the loss of the resource transmission sample set, the first loss function is used for analyzing the lost data, so as to record the lost data as the first loss data, the adaptation function is used for responding in the first loss data input function, namely when the first loss function is input to the first loss function, the response time is changed, namely the response time is improved, the response is improved, namely the response is realized, the response is controlled according to the response to the data in the first loss layer, and the adaptation result is recorded, the response is improved, and the response is achieved, and the response is improved, and the response is achieved.
Step S600: and controlling the resource transmission of the first computer cluster according to the plurality of control sub-models.
Specifically, according to the method, a plurality of control sub-models which are output by migration learning of a plurality of split sub-networks and respectively correspond to the split sub-networks are used as the basis of resource transmission control, the resource transmission control among the first computer nodes in the first computer network is performed on the first computer cluster in the target enterprise, namely, the resource transmission task types required to be transmitted are matched to the corresponding control sub-models, the resource transmission path, the resource packet size, the resource storage type, the resource storage queue and the resource transmission time are performed on the resource sending end of the first computer node and the resource receiving end of the first computer node in the matched control sub-models, and the convergence index of the identification resource delay is extracted on the basis, so that the intelligent control for realizing the more accurate resource transmission in the target enterprise in the later period is achieved.
In summary, the intelligent control method based on machine learning provided by the embodiment of the application at least includes the following technical effects, so that reasonable intelligent control on resource transmission based on machine learning is realized, and further resource transmission efficiency is improved.
Example two
Based on the same inventive concept as the intelligent control method based on machine learning in the foregoing embodiment, as shown in fig. 5, the present application provides an intelligent control system based on machine learning, the system comprising:
the cluster acquisition module 1 is used for acquiring a first computer cluster of a target enterprise;
the network acquisition module 2 is used for generating a first computer network according to the connection relation of the first computer cluster;
the characteristic identification module 3 is used for acquiring a historical resource transmission path of the first computer network, carrying out characteristic identification on the historical resource transmission path and acquiring path distribution characteristics;
the splitting module 4 is configured to split the first computer network according to the path distribution characteristics, so as to generate a plurality of split sub-networks;
the migration learning module 5 is used for performing migration learning on the plurality of split sub-networks and outputting a plurality of control sub-models respectively corresponding to the plurality of split sub-networks;
and the control module 6 is used for controlling the resource transmission of the first computer cluster according to the plurality of control submodels.
Further, the system further comprises:
the distribution feature module is used for acquiring the path distribution feature, wherein the path distribution feature comprises the number of path transmission objects, the number of path transmission target objects and path transmission frequency;
the density identification module is used for carrying out density identification on each node in the first computer network according to the number of the path transmission objects, the number of the path transmission target objects and the path transmission frequency to obtain a density identification index;
the first network splitting module is used for splitting the first computer network according to the density identification index.
Further, the system further comprises:
the identification module is used for identifying the density identification index, acquiring nodes which are larger than or equal to a preset density identification index and determining a plurality of clustering centers;
the clustering module is used for clustering the first computer network by the plurality of clustering centers to obtain a clustering result;
the edge node module is used for determining a split node set according to the edge nodes of the adjacent clusters in the clustering result;
and the second network splitting module is used for splitting the first computer network according to the splitting node set.
Further, the system further comprises:
the sub-network module is used for determining a first sub-network according to the plurality of split sub-networks;
the first task type module is used for generating a resource transmission training sample set and a resource transmission training test sample set according to the resource transmission task type corresponding to the first sub-network, wherein each sample set comprises a resource packet size, a resource storage type, a resource storage queue, a resource transmission time and a convergence index for identifying resource delay;
and the scheduling control module is used for performing model training by using the resource transmission training sample set and the resource transmission training test sample set, outputting a control sub-model corresponding to the first sub-network and performing scheduling control on the resource transmission task in the first sub-network.
Further, the system further comprises:
the second task type module is used for acquiring resource transmission task types corresponding to the rest sub-networks in the plurality of split sub-networks;
the comparison module is used for comparing the resource transmission task type corresponding to the residual sub-network with the resource transmission task type corresponding to the first sub-network to obtain a first similarity;
and the coefficient adjusting module is used for controlling the transfer learning weight layers of the resource transmission training sample set and the resource transmission training test sample set to carry out coefficient adjustment according to the first similarity and outputting a plurality of control sub-models corresponding to the plurality of split sub-networks.
Further, the system further comprises:
the hierarchy building module is used for obtaining a difference sub-network with similarity smaller than preset similarity and building a migration filter layer according to the difference sub-network;
the filtering module is used for filtering the sample set for transfer learning according to the transfer filter layer to obtain a transferable sample set;
and the training module is used for training the difference sub-network according to the movable sample set and the newly added sample set and outputting a control sub-model corresponding to the difference sub-network.
Further, the system further comprises:
the function building module is used for taking the first similarity as an adaptation target, taking a training weight layer of the resource transmission training sample set as an input target and building an adaptation function;
the similarity analysis module is used for introducing a first loss function to analyze the first similarity to obtain first loss data;
and the response module is used for inputting the first loss data into the adaptive function to respond, and obtaining a first response result, wherein the first response result comprises a weight layer with the coefficient adjusted.
The foregoing detailed description of the intelligent control method based on machine learning will be clear to those skilled in the art, and the intelligent control system based on machine learning in this embodiment is relatively simple for the device disclosed in the embodiment, and the relevant points refer to the description of the method section because it corresponds to the method disclosed in the embodiment.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (8)
1. An intelligent control method based on machine learning, which is characterized by comprising the following steps:
acquiring a first computer cluster of a target enterprise;
generating a first computer network according to the connection relation of the first computer cluster;
collecting a historical resource transmission path of the first computer network, and carrying out feature recognition on the historical resource transmission path to obtain path distribution features;
splitting the first computer network according to the path distribution characteristics to generate a plurality of split sub-networks;
performing migration learning on the plurality of split sub-networks, and outputting a plurality of control sub-models respectively corresponding to the plurality of split sub-networks;
and controlling the resource transmission of the first computer cluster according to the plurality of control sub-models.
2. The method of claim 1, wherein the first computer network is split according to the path distribution characteristics, the method further comprising:
the path distribution characteristics are obtained, wherein the path distribution characteristics comprise the number of path transmission objects, the number of path transmission target objects and the path transmission frequency;
performing density identification on each node in the first computer network according to the number of path transmission objects, the number of path transmission target objects and the path transmission frequency to obtain a density identification index;
splitting the first computer network according to the density identification index.
3. The method of claim 2, wherein the method further comprises:
identifying the density identification index, obtaining nodes which are larger than or equal to a preset density identification index, and determining a plurality of clustering centers;
clustering the first computer network by using the plurality of clustering centers to obtain a clustering result;
determining a split node set according to edge nodes of adjacent clusters in the clustering result;
splitting the first computer network according to the splitting node set.
4. The method of claim 1, wherein the plurality of split molecular networks are subject to transfer learning, the method further comprising:
determining a first sub-network according to the plurality of split sub-networks;
generating a resource transmission training sample set and a resource transmission training test sample set according to the resource transmission task type corresponding to the first sub-network, wherein each sample set comprises a resource packet size, a resource storage type, a resource storage queue, a resource transmission time and a convergence index for identifying resource delay;
and performing model training by using the resource transmission training sample set and the resource transmission training test sample set, and outputting a control sub-model corresponding to the first sub-network for scheduling and controlling the resource transmission task in the first sub-network.
5. The method of claim 4, wherein the method further comprises:
acquiring resource transmission task types corresponding to the rest sub-networks in the plurality of split sub-networks;
comparing the resource transmission task type corresponding to the residual sub-network with the resource transmission task type corresponding to the first sub-network to obtain a first similarity;
and according to the first similarity, controlling the transfer learning weight layers of the resource transmission training sample set and the resource transmission training test sample set to carry out coefficient adjustment, and outputting a plurality of control sub-models corresponding to the plurality of split sub-networks.
6. The method of claim 5, wherein the method further comprises:
obtaining a difference sub-network with similarity smaller than preset similarity, and establishing a migration filter layer according to the difference sub-network;
filtering the sample set for transfer learning according to the transfer filter layer to obtain a transferable sample set;
training the difference sub-network according to the movable sample set and the newly added sample set, and outputting a control sub-model corresponding to the difference sub-network.
7. The method of claim 5, wherein the method further comprises:
taking the first similarity as an adaptation target, taking a training weight layer of the resource transmission training sample set as an input target, and establishing an adaptation function;
introducing a first loss function to analyze the first similarity to obtain first loss data;
and inputting the first loss data into the adaptive function to respond, and obtaining a first response result, wherein the first response result comprises a weight layer with the coefficient adjusted.
8. An intelligent control system based on machine learning, the system comprising:
the cluster acquisition module is used for acquiring a first computer cluster of a target enterprise;
the network acquisition module is used for generating a first computer network according to the connection relation of the first computer cluster;
the characteristic identification module is used for acquiring a historical resource transmission path of the first computer network, carrying out characteristic identification on the historical resource transmission path and acquiring path distribution characteristics;
the splitting module is used for splitting the first computer network according to the path distribution characteristics to generate a plurality of splitting sub-networks;
the migration learning module is used for performing migration learning on the plurality of split sub-networks and outputting a plurality of control sub-models respectively corresponding to the plurality of split sub-networks;
and the control module is used for controlling the resource transmission of the first computer cluster according to the plurality of control submodels.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310691791.6A CN116436802B (en) | 2023-06-13 | 2023-06-13 | Intelligent control method and system based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310691791.6A CN116436802B (en) | 2023-06-13 | 2023-06-13 | Intelligent control method and system based on machine learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116436802A true CN116436802A (en) | 2023-07-14 |
| CN116436802B CN116436802B (en) | 2023-08-18 |
Family
ID=87085841
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310691791.6A Active CN116436802B (en) | 2023-06-13 | 2023-06-13 | Intelligent control method and system based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116436802B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117196996A (en) * | 2023-10-17 | 2023-12-08 | 山东鸿业信息科技有限公司 | Interface-free interaction management method and system for data resources |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130322740A1 (en) * | 2012-05-31 | 2013-12-05 | Lihui Chen | Method of Automatically Training a Classifier Hierarchy by Dynamic Grouping the Training Samples |
| CN108848561A (en) * | 2018-04-11 | 2018-11-20 | 湖北工业大学 | A kind of isomery cellular network combined optimization method based on deeply study |
| CN109803269A (en) * | 2017-11-17 | 2019-05-24 | 华为技术有限公司 | Network-building method and device, method for network access and user equipment |
| US20200111030A1 (en) * | 2018-10-05 | 2020-04-09 | Cisco Technology, Inc. | Distributed random forest training with a predictor trained to balance tasks |
| CN111694656A (en) * | 2020-04-22 | 2020-09-22 | 北京大学 | Cluster resource scheduling method and system based on multi-agent deep reinforcement learning |
| US20210026707A1 (en) * | 2019-07-24 | 2021-01-28 | Red Hat, Inc. | Predictive analysis for migration schedulers |
| US20210190882A1 (en) * | 2019-12-10 | 2021-06-24 | Wuhan University | Transformer failure identification and location diagnosis method based on multi-stage transfer learning |
| CN113496061A (en) * | 2021-03-03 | 2021-10-12 | 安徽大学 | SOM network clustering electromechanical device bearing fault analysis method based on transfer learning and manifold distance |
| CN115759295A (en) * | 2022-11-14 | 2023-03-07 | 成都理工大学 | Collaborative training method and device based on longitudinal federal learning and storage medium |
| CN115907038A (en) * | 2022-09-09 | 2023-04-04 | 南开大学 | Multivariate control decision-making method based on federated split learning framework |
| CN116151348A (en) * | 2021-11-22 | 2023-05-23 | 华为技术有限公司 | Model training method and related device |
-
2023
- 2023-06-13 CN CN202310691791.6A patent/CN116436802B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130322740A1 (en) * | 2012-05-31 | 2013-12-05 | Lihui Chen | Method of Automatically Training a Classifier Hierarchy by Dynamic Grouping the Training Samples |
| CN109803269A (en) * | 2017-11-17 | 2019-05-24 | 华为技术有限公司 | Network-building method and device, method for network access and user equipment |
| CN108848561A (en) * | 2018-04-11 | 2018-11-20 | 湖北工业大学 | A kind of isomery cellular network combined optimization method based on deeply study |
| US20200111030A1 (en) * | 2018-10-05 | 2020-04-09 | Cisco Technology, Inc. | Distributed random forest training with a predictor trained to balance tasks |
| US20210026707A1 (en) * | 2019-07-24 | 2021-01-28 | Red Hat, Inc. | Predictive analysis for migration schedulers |
| US20210190882A1 (en) * | 2019-12-10 | 2021-06-24 | Wuhan University | Transformer failure identification and location diagnosis method based on multi-stage transfer learning |
| CN111694656A (en) * | 2020-04-22 | 2020-09-22 | 北京大学 | Cluster resource scheduling method and system based on multi-agent deep reinforcement learning |
| CN113496061A (en) * | 2021-03-03 | 2021-10-12 | 安徽大学 | SOM network clustering electromechanical device bearing fault analysis method based on transfer learning and manifold distance |
| CN116151348A (en) * | 2021-11-22 | 2023-05-23 | 华为技术有限公司 | Model training method and related device |
| CN115907038A (en) * | 2022-09-09 | 2023-04-04 | 南开大学 | Multivariate control decision-making method based on federated split learning framework |
| CN115759295A (en) * | 2022-11-14 | 2023-03-07 | 成都理工大学 | Collaborative training method and device based on longitudinal federal learning and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 夏胜平;刘建军;袁振涛;虞华;张乐锋;郁文贤;: "基于集群的增量分布式RSOM聚类方法", 电子学报, no. 03 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117196996A (en) * | 2023-10-17 | 2023-12-08 | 山东鸿业信息科技有限公司 | Interface-free interaction management method and system for data resources |
| CN117196996B (en) * | 2023-10-17 | 2024-06-04 | 山东鸿业信息科技有限公司 | Interface-free interaction management method and system for data resources |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116436802B (en) | 2023-08-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116436802B (en) | Intelligent control method and system based on machine learning | |
| CN109617888B (en) | Abnormal flow detection method and system based on neural network | |
| CN104102700A (en) | Categorizing method oriented to Internet unbalanced application flow | |
| WO2011130957A1 (en) | Method and apparatus for online distinguishing transmission control protocol traffic by using data flow head characteristics | |
| CN111756600B (en) | Multi-communication system and method for realizing multiple switch test machines | |
| CN111953669A (en) | Tor flow tracing and application type identification method and system suitable for SDN | |
| CN110034970A (en) | The network equipment distinguishes method of discrimination and device | |
| Zhao et al. | A few-shot learning based approach to IoT traffic classification | |
| CN116684877A (en) | GYAC-LSTM-based 5G network traffic anomaly detection method and system | |
| CN104965846A (en) | Virtual human establishing method on MapReduce platform | |
| CN113449682B (en) | Method for identifying radio frequency fingerprints in civil aviation field based on dynamic fusion model | |
| CN112235254B (en) | Rapid identification method for Tor network bridge in high-speed backbone network | |
| CN111669396A (en) | Self-learning security defense method and system for software-defined Internet of things | |
| Bebensee et al. | Applying dynamic Bayesian networks for automated modeling in ArchiMate: a realization study | |
| CN113726809B (en) | Internet of things equipment identification method based on flow data | |
| CN113537053B (en) | Method for constructing radio frequency fingerprint identification model in civil aviation field | |
| CN114529096A (en) | Social network link prediction method and system based on ternary closure graph embedding | |
| CN113794653A (en) | High-speed network traffic classification method based on sampling data flow | |
| CN111586052B (en) | Multi-level-based crowd sourcing contract abnormal transaction identification method and identification system | |
| CN115002031A (en) | Unbalanced data distribution-based federated learning network traffic classification model training method, model and classification method | |
| Zhenxiang et al. | Research of P2P traffic comprehensive identification method | |
| Erokhin et al. | The Dataset Features Selection for Detecting and Classifying Network Attacks | |
| CN110544182A (en) | Power distribution communication network fusion control method and system based on machine learning technology | |
| CN111600779A (en) | Test platform compatible with various switches | |
| CN114726800B (en) | Rule type application flow classification method and system based on model interpretation |
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 |