CN111461184A - XGB multi-dimensional operation and maintenance data anomaly detection method based on multivariate feature matrix - Google Patents

XGB multi-dimensional operation and maintenance data anomaly detection method based on multivariate feature matrix Download PDF

Info

Publication number
CN111461184A
CN111461184A CN202010194474.XA CN202010194474A CN111461184A CN 111461184 A CN111461184 A CN 111461184A CN 202010194474 A CN202010194474 A CN 202010194474A CN 111461184 A CN111461184 A CN 111461184A
Authority
CN
China
Prior art keywords
data
maintenance data
xgb
training
sample
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.)
Withdrawn
Application number
CN202010194474.XA
Other languages
Chinese (zh)
Inventor
朱耀琴
韩仁松
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.)
Nanjing University of Science and Technology
Original Assignee
Nanjing University of Science and Technology
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 Nanjing University of Science and Technology filed Critical Nanjing University of Science and Technology
Priority to CN202010194474.XA priority Critical patent/CN111461184A/en
Publication of CN111461184A publication Critical patent/CN111461184A/en
Withdrawn legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/20—Ensemble learning

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Software Systems (AREA)
  • Medical Informatics (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention discloses an XGB multi-dimensional operation and maintenance data anomaly detection method based on a multivariate feature matrix, which comprises the following steps of: constructing a METRIC table for acquiring data according to the product service data; acquiring operation and maintenance data of product service according to a METRIC table, and constructing a training sample; constructing a data set according to the training samples; training an XGB model by using a data set; and carrying out anomaly detection on the operation and maintenance data to be detected by using the trained XGB model. The invention firstly proposes to carry out anomaly detection on the multidimensional operation and maintenance data, designs the METRIC table structure through the multidimensional operation and maintenance data, achieves the directionality target of data monitoring and acquisition, fully considers the integrity of the service, has complete structural logic, and is more favorable for product supervision and data acquisition. In addition, the XGB model is adopted for classification detection, and a multivariate feature matrix is constructed to train the XGB model by comprehensively extracting data features, so that the reliability of the XGB model is improved, and the detection precision is further improved.

Description

XGB multi-dimensional operation and maintenance data anomaly detection method based on multivariate feature matrix
Technical Field
The invention belongs to the field of intelligent operation and maintenance abnormity detection, and particularly relates to an XGB multi-dimensional operation and maintenance data abnormity detection method based on a multivariate feature matrix.
Background
In recent years, the application of high and new technologies such as mobile internet, cloud computing, big data and the like is continuously and maturely evolving, and enterprises are required to follow the era wave, complete digital transformation and avoid being eliminated by the era. The enterprise carries out digital transformation, which brings about a great development opportunity for the enterprise and also faces greater challenges, and the IT environment and fault handling become unprecedented and complicated. In order to provide more efficient and rapid emergency service capability, more and more enterprises begin to deploy intelligent operation and maintenance. In popular terms, intelligent operation and maintenance are deployed to ensure normal operation of large-scale complex systems and product services, wherein timely discovery of system or service abnormalities is an important link in operation and maintenance. During the operation process of the system and the service process of the product service, a plurality of operation and maintenance data describing the operation state of each part can be collected, such as CPU utilization rate, product service processing capacity and the like. The anomaly detection aims to automatically find the anomalies in the operation and maintenance data through a certain method and provide decision basis for subsequent alarm, automatic loss stopping, root cause analysis and the like.
The traditional abnormality detection method of enterprises is a threshold-based abnormality detection method, and the main idea of the method is that an area higher than a high threshold or lower than a low threshold is an abnormal area. The oldest method is an anomaly detection method in which a static threshold is directly set, and this detection method causes a large number of false alarms or non-alarms of abnormal conditions. The rolling time window-based moving average anomaly detection method can realize the idea of dynamic threshold, and in a scene with small fluctuation of sequence values along with time, the moving average value is calculated by using the moving average method and is set as an accurate value which should exist at the current moment. And then, comparing with the true value of the time, and if the difference value exceeds a certain threshold value, judging that the time node is abnormal. Due to the fact that different operation and maintenance data threshold values are set to have great differences, the range of a part of operation and maintenance data value ranges is great, if different threshold values are set to process the problem of abnormal detection of multi-dimensional operation and maintenance data, detection time is increased greatly, and if interference between the operation and maintenance data (for example, the success rate and the failure rate of product business service are mutually opposite), a more serious false alarm situation is caused. Therefore, the anomaly detection method based on the threshold is not suitable for anomaly detection of multidimensional operation and maintenance data.
The XGB model adopts a boosting integrated learning method, a plurality of weak classifiers are iterated into a strong classifier, the strong classifier can discover the correlation between data and learn rules from a large amount of data in a centralized manner so as to obtain a more accurate classification effect, the XGB model can play an important role in the results, in the problem of storage and marketing prediction, physical event classification, webpage text classification, customer behavior prediction, click rate prediction, motivation detection, product classification and the like, in the problem of XGB improved Supervised Outvier detection with unsaved recovery L earning, the problem that the time sequence anomaly may be well hidden in certain subspaces or can only be recognized under specific assumptions, a label is utilized to construct a final monitoring model, the final monitoring and verification result is realized, the multiple-dimensional humidity detection results are obtained by an XGB model, the multiple-classification problem of the XGB model is generated based on the short-term rain load reduction, the three-dimensional humidity detection and the like, and the multiple-dimensional humidity detection and the three-dimensional humidity detection model is provided for the research on the short-term rainfall load and the super-dimensional humidity detection and the three-dimensional humidity detection and the super-dimensional humidity detection and the three-dimensional humidity research.
Based on the facts, a large amount of operation and maintenance data exist in the product service process, and the XGB model can have a very good effect on multi-dimensional time series analysis.
Disclosure of Invention
The invention aims to provide an XGB multi-dimensional operation and maintenance data anomaly detection method based on a multivariate feature matrix.
The technical solution for realizing the purpose of the invention is as follows: an XGB multi-dimensional operation and maintenance data anomaly detection method based on a multivariate feature matrix comprises the following steps:
step 1, constructing a METRIC table for acquiring data according to product service data;
step 2, acquiring operation and maintenance data of product service according to a METRIC table, and constructing a training sample;
step 3, constructing a data set according to the training samples;
step 4, training the XGB model by using the data set;
and 5, carrying out anomaly detection on the operation and maintenance data to be detected by using the trained XGB model.
Further, the abscissa of the METRIC table in step 1 represents all product service data names, and the ordinate represents time nodes.
Further, the step 2 of collecting operation and maintenance data of the product service according to the METRIC table and constructing a training sample includes the following specific processes:
step 2-1, setting a time window length standard for collecting operation and maintenance data, wherein the time window length standard is expressed by a tuple as follows:
(WindowLength，TimeUnit，TimeInterval，TimePeriod，TimeLength)
in the formula, Window L ength represents the length of a Time Window, TimeUnit represents the Time unit for collecting the operation and maintenance data, TimeInterval represents the Time interval for collecting the operation and maintenance data, TimePeriod represents the Time period of the operation and maintenance data, Time L ength represents the length of the Time period for collecting the operation and maintenance data, wherein,
Figure BDA0002417104020000031
step 2-2, acquiring operation and maintenance data of product service according to a time window length standard and a METRIC table, and constructing a plurality of training samples; each training sample corresponds to a METRIC table containing operation and maintenance data;
and 2-3, marking each training sample with a sample, dividing a normal sample and an abnormal sample to be used as a negative sample and a positive sample respectively.
Further, the step 3 of constructing a data set according to the training samples includes:
step 3-1, arranging all training samples according to the time node sequence to form a training sample sequence;
and 3-2, extracting training samples from the training sample sequence by using a sliding window-based data set training method, and constructing a data set.
Further, in step 4, the XGB model is trained by using the data set, and the specific process includes:
step 4-1, performing undersampling treatment on negative samples in the data set;
step 4-2, performing normalization processing on all operation and maintenance data of each sample in the data set;
4-3, extracting data characteristics of each type of operation and maintenance data in each sample, wherein the data characteristics comprise statistical characteristics, classification characteristics and sequence characteristics; wherein, the sequence features are extracted based on a time sequence analysis method;
step 4-4, aiming at each sample, constructing a multivariate feature matrix MF:
Figure BDA0002417104020000032
in the formula (I), the compound is shown in the specification,
Figure BDA0002417104020000034
the statistical feature vector of the ith dimension operation and maintenance data,
Figure BDA0002417104020000035
the classification feature vector of the ith dimension operation and maintenance data,
Figure BDA0002417104020000036
is the sequence feature vector of the ith dimension operation and maintenance data, i ∈ [1, n]；
Step 4-5, longitudinally processing the MF to obtain a new multivariate feature matrix MF':
Figure BDA0002417104020000037
in the formula (I), the compound is shown in the specification,
Figure BDA0002417104020000033
Figure BDA0002417104020000041
4-6, initializing training parameters of the XGB model;
step 4-7, respectively distributing weight omega for statistical feature, classification feature and sequence feature1、ω2、ω3，ω1+ω2+ω3And (5) inputting the multivariate feature matrix corresponding to all samples into the XGB model for training, updating the parameters of the XGB model until a preset training end condition is reached, and outputting a final XGB model and the parameters thereof.
Compared with the prior art, the invention has the following remarkable advantages: 1) the method has the advantages that the abnormity detection of the multi-dimensional operation and maintenance data is firstly proposed, and the high-precision and high-efficiency detection is realized; 2) under the condition of complex application of software and hardware matching of products in various current production scenes, a METRIC table structure is designed through multi-dimensional operation and maintenance data, the directionality target of data monitoring and acquisition is achieved, the integrity of services is fully considered, the structural logic is complete, and product supervision and data acquisition are facilitated; 3) the XGB model is adopted, so that the effect from a weak classifier to a strong classifier can be realized, and the precision of classification detection is improved; 4) the data used for model training is acquired through the sliding window, so that the autonomy is improved, the data quantity needing to be processed is reduced, and the training and detection efficiency is improved; 5) the problem of unbalance of a data set used for model training is solved, and the regularization result of the training is improved; 6) by comprehensively extracting data features and constructing a multi-element feature matrix to train the model, the reliability of the XGB model is improved, and the detection precision is further improved.
The present invention is described in further detail below with reference to the attached drawing figures.
Drawings
FIG. 1 is a flowchart of an XGB multidimensional operation and maintenance data anomaly detection method based on a multivariate feature matrix.
Detailed Description
With reference to fig. 1, the invention provides an XGB multidimensional operation and maintenance data anomaly detection method based on a multivariate feature matrix, which includes the following steps:
step 1, constructing a METRIC table for acquiring data according to product service data;
here, the product service data includes the underlying data: CPU utilization, etc., middle layer data: number of inter-service calls, etc., top level data: the number of business services, etc.;
step 2, acquiring operation and maintenance data of product service according to a METRIC table, and constructing a training sample;
step 3, constructing a data set according to the training samples;
step 4, training the XGB model by using the data set;
and 5, carrying out anomaly detection on the operation and maintenance data to be detected by using the trained XGB model.
Further, in one embodiment, the abscissa of the METRIC table in step 1 represents all product service data names, and the ordinate represents time nodes.
Further, in one embodiment, the step 2 acquires the operation and maintenance data of the product service according to the METRIC table, and constructs the training sample, and the specific process includes:
step 2-1, setting a time window length standard for collecting operation and maintenance data, wherein the time window length standard is expressed by a tuple as follows:
(WindowLength，TimeUnit，TimeInterval，TimePeriod，TimeLength)
in the formula, Window L ength represents the length of a Time Window, TimeUnit represents the Time unit for collecting the operation and maintenance data, TimeInterval represents the Time interval for collecting the operation and maintenance data, TimePeriod represents the Time period of the operation and maintenance data, Time L ength represents the length of the Time period for collecting the operation and maintenance data, wherein,
Figure BDA0002417104020000051
here, the Time L ength period is set with reference to the TimeUnit unit and the TimePeriod size.
Step 2-2, acquiring operation and maintenance data of product service according to a time window length standard and a METRIC table, and constructing a plurality of training samples; each training sample corresponds to a METRIC table containing operation and maintenance data;
and 2-3, marking each training sample with a sample, dividing a normal sample and an abnormal sample to be used as a negative sample and a positive sample respectively.
Further, in one embodiment, the step 3 constructs a data set according to the training samples, and the specific process includes:
step 3-1, arranging all training samples according to the time node sequence to form a training sample sequence;
and 3-2, extracting training samples from the training sample sequence by using a sliding window-based data set training method, and constructing a data set.
In the method, the data in any time period can be selected to construct the data set by adopting the sliding window according to actual requirements, so that the autonomy is stronger, and the data volume needing to be processed is reduced.
Further, in one embodiment, the step 4 trains the XGB model by using the data set, and the specific process includes:
step 4-1, performing undersampling treatment on negative samples in the data set;
step 4-2, performing normalization processing on all operation and maintenance data of each sample in the data set, so that each data is located between 0 and 1;
4-3, extracting data characteristics of each type of operation and maintenance data in each sample, wherein the data characteristics comprise statistical characteristics, classification characteristics and sequence characteristics; wherein, the sequence features are extracted based on a time sequence analysis method;
here, the statistical features include: maximum, minimum, mean, median, repeat, standard deviation, skewness, kurtosis, identity, ring ratio, autocorrelation coefficient, coefficient of variation, and the like;
the classification features include: the number of the points which are more than the mean value and less than the mean value, the entropy value, the wavelet analysis value, the minimum value position, the maximum value position and the like;
step 4-4, aiming at each sample, constructing a multivariate feature matrix MF:
Figure BDA0002417104020000061
in the formula (I), the compound is shown in the specification,
Figure BDA0002417104020000064
the statistical feature vector of the ith dimension operation and maintenance data,
Figure BDA0002417104020000065
the classification feature vector of the ith dimension operation and maintenance data,
Figure BDA0002417104020000066
is the sequence feature vector of the ith dimension operation and maintenance data, i ∈ [1, n]；
Step 4-5, longitudinally processing the MF to obtain a new multivariate feature matrix MF':
Figure BDA0002417104020000067
in the formula (I), the compound is shown in the specification,
Figure BDA0002417104020000062
Figure BDA0002417104020000063
4-6, initializing training parameters of the XGB model;
step 4-7, respectively distributing weight omega for statistical feature, classification feature and sequence feature1、ω2、ω3，ω1+ω2+ω3And (5) inputting the multivariate feature matrix corresponding to all samples into the XGB model for training, updating the parameters of the XGB model until a preset training end condition is reached, and outputting a final XGB model and the parameters thereof.
The invention firstly proposes to carry out anomaly detection on the multidimensional operation and maintenance data, designs the METRIC table structure through the multidimensional operation and maintenance data, achieves the directionality target of data monitoring and acquisition, fully considers the integrity of the service, has complete structural logic, and is more favorable for product supervision and data acquisition. In addition, the XGB model is adopted for classification detection, and a multivariate feature matrix is constructed to train the XGB model by comprehensively extracting data features, so that the reliability of the XGB model is improved, and the detection precision is further improved.

Claims (5)

1. An XGB multi-dimensional operation and maintenance data anomaly detection method based on a multivariate feature matrix is characterized by comprising the following steps:
step 1, constructing a METRIC table for acquiring data according to product service data;
step 2, acquiring operation and maintenance data of product service according to a METRIC table, and constructing a training sample;
step 3, constructing a data set according to the training samples;
step 4, training the XGB model by using the data set;
and 5, carrying out anomaly detection on the operation and maintenance data to be detected by using the trained XGB model.
2. The XGB multidimensional operation and maintenance data anomaly detection method based on the multivariate feature matrix as claimed in claim 1, wherein the abscissa of the METRIC table in step 1 represents all product service data names, and the ordinate represents time nodes.
3. The XGB multi-dimensional operation and maintenance data anomaly detection method based on the multivariate feature matrix as claimed in claim 1 or 2, wherein the step 2 of acquiring operation and maintenance data of product services according to a METRIC table and constructing training samples comprises the following specific processes:
step 2-1, setting a time window length standard for collecting operation and maintenance data, wherein the time window length standard is expressed by a tuple as follows:
(WindowLength，TimeUnit，TimeInterval，TimePeriod，TimeLength)
in the formula, Window L ength represents the length of a Time Window, TimeUnit represents the Time unit for collecting the operation and maintenance data, TimeInterval represents the Time interval for collecting the operation and maintenance data, TimePeriod represents the Time period of the operation and maintenance data, Time L ength represents the length of the Time period for collecting the operation and maintenance data, wherein,
Figure FDA0002417104010000011
step 2-2, acquiring operation and maintenance data of product service according to a time window length standard and a METRIC table, and constructing a plurality of training samples; each training sample corresponds to a METRIC table containing operation and maintenance data;
and 2-3, marking each training sample with a sample, dividing a normal sample and an abnormal sample to be used as a negative sample and a positive sample respectively.
4. The XGB multidimensional operation and maintenance data anomaly detection method based on the multivariate feature matrix as claimed in claim 3, wherein the step 3 of constructing the data set according to the training samples comprises the following specific processes:
step 3-1, arranging all training samples according to the time node sequence to form a training sample sequence;
and 3-2, extracting training samples from the training sample sequence by using a sliding window-based data set training method, and constructing a data set.
5. The XGB multidimensional operation and maintenance data anomaly detection method based on the multivariate feature matrix as claimed in claim 4, wherein the step 4 of training the XGB model by using the data set comprises the following specific processes:
step 4-1, performing undersampling treatment on negative samples in the data set;
step 4-2, performing normalization processing on all operation and maintenance data of each sample in the data set;
4-3, extracting data characteristics of each type of operation and maintenance data in each sample, wherein the data characteristics comprise statistical characteristics, classification characteristics and sequence characteristics; wherein, the sequence features are extracted based on a time sequence analysis method;
step 4-4, aiming at each sample, constructing a multivariate feature matrix MF:
Figure FDA0002417104010000021
in the formula (I), the compound is shown in the specification,
Figure FDA0002417104010000022
the statistical feature vector of the ith dimension operation and maintenance data,
Figure FDA0002417104010000023
the classification feature vector of the ith dimension operation and maintenance data,
Figure FDA0002417104010000024
is the sequence feature vector of the ith dimension operation and maintenance data, i ∈ [1, n]；
Step 4-5, longitudinally processing the MF to obtain a new multivariate feature matrix MF':
Figure FDA0002417104010000025
in the formula (I), the compound is shown in the specification,
Figure FDA0002417104010000026
Figure FDA0002417104010000027
4-6, initializing training parameters of the XGB model;
step 4-7, respectively distributing weight omega for statistical feature, classification feature and sequence feature1、ω2、ω3，ω1+ω2+ω3And (5) inputting the multivariate feature matrix corresponding to all samples into the XGB model for training, updating the parameters of the XGB model until a preset training end condition is reached, and outputting a final XGB model and the parameters thereof.
CN202010194474.XA 2020-03-19 2020-03-19 XGB multi-dimensional operation and maintenance data anomaly detection method based on multivariate feature matrix Withdrawn CN111461184A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010194474.XA CN111461184A (en) 2020-03-19 2020-03-19 XGB multi-dimensional operation and maintenance data anomaly detection method based on multivariate feature matrix

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010194474.XA CN111461184A (en) 2020-03-19 2020-03-19 XGB multi-dimensional operation and maintenance data anomaly detection method based on multivariate feature matrix

Publications (1)

Publication Number Publication Date
CN111461184A true CN111461184A (en) 2020-07-28

Family

ID=71680857

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010194474.XA Withdrawn CN111461184A (en) 2020-03-19 2020-03-19 XGB multi-dimensional operation and maintenance data anomaly detection method based on multivariate feature matrix

Country Status (1)

Country Link
CN (1) CN111461184A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111985567A (en) * 2020-08-21 2020-11-24 河北先河环保科技股份有限公司 Automatic pollution source type identification method based on machine learning
WO2022127597A1 (en) * 2020-12-17 2022-06-23 Telefonaktiebolaget Lm Ericsson (Publ) Methods and devices for anomaly detection

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111985567A (en) * 2020-08-21 2020-11-24 河北先河环保科技股份有限公司 Automatic pollution source type identification method based on machine learning
WO2022127597A1 (en) * 2020-12-17 2022-06-23 Telefonaktiebolaget Lm Ericsson (Publ) Methods and devices for anomaly detection

Similar Documents

Publication Publication Date Title
CN108388927B (en) Small sample polarization SAR terrain classification method based on deep convolution twin network
CN110263846B (en) Fault diagnosis method based on fault data deep mining and learning
Mao et al. Online detection for bearing incipient fault based on deep transfer learning
CN110084151B (en) Video abnormal behavior discrimination method based on non-local network deep learning
CN110287983B (en) Single-classifier anomaly detection method based on maximum correlation entropy deep neural network
Ko et al. Fault classification in high-dimensional complex processes using semi-supervised deep convolutional generative models
Yin et al. Wasserstein generative adversarial network and convolutional neural network (WG-CNN) for bearing fault diagnosis
CN111476263A (en) Bearing defect identification method based on SDAE and improved GWO-SVM
CN111562108A (en) Rolling bearing intelligent fault diagnosis method based on CNN and FCMC
CN113505655B (en) Intelligent bearing fault diagnosis method for digital twin system
CN111352977B (en) Time sequence data monitoring method based on self-attention bidirectional long-short term memory network
CN104931263A (en) Bearing fault diagnosis method based on symbolic probabilistic finite state machine
CN105607631B (en) The weak fault model control limit method for building up of batch process and weak fault monitoring method
CN111352086A (en) Unknown target identification method based on deep convolutional neural network
CN113158722A (en) Rotary machine fault diagnosis method based on multi-scale deep neural network
CN111461184A (en) XGB multi-dimensional operation and maintenance data anomaly detection method based on multivariate feature matrix
CN112949821B (en) Network security situation awareness method based on dual-attention mechanism
CN102346851A (en) Image segmentation method based on NJW (Ng-Jordan-Weiss) spectral clustering mark
CN115358259A (en) Self-learning-based unsupervised cross-working-condition bearing fault diagnosis method
WO2019178930A1 (en) Fault diagnosis method for mechanical device
Wang et al. Uncertainty quantification of bearing remaining useful life based on convolutional neural network
CN116152678A (en) Marine disaster-bearing body identification method based on twin neural network under small sample condition
CN109615027B (en) Intelligent prediction method for extracting wind speed characteristics along high-speed railway
CN113780432B (en) Intelligent detection method for operation and maintenance abnormity of network information system based on reinforcement learning
Mobtahej et al. An lstm-autoencoder architecture for anomaly detection applied on compressors audio data

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
WW01 Invention patent application withdrawn after publication

Application publication date: 20200728

WW01 Invention patent application withdrawn after publication