CN112732785A - Time series data abnormity detection method, device, equipment and storage medium - Google Patents
Time series data abnormity detection method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN112732785A CN112732785A CN202011635793.6A CN202011635793A CN112732785A CN 112732785 A CN112732785 A CN 112732785A CN 202011635793 A CN202011635793 A CN 202011635793A CN 112732785 A CN112732785 A CN 112732785A
- Authority
- CN
- China
- Prior art keywords
- time sequence
- time
- matrix
- preset
- data
- 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.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 45
- 239000011159 matrix material Substances 0.000 claims abstract description 233
- 238000006243 chemical reaction Methods 0.000 claims abstract description 59
- 230000002159 abnormal effect Effects 0.000 claims abstract description 39
- 238000000034 method Methods 0.000 claims abstract description 34
- 230000006870 function Effects 0.000 claims description 31
- 238000005457 optimization Methods 0.000 claims description 28
- 230000005856 abnormality Effects 0.000 claims description 21
- 238000011176 pooling Methods 0.000 claims description 17
- 230000002123 temporal effect Effects 0.000 claims description 5
- 230000009466 transformation Effects 0.000 claims description 5
- 238000012423 maintenance Methods 0.000 abstract description 4
- NJPPVKZQTLUDBO-UHFFFAOYSA-N novaluron Chemical compound C1=C(Cl)C(OC(F)(F)C(OC(F)(F)F)F)=CC=C1NC(=O)NC(=O)C1=C(F)C=CC=C1F NJPPVKZQTLUDBO-UHFFFAOYSA-N 0.000 abstract 1
- 230000000875 corresponding effect Effects 0.000 description 15
- 238000010586 diagram Methods 0.000 description 7
- 238000012549 training Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 5
- 238000012544 monitoring process Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- 230000006835 compression Effects 0.000 description 2
- 238000007906 compression Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000002085 persistent effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
- G06F16/2474—Sequence data queries, e.g. querying versioned data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/15—Correlation function computation including computation of convolution operations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Computing Systems (AREA)
- Algebra (AREA)
- Computational Linguistics (AREA)
- Probability & Statistics with Applications (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Operations Research (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Fuzzy Systems (AREA)
- Debugging And Monitoring (AREA)
Abstract
The invention relates to the field of pedestal operation and maintenance, and discloses a method, a device, equipment and a storage medium for detecting time sequence data abnormity. The method comprises the following steps: collecting time sequence data to be detected, and extracting sub time sequence data with fixed time length through a preset sliding time window; calling a preset gram angular field algorithm, and performing 2D conversion on the sub-time sequence data to obtain a first time sequence angular field matrix; inputting the first time sequence angle field matrix into a preset time sequence reconstruction model for time sequence data reconstruction to obtain a second time sequence angle field matrix; calling a preset loss function, calculating loss values of the first time sequence angle field matrix and the second time sequence angle field matrix, and judging whether the loss values exceed a preset loss threshold value; and if so, determining that the sub time sequence data is abnormal. The invention utilizes the dependency information between time sequences to reconstruct the time sequences, thereby improving the accuracy of the abnormal detection.
Description
Technical Field
The invention relates to the field of operation and maintenance of base frames, in particular to a method, a device, equipment and a storage medium for detecting time sequence data abnormity.
Background
The time sequence is a sequence formed by arranging numerical values of a certain statistical index in a certain field at different moments according to a time sequence. The time series data abnormity detection is always a relatively concerned problem in academic and industrial fields, is commonly used in operation and maintenance monitoring, traffic management and other scenes, detects the time series of monitoring indexes, and timely alarms for the found abnormity to prompt a worker to pay attention to the processing.
The conventional time series detection is to count important information in the time data, such as average number, median number, etc., by a sliding window, and then predict the value of the next time point according to the key information. And judging whether the time series is abnormal or not according to the predicted value and the actual value in the time series. In the conventional time series detection method, when the time series is long, a detection result has a large error.
Disclosure of Invention
The invention mainly aims to solve the technical problem of low detection accuracy caused by large detection error of time sequence data abnormity.
The invention provides a time series data abnormity detection method in a first aspect, which comprises the following steps:
collecting time sequence data to be detected, and extracting sub time sequence data with fixed time length through a preset sliding time window;
calling a preset gram angular field algorithm, and performing 2D conversion on the sub-time sequence data to obtain a first time sequence angular field matrix;
inputting the first time sequence angle field matrix into a preset time sequence reconstruction model for time sequence data reconstruction to obtain a second time sequence angle field matrix;
calling a preset loss function, calculating loss values of the first time sequence angle field matrix and the second time sequence angle field matrix, and judging whether the loss values exceed a preset loss threshold value;
and if so, determining that the sub time sequence data is abnormal.
Optionally, in a first implementation manner of the first aspect of the present invention, the time-series reconstruction model includes an encoder and a decoder, where the encoder is composed of multiple layers of one-dimensional convolutional layers and pooling layers, and the decoder is composed of multiple layers of one-dimensional deconvolution layers and convolutional layers, and before the acquiring the time-series data to be detected, the method further includes:
collecting abnormal time sequence data, and extracting abnormal time sequence data with fixed time length through the sliding time window;
calling a preset gram angular field algorithm, and performing 2D conversion on the abnormal-free sub time sequence data to obtain a third time sequence angular field matrix;
inputting the third time sequence angle field matrix into the multilayer one-dimensional convolution layer to carry out convolution operation to obtain a first characteristic matrix;
inputting the first feature matrix into the pooling layer for mean value coding to obtain a first coding matrix;
inputting the first coding matrix into the multilayer one-dimensional deconvolution layer for feature conversion to obtain a first conversion matrix;
inputting the first conversion matrix into the convolution layer for characteristic reconstruction to obtain a fourth time sequence angular field matrix;
calling a preset loss function, and calculating loss values of the third time sequence angle field matrix and the fourth time sequence angle field matrix;
and calling a preset optimization algorithm to perform parameter optimization on the encoder and the decoder according to the loss value until the encoder and the decoder are converged to obtain a time sequence reconstruction model.
Optionally, in a second implementation manner of the first aspect of the present invention, the invoking a preset optimization algorithm to perform parameter optimization on the encoder and the decoder according to the loss value until the encoder and the decoder converge to obtain a time sequence reconstruction model includes:
calling a preset optimization algorithm to calculate first moment estimation and second moment estimation of the encoder and the decoder according to the loss value;
and correcting the first moment estimation and the second moment estimation, and updating parameters through the corrected first moment estimation and second moment estimation until the encoder and the decoder converge to obtain a time sequence reconstruction model.
Optionally, in a third implementation manner of the first aspect of the present invention, the acquiring the time-series data to be detected, and extracting the sub-time-series data with a fixed time length through the sliding time window includes:
collecting time sequence data to be detected, and adjusting the window length of the sliding time window;
and acquiring a time sequence acquisition value of a fixed time length forwards at intervals of preset fixed time through the adjusted sliding time window to obtain the sub-time sequence data of the time sequence data to be detected.
Optionally, in a fourth implementation manner of the first aspect of the present invention, the invoking a preset glam angle field algorithm to perform 2D conversion on the sub-timing sequence data to obtain a first timing angle field matrix includes:
normalizing the time sequence value in the sub-time sequence data to obtain normalized time sequence data;
performing polar coordinate conversion on the normalized time sequence data to obtain polar coordinate time sequence data;
and intercepting a plurality of corresponding values of the timestamps in the polar coordinate time sequence data, and performing trigonometric function transformation on the corresponding values of the timestamps to obtain a first time sequence angular field matrix.
Optionally, in a fifth implementation manner of the first aspect of the present invention, the inputting the first time sequence angle field matrix into a preset time sequence reconstruction model for time sequence data reconstruction, and obtaining a second time sequence angle field matrix includes:
inputting the first time sequence angle field matrix into the multilayer one-dimensional convolution layer for convolution operation to obtain a second characteristic matrix;
inputting the second feature matrix into the pooling layer for mean value coding to obtain a second coding matrix;
inputting the second coding matrix into the multilayer one-dimensional deconvolution layer for feature conversion to obtain a second conversion matrix;
and inputting the second conversion matrix into the convolution layer for characteristic reconstruction to obtain a second time sequence angle field matrix.
Optionally, in a sixth implementation manner of the first aspect of the present invention, the invoking a preset penalty function, and calculating penalty values of the first time series angle field matrix and the second time series angle field matrix includes:
calculating matrix element errors of corresponding positions in the first time sequence angle field matrix and the second time sequence angle field matrix to obtain a plurality of matrix element error values;
calling a preset loss function, and calculating the sum of squares of the error values of the matrix elements;
and calculating the average value of the square sum of the error values of the elements of each matrix to obtain the loss values of the first time sequence angle field matrix and the second time sequence angle field matrix.
A second aspect of the present invention provides a time series data abnormality detection apparatus including:
the acquisition module is used for acquiring time sequence data to be detected and extracting sub-time sequence data with fixed time length through a preset sliding time window;
the 2D conversion module is used for calling a preset Graham angular field algorithm and carrying out 2D conversion on the sub-time sequence data to obtain a first time sequence angular field matrix;
the time sequence reconstruction module is used for inputting the first time sequence angle field matrix into a preset time sequence reconstruction model to reconstruct time sequence data to obtain a second time sequence angle field matrix;
the judging module is used for calling a preset loss function, calculating the loss values of the first time sequence angle field matrix and the second time sequence angle field matrix and judging whether the loss values exceed a preset loss threshold value or not;
and the output module is used for determining that the sub-time sequence data is abnormal if the loss value exceeds the loss threshold value.
Optionally, in a first implementation manner of the second aspect of the present invention, the time-series reconstruction model includes an encoder and a decoder, where the encoder is composed of multiple layers of one-dimensional convolutional layers and pooling layers, and the decoder is composed of multiple layers of one-dimensional deconvolution layers and convolutional layers, and the time-series data anomaly detection apparatus further includes:
the sample processing module is used for acquiring abnormal time sequence data and extracting abnormal time sequence data with fixed time length through the sliding time window; calling a preset gram angular field algorithm, and performing 2D conversion on the abnormal-free sub time sequence data to obtain a third time sequence angular field matrix;
the model training module is used for inputting the third time sequence angle field matrix into the multilayer one-dimensional convolution layer to carry out convolution operation to obtain a first characteristic matrix; inputting the first feature matrix into the pooling layer for mean value coding to obtain a first coding matrix; inputting the first coding matrix into the multilayer one-dimensional deconvolution layer for feature conversion to obtain a first conversion matrix; inputting the first conversion matrix into the convolution layer for characteristic reconstruction to obtain a fourth time sequence angular field matrix; calling a preset loss function, and calculating loss values of the third time sequence angle field matrix and the fourth time sequence angle field matrix;
and the model optimization module is used for calling a preset optimization algorithm to carry out parameter optimization on the encoder and the decoder according to the loss value until the encoder and the decoder are converged to obtain a time sequence reconstruction model.
Optionally, in a second implementation manner of the second aspect of the present invention, the model optimization module is specifically configured to:
calling a preset optimization algorithm to calculate first moment estimation and second moment estimation of the encoder and the decoder according to the loss value;
and correcting the first moment estimation and the second moment estimation, and updating parameters through the corrected first moment estimation and second moment estimation until the encoder and the decoder converge to obtain a time sequence reconstruction model.
Optionally, in a third implementation manner of the second aspect of the present invention, the acquisition module is specifically configured to:
collecting time sequence data to be detected, and adjusting the window length of the sliding time window;
and acquiring a time sequence acquisition value of a fixed time length forwards at intervals of preset fixed time through the adjusted sliding time window to obtain the sub-time sequence data of the time sequence data to be detected.
Optionally, in a fourth implementation manner of the second aspect of the present invention, the 2D conversion module is specifically configured to:
normalizing the time sequence value in the sub-time sequence data to obtain normalized time sequence data;
performing polar coordinate conversion on the normalized time sequence data to obtain polar coordinate time sequence data;
and intercepting a plurality of corresponding values of the timestamps in the polar coordinate time sequence data, and performing trigonometric function transformation on the corresponding values of the timestamps to obtain a first time sequence angular field matrix.
Optionally, in a fifth implementation manner of the second aspect of the present invention, the timing sequence reconstructing module is specifically configured to:
inputting the first time sequence angle field matrix into the multilayer one-dimensional convolution layer for convolution operation to obtain a second characteristic matrix;
inputting the second feature matrix into the pooling layer for mean value coding to obtain a second coding matrix;
inputting the second coding matrix into the multilayer one-dimensional deconvolution layer for feature conversion to obtain a second conversion matrix;
and inputting the second conversion matrix into the convolution layer for characteristic reconstruction to obtain a second time sequence angle field matrix.
Optionally, in a sixth implementation manner of the second aspect of the present invention, the determining module is specifically configured to:
calculating matrix element errors of corresponding positions in the first time sequence angle field matrix and the second time sequence angle field matrix to obtain a plurality of matrix element error values;
calling a preset loss function, and calculating the sum of squares of the error values of the matrix elements;
and calculating the average value of the square sum of the error values of the elements of each matrix to obtain the loss values of the first time sequence angle field matrix and the second time sequence angle field matrix.
A third aspect of the present invention provides a time series data abnormality detection apparatus, including: a memory and at least one processor, the memory having instructions stored therein; the at least one processor calls the instructions in the memory to cause the time series data abnormality detection apparatus to execute the time series data abnormality detection method described above.
A fourth aspect of the present invention provides a computer-readable storage medium having stored therein instructions, which, when run on a computer, cause the computer to execute the above-described time-series data abnormality detection method.
In the technical scheme provided by the invention, the anomaly detection is carried out by combining the time series 2D model and the time series reconstruction model, the calculation detection can be carried out according to the input time series, and the time when the anomaly exists is identified, and the method specifically comprises the following steps: the method comprises the steps of collecting time sequence data to be detected, and extracting sub-time sequence data with fixed time length through a preset sliding time window, wherein the collection of longer time sequence is beneficial to inputting more information to a model; calling a preset gram angular field algorithm, converting the sub-time sequence data into a 2D time sequence angular field matrix, and acquiring the dependency information of the time sequence in the time sequence data through 2D; inputting the 2D time sequence angle field matrix into a preset time sequence reconstruction model to reconstruct time sequence data to obtain a reconstructed time sequence angle field matrix, wherein the time sequence reconstruction model has better reconstruction capability on a time sequence without abnormality, but the loss value for reconstructing an abnormal time sequence is larger, so that the model is used for time sequence reconstruction, and the time sequence data with the loss value larger than a preset threshold value is judged to be abnormal. The invention carries out time sequence reconstruction on the time sequence dependence information, and improves the accuracy of time sequence data anomaly detection.
Drawings
FIG. 1 is a schematic diagram of a first embodiment of a method for detecting an anomaly in time series data according to an embodiment of the present invention;
FIG. 2 is a diagram of a second embodiment of a method for detecting an anomaly in time series data according to an embodiment of the present invention;
FIG. 3 is a diagram of a third embodiment of a method for detecting an anomaly in time series data according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a first embodiment of an apparatus for detecting an abnormality in time series data according to an embodiment of the present invention;
FIG. 5 is a diagram of a second embodiment of a device for detecting an abnormality in timing data according to an embodiment of the present invention;
fig. 6 is a schematic diagram of an embodiment of a time series data anomaly detection device in an embodiment of the present invention.
Detailed Description
The embodiment of the invention provides a method, a device, equipment and a storage medium for detecting time series data abnormity. The terms "first," "second," "third," "fourth," and the like in the description and in the claims, as well as in the drawings, if any, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It will be appreciated that the data so used may be interchanged under appropriate circumstances such that the embodiments described herein may be practiced otherwise than as specifically illustrated or described herein. Furthermore, the terms "comprises," "comprising," or "having," and any variations thereof, are intended to cover non-exclusive inclusions, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
For convenience of understanding, a detailed flow of the embodiment of the present invention is described below, and referring to fig. 1, a first embodiment of the method for detecting an abnormal time series data according to the embodiment of the present invention includes:
101. collecting time sequence data to be detected, and extracting sub time sequence data with fixed time length through a preset sliding time window;
it is to be understood that the execution subject of the present invention may be a time series data abnormality detection apparatus, and may also be a terminal or a server, which is not limited herein. The embodiment of the present invention is described by taking a server as an execution subject.
In this embodiment, time series data that is highly correlated with time and is generated by various types of monitoring index data in the monitoring system is used as data to be detected, longer time series information is also added in this embodiment, and a time series with a length of 3N (2N length is extended forward) is taken as input.
Optionally, in an embodiment, the acquiring the time series data to be detected, and extracting the sub-time series data with a fixed time length through the sliding time window includes:
collecting time sequence data to be detected, and adjusting the window length of the sliding time window;
and acquiring a time sequence acquisition value of a fixed time length forwards at intervals of preset fixed time through the adjusted sliding time window to obtain the sub-time sequence data of the time sequence data to be detected.
In this embodiment, the window of the sliding time window is adjusted to a certain length, for example: the time sequence collection granularity is 1min, the collection time length is 60min, the length is 60 (including 60 collection values), and the time sequence data with the length of 2 x 60 is obtained by extending forwards.
102. Calling a preset gram angular field algorithm, and performing 2D conversion on the sub-time sequence data to obtain a first time sequence angular field matrix;
in this embodiment, the gram angular field algorithm is used to measure the characteristics of each dimension and the relationship between each dimension, diagonal elements provide respective information of different feature maps and other elements provide related information between different feature maps in a multi-scale matrix obtained after inner product, and the obtained time sequence angular field matrix can embody which features are present and the tightness degree between different features.
103. Inputting the first time sequence angle field matrix into a preset time sequence reconstruction model for time sequence data reconstruction to obtain a second time sequence angle field matrix;
in this embodiment, the time-series reconstruction model includes an encoder and a decoder, where the encoder is composed of multiple one-dimensional convolutional layers and pooling layers, and the decoder is composed of multiple one-dimensional deconvolution layers and convolutional layers. The one-dimensional convolutional layer is a layer that compresses input data, for example: the dimension of input data is 8, the dimension of the filter is 5, and the dimension of output data after convolution is 8-5+ 1-4. The main functions of the pooling layer are down-sampling, dimension reduction, redundant information removal, feature compression, network complexity simplification, calculation amount reduction and memory consumption reduction. The role of the one-dimensional deconvolution layer is to restore the compressed matrix, for example: one step size is 2, the number of deconvolution outputs of size 3 is twice the input, and the coefficient becomes 4. The convolutional layer outputs a second timing angle field matrix of [ N, 3N ].
Optionally, in an embodiment, the inputting the first time series angle field matrix into a preset time series reconstruction model for time series data reconstruction, and obtaining a second time series angle field matrix includes:
inputting the first time sequence angle field matrix into the multilayer one-dimensional convolution layer for convolution operation to obtain a second characteristic matrix;
inputting the second feature matrix into the pooling layer for mean value coding to obtain a second coding matrix;
inputting the second coding matrix into the multilayer one-dimensional deconvolution layer for feature conversion to obtain a second conversion matrix;
and inputting the second conversion matrix into the convolution layer for characteristic reconstruction to obtain a second time sequence angle field matrix.
In this embodiment, the encoder includes multiple layers of one-dimensional convolutional layers and pooling layers, with the dimension of [ N, 3N ]]Is encoded as a dimension [ L, C ]]X of1And x is1Then decoded into dimension [ N, 3N ] by decoder (including multi-layer one-dimensional deconvolution layer and convolution layer)]The second timing angle field matrix. L and C are dimensions of the encoder after length compression and dimension increase, namely L<N and C>3N。
104. Calling a preset loss function, calculating loss values of the first time sequence angle field matrix and the second time sequence angle field matrix, and judging whether the loss values exceed a preset loss threshold value;
in this embodiment, the loss function is MSE, and whether there is an abnormality in the time series data is determined by solving the mean variance between the 2D time series angular field matrix and the reconstructed time series angular field matrix.
Optionally, in an embodiment, the invoking a preset penalty function, and calculating penalty values of the first timing angle field matrix and the second timing angle field matrix includes:
calculating matrix element errors of corresponding positions in the first time sequence angle field matrix and the second time sequence angle field matrix to obtain a plurality of matrix element error values;
calling a preset loss function, and calculating the sum of squares of the error values of the matrix elements;
and calculating the average value of the square sum of the error values of the elements of each matrix to obtain the loss values of the first time sequence angle field matrix and the second time sequence angle field matrix.
In this embodiment, the loss function is:
where MSE represents the mean variance, n represents n elements in the matrix, x represents matrix element values in the second temporal angular field matrix, and x' represents matrix element values in the first temporal angular field matrix.
105. And if so, determining that the sub time sequence data is abnormal.
In this embodiment, if the loss value exceeds the preset loss threshold, it is determined that the sub-sequence data is abnormal, and the operation and maintenance system finds the relevant abnormality and gives an early warning in time.
The embodiment of the invention carries out anomaly detection by combining time series 2D and time series reconstruction models, can carry out calculation detection according to the input time series, and identifies the time when the anomaly exists, and specifically comprises the following steps: the method comprises the steps of collecting time sequence data to be detected, and extracting sub-time sequence data with fixed time length through a preset sliding time window, wherein the collection of longer time sequence is beneficial to inputting more information to a model; calling a preset gram angular field algorithm, converting the sub-time sequence data into a 2D time sequence angular field matrix, and acquiring the dependency information of the time sequence in the time sequence data through 2D; inputting the 2D time sequence angle field matrix into a preset time sequence reconstruction model to reconstruct time sequence data to obtain a reconstructed time sequence angle field matrix, wherein the time sequence reconstruction model has better reconstruction capability on a time sequence without abnormality, but the loss value for reconstructing an abnormal time sequence is larger, so that the model is used for time sequence reconstruction, and the time sequence data with the loss value larger than a preset threshold value is judged to be abnormal. The invention carries out time sequence reconstruction on the time sequence dependence information, and improves the accuracy of time sequence data anomaly detection.
Referring to fig. 2, a second embodiment of the method for detecting an abnormal time series data according to the embodiment of the present invention includes:
201. collecting abnormal time sequence data, and extracting abnormal time sequence data with fixed time length through the sliding time window;
202. calling a preset gram angular field algorithm, and performing 2D conversion on the abnormal-free sub time sequence data to obtain a third time sequence angular field matrix;
203. inputting the third time sequence angle field matrix into the multilayer one-dimensional convolution layer to carry out convolution operation to obtain a first characteristic matrix;
204. inputting the first feature matrix into the pooling layer for mean value coding to obtain a first coding matrix;
205. inputting the first coding matrix into the multilayer one-dimensional deconvolution layer for feature conversion to obtain a first conversion matrix;
206. inputting the first conversion matrix into the convolution layer for characteristic reconstruction to obtain a fourth time sequence angular field matrix;
207. calling a preset loss function, and calculating loss values of the third time sequence angle field matrix and the fourth time sequence angle field matrix;
in this embodiment, the loss function is:
where MSE represents the mean variance, n represents n elements in the matrix, y represents matrix element values in the fourth time-series angular field matrix, and y' represents matrix element values in the third time-series angular field matrix.
208. Calling a preset optimization algorithm to perform parameter optimization on the encoder and the decoder according to the loss value until the encoder and the decoder are converged to obtain a time sequence reconstruction model;
optionally, in an embodiment, the invoking a preset optimization algorithm to perform parameter optimization on the encoder and the decoder according to the loss value until the encoder and the decoder converge to obtain a time sequence reconstruction model includes:
calling a preset optimization algorithm to calculate first moment estimation and second moment estimation of the encoder and the decoder according to the loss value;
and correcting the first moment estimation and the second moment estimation, and updating parameters through the corrected first moment estimation and second moment estimation until the encoder and the decoder converge to obtain a time sequence reconstruction model.
In this embodiment, the optimization algorithm adopts an Adam optimization algorithm, the Adam optimization algorithm calculates adaptive learning rates of different parameters from budgets of the first moment and the second moment of the gradient, and the execution process of the Adam algorithm is as follows:
first moment estimate V for solving gradientdW、VdbAnd second moment estimation SdW、Sdb：
VdW：＝β1VdW+(1-β1)dW；
Vdb：＝β1Vdb+(1-β1)db；
SdW：＝β2SdW+(1-β2)dW2；
Sdb：＝β2Sdb+(1-β2)db2；
Wherein, beta1And beta2Exponential decay rates of the first and second order moment estimates, the matrices V and S on the right are the moment estimates of the last iteration, the sign ": and "stands for assignment operation. And correcting the first moment estimation and the second moment estimation, and updating the parameters through the corrected moment estimation. And judging whether the precision of the training model meets the requirement. If the time sequence of the training is consistent with the time sequence of the training, outputting the trained network model as a time sequence reconstruction model; otherwise, network training is carried out again until the training model converges to obtain a time sequence reconstruction model.
209. Collecting time sequence data to be detected, and extracting sub time sequence data with fixed time length through a preset sliding time window;
210. calling a preset gram angular field algorithm, and performing 2D conversion on the sub-time sequence data to obtain a first time sequence angular field matrix;
211. inputting the first time sequence angle field matrix into a preset time sequence reconstruction model for time sequence data reconstruction to obtain a second time sequence angle field matrix;
212. calling a preset loss function, calculating loss values of the first time sequence angle field matrix and the second time sequence angle field matrix, and judging whether the loss values exceed a preset loss threshold value;
213. and if so, determining that the sub time sequence data is abnormal.
In the embodiment of the invention, the time sequence reconstruction model is trained by adopting the abnormal time sequence data, so that the time sequence reconstruction model has a better reconstruction effect on the abnormal time sequence data, but the abnormal time sequence cannot be reconstructed and the loss value is larger because the abnormal data is not trained. Based on the principle, the model is used for time sequence reconstruction, and the sample with the loss value larger than the preset threshold value is judged to be abnormal.
Referring to fig. 3, a third embodiment of the method for detecting an abnormal time series data according to the embodiment of the present invention includes:
301. collecting time sequence data to be detected, and extracting sub time sequence data with fixed time length through a preset sliding time window;
302. normalizing the time sequence value in the sub-time sequence data to obtain normalized time sequence data;
303. performing polar coordinate conversion on the normalized time sequence data to obtain polar coordinate time sequence data;
304. intercepting a plurality of corresponding values of the timestamps in the polar coordinate time sequence data, and performing trigonometric function transformation on the corresponding values of the timestamps to obtain a first time sequence angular field matrix;
in this embodiment, for a time sequence with a length of N, the gram angular field algorithm quantizes the angular relationship between any 2 timestamps in the input time sequence through a gram matrix, so as to obtain an N × N matrix. And 2D conversion is carried out by adopting a gram angular field algorithm, and the time sequence with the length of N is directly substituted into a matrix which generates N x N, and then substituted into a time sequence reconstruction model. In this embodiment, to add longer time series information, a time series with a length of 3N (extending forward by a length of 2N) is substituted, a matrix of 3N × 3N is generated, and then only the last N time-stamp corresponding values, that is, a matrix of N × 3N, are truncated. The calculation process can be summarized as follows:
x＝GAF(t)[-N:]；
where t is a time series of length 3N and x is a 2D matrix of [ N, 3N ], which is then used to detect anomalies for the last N timetags.
305. Inputting the first time sequence angle field matrix into a preset time sequence reconstruction model for time sequence data reconstruction to obtain a second time sequence angle field matrix;
306. calling a preset loss function, calculating loss values of the first time sequence angle field matrix and the second time sequence angle field matrix, and judging whether the loss values exceed a preset loss threshold value;
307. and if so, determining that the sub time sequence data is abnormal.
In the embodiment of the invention, after the time sequence data is subjected to 2D conversion by the Gelam angular field algorithm, the dimensionality is expanded, the original spatial information of the data is enhanced, meanwhile, because the format and the time have high relevance, more dependency information among time sequences can be obtained, and the problem of low identification rate of the common abnormal detection method on the time sequence data with longer sequence is solved.
With reference to fig. 4, the method for detecting a time series data anomaly in an embodiment of the present invention is described above, and a time series data anomaly detection apparatus in an embodiment of the present invention is described below, where a first embodiment of the time series data anomaly detection apparatus in an embodiment of the present invention includes:
the acquisition module 401 is configured to acquire time sequence data to be detected and extract sub-time sequence data with a fixed time length through a preset sliding time window;
a 2D conversion module 402, configured to invoke a preset glaham angular field algorithm, perform 2D conversion on the sub-timing sequence data to obtain a first timing sequence angular field matrix;
a time sequence reconstruction module 403, configured to input the first time sequence angle field matrix into a preset time sequence reconstruction model to perform time sequence data reconstruction, so as to obtain a second time sequence angle field matrix;
a judging module 404, configured to invoke a preset loss function, calculate a loss value of the first time sequence angle field matrix and the second time sequence angle field matrix, and judge whether the loss value exceeds a preset loss threshold;
an output module 405, configured to determine that the sub-timing data is abnormal if the loss value exceeds the loss threshold.
Referring to fig. 5, a second embodiment of the apparatus for detecting an abnormal time series data according to the present invention includes:
the acquisition module 401 is configured to acquire time sequence data to be detected and extract sub-time sequence data with a fixed time length through a preset sliding time window;
a 2D conversion module 402, configured to invoke a preset glaham angular field algorithm, perform 2D conversion on the sub-timing sequence data to obtain a first timing sequence angular field matrix;
a time sequence reconstruction module 403, configured to input the first time sequence angle field matrix into a preset time sequence reconstruction model to perform time sequence data reconstruction, so as to obtain a second time sequence angle field matrix;
a judging module 404, configured to invoke a preset loss function, calculate a loss value of the first time sequence angle field matrix and the second time sequence angle field matrix, and judge whether the loss value exceeds a preset loss threshold;
an output module 405, configured to determine that the sub-timing data is abnormal if the loss value exceeds the loss threshold.
Optionally, in an embodiment, the time-series reconstruction model includes an encoder and a decoder, where the encoder is composed of multiple layers of one-dimensional convolutional layers and pooling layers, and the decoder is composed of multiple layers of one-dimensional deconvolution layers and convolutional layers, and the time-series data anomaly detection apparatus further includes:
the sample processing module 406 is configured to collect abnormal-free time sequence data, and extract abnormal-free time sequence data with a fixed time length through the sliding time window; calling a preset gram angular field algorithm, and performing 2D conversion on the abnormal-free sub time sequence data to obtain a third time sequence angular field matrix;
a model training module 407, configured to input the third time sequence angle field matrix into the multilayer one-dimensional convolution layer to perform convolution operation, so as to obtain a first feature matrix; inputting the first feature matrix into the pooling layer for mean value coding to obtain a first coding matrix; inputting the first coding matrix into the multilayer one-dimensional deconvolution layer for feature conversion to obtain a first conversion matrix; inputting the first conversion matrix into the convolution layer for characteristic reconstruction to obtain a fourth time sequence angular field matrix; calling a preset loss function, and calculating loss values of the third time sequence angle field matrix and the fourth time sequence angle field matrix;
and the model optimization module 408 is configured to invoke a preset optimization algorithm to perform parameter optimization on the encoder and the decoder according to the loss value until the encoder and the decoder converge, so as to obtain a time sequence reconstruction model.
Optionally, in an embodiment, the model optimization module 408 is specifically configured to:
calling a preset optimization algorithm to calculate first moment estimation and second moment estimation of the encoder and the decoder according to the loss value;
and correcting the first moment estimation and the second moment estimation, and updating parameters through the corrected first moment estimation and second moment estimation until the encoder and the decoder converge to obtain a time sequence reconstruction model.
Optionally, in an embodiment, the acquisition module 401 is specifically configured to:
collecting time sequence data to be detected, and adjusting the window length of the sliding time window;
and acquiring a time sequence acquisition value of a fixed time length forwards at intervals of preset fixed time through the adjusted sliding time window to obtain the sub-time sequence data of the time sequence data to be detected.
Optionally, in an embodiment, the 2D conversion module 402 is specifically configured to:
normalizing the time sequence value in the sub-time sequence data to obtain normalized time sequence data;
performing polar coordinate conversion on the normalized time sequence data to obtain polar coordinate time sequence data;
and intercepting a plurality of corresponding values of the timestamps in the polar coordinate time sequence data, and performing trigonometric function transformation on the corresponding values of the timestamps to obtain a first time sequence angular field matrix.
Optionally, in an embodiment, the timing reconstruction module 403 is specifically configured to:
inputting the first time sequence angle field matrix into the multilayer one-dimensional convolution layer for convolution operation to obtain a second characteristic matrix;
inputting the second feature matrix into the pooling layer for mean value coding to obtain a second coding matrix;
inputting the second coding matrix into the multilayer one-dimensional deconvolution layer for feature conversion to obtain a second conversion matrix;
and inputting the second conversion matrix into the convolution layer for characteristic reconstruction to obtain a second time sequence angle field matrix.
Optionally, in an embodiment, the determining module 404 is specifically configured to:
calculating matrix element errors of corresponding positions in the first time sequence angle field matrix and the second time sequence angle field matrix to obtain a plurality of matrix element error values;
calling a preset loss function, and calculating the sum of squares of the error values of the matrix elements;
and calculating the average value of the square sum of the error values of the elements of each matrix to obtain the loss values of the first time sequence angle field matrix and the second time sequence angle field matrix.
The embodiment of the invention carries out anomaly detection by combining time series 2D and time series reconstruction models, can carry out calculation detection according to the input time series, and identifies the time when the anomaly exists, and specifically comprises the following steps: the method comprises the steps of collecting time sequence data to be detected, and extracting sub-time sequence data with fixed time length through a preset sliding time window, wherein the collection of longer time sequence is beneficial to inputting more information to a model; calling a preset gram angular field algorithm, converting the sub-time sequence data into a 2D time sequence angular field matrix, and acquiring the dependency information of the time sequence in the time sequence data through 2D; inputting the 2D time sequence angle field matrix into a preset time sequence reconstruction model to reconstruct time sequence data to obtain a reconstructed time sequence angle field matrix, wherein the time sequence reconstruction model has better reconstruction capability on a time sequence without abnormality, but the loss value for reconstructing an abnormal time sequence is larger, so that the model is used for time sequence reconstruction, and the time sequence data with the loss value larger than a preset threshold value is judged to be abnormal. The invention carries out time sequence reconstruction on the time sequence dependence information, and improves the accuracy of time sequence data anomaly detection.
Fig. 4 and 5 describe the time-series data anomaly detection device in the embodiment of the present invention in detail from the perspective of the modular functional entity, and the time-series data anomaly detection device in the embodiment of the present invention is described in detail from the perspective of hardware processing.
Fig. 6 is a schematic structural diagram of a time-series data abnormality detecting apparatus according to an embodiment of the present invention, where the time-series data abnormality detecting apparatus 600 may generate relatively large differences due to different configurations or performances, and may include one or more processors (CPUs) 610 (e.g., one or more processors) and a memory 620, and one or more storage media 630 (e.g., one or more mass storage devices) storing applications 633 or data 632. Memory 620 and storage medium 630 may be, among other things, transient or persistent storage. The program stored in the storage medium 630 may include one or more modules (not shown), each of which may include a series of instruction operations in the temporal data anomaly detection apparatus 600. Still further, the processor 610 may be configured to communicate with the storage medium 630 to execute a series of instruction operations in the storage medium 630 on the time series data anomaly detection device 600.
The time series data anomaly detection apparatus 600 may also include one or more power supplies 640, one or more wired or wireless network interfaces 650, one or more input-output interfaces 660, and/or one or more operating systems 631, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, and the like. Those skilled in the art will appreciate that the configuration of the time series data abnormality detecting apparatus shown in fig. 6 does not constitute a limitation to the time series data abnormality detecting apparatus, and may include more or less components than those shown, or some components in combination, or a different arrangement of components.
The invention also provides a time series data abnormity detection device, which comprises a memory and a processor, wherein the memory stores computer readable instructions, and the computer readable instructions, when executed by the processor, enable the processor to execute the steps of the time series data abnormity detection method in the above embodiments.
The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium, and which may also be a volatile computer-readable storage medium, having stored therein instructions, which, when run on a computer, cause the computer to perform the steps of the time-series data anomaly detection method.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a read-only memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A time series data abnormity detection method is characterized by comprising the following steps:
collecting time sequence data to be detected, and extracting sub time sequence data with fixed time length through a preset sliding time window;
calling a preset gram angular field algorithm, and performing 2D conversion on the sub-time sequence data to obtain a first time sequence angular field matrix;
inputting the first time sequence angle field matrix into a preset time sequence reconstruction model for time sequence data reconstruction to obtain a second time sequence angle field matrix;
calling a preset loss function, calculating loss values of the first time sequence angle field matrix and the second time sequence angle field matrix, and judging whether the loss values exceed a preset loss threshold value;
and if so, determining that the sub time sequence data is abnormal.
2. The method according to claim 1, wherein the time series reconstruction model comprises an encoder and a decoder, wherein the encoder is composed of a plurality of layers of one-dimensional convolutional layers and pooling layers, and the decoder is composed of a plurality of layers of one-dimensional deconvolution layers and convolutional layers, and before the acquiring the time series data to be detected, the method further comprises:
collecting abnormal time sequence data, and extracting abnormal time sequence data with fixed time length through the sliding time window;
calling a preset gram angular field algorithm, and performing 2D conversion on the abnormal-free sub time sequence data to obtain a third time sequence angular field matrix;
inputting the third time sequence angle field matrix into the multilayer one-dimensional convolution layer to carry out convolution operation to obtain a first characteristic matrix;
inputting the first feature matrix into the pooling layer for mean value coding to obtain a first coding matrix;
inputting the first coding matrix into the multilayer one-dimensional deconvolution layer for feature conversion to obtain a first conversion matrix;
inputting the first conversion matrix into the convolution layer for characteristic reconstruction to obtain a fourth time sequence angular field matrix;
calling a preset loss function, and calculating loss values of the third time sequence angle field matrix and the fourth time sequence angle field matrix;
and calling a preset optimization algorithm to perform parameter optimization on the encoder and the decoder according to the loss value until the encoder and the decoder are converged to obtain a time sequence reconstruction model.
3. The method for detecting the abnormal time series data according to claim 2, wherein the step of calling a preset optimization algorithm to perform parameter optimization on the encoder and the decoder according to the loss value until the encoder and the decoder converge to obtain a time series reconstruction model comprises the steps of:
calling a preset optimization algorithm to calculate first moment estimation and second moment estimation of the encoder and the decoder according to the loss value;
and correcting the first moment estimation and the second moment estimation, and updating parameters through the corrected first moment estimation and second moment estimation until the encoder and the decoder converge to obtain a time sequence reconstruction model.
4. The method for detecting the abnormality of the time series data according to claim 1, wherein the collecting the time series data to be detected and extracting the sub time series data of a fixed time length through the sliding time window comprises:
collecting time sequence data to be detected, and adjusting the window length of the sliding time window;
and acquiring a time sequence acquisition value of a fixed time length forwards at intervals of preset fixed time through the adjusted sliding time window to obtain the sub-time sequence data of the time sequence data to be detected.
5. The method for detecting the abnormal time series data according to claim 1, wherein the calling a preset gram corner field algorithm to perform 2D conversion on the sub time series data to obtain a first time series corner field matrix comprises:
normalizing the time sequence value in the sub-time sequence data to obtain normalized time sequence data;
performing polar coordinate conversion on the normalized time sequence data to obtain polar coordinate time sequence data;
and intercepting a plurality of corresponding values of the timestamps in the polar coordinate time sequence data, and performing trigonometric function transformation on the corresponding values of the timestamps to obtain a first time sequence angular field matrix.
6. The method for detecting the abnormality of the time series data according to claim 1 or 5, wherein the step of inputting the first time series angle field matrix into a preset time series reconstruction model for reconstructing the time series data to obtain a second time series angle field matrix comprises:
inputting the first time sequence angle field matrix into the multilayer one-dimensional convolution layer for convolution operation to obtain a second characteristic matrix;
inputting the second feature matrix into the pooling layer for mean value coding to obtain a second coding matrix;
inputting the second coding matrix into the multilayer one-dimensional deconvolution layer for feature conversion to obtain a second conversion matrix;
and inputting the second conversion matrix into the convolution layer for characteristic reconstruction to obtain a second time sequence angle field matrix.
7. The method of claim 6, wherein the invoking a pre-set penalty function to calculate penalty values for the first and second temporal angular field matrices comprises:
calculating matrix element errors of corresponding positions in the first time sequence angle field matrix and the second time sequence angle field matrix to obtain a plurality of matrix element error values;
calling a preset loss function, and calculating the sum of squares of the error values of the matrix elements;
and calculating the average value of the square sum of the error values of the elements of each matrix to obtain the loss values of the first time sequence angle field matrix and the second time sequence angle field matrix.
8. A time series data abnormality detection apparatus, characterized by comprising:
the acquisition module is used for acquiring time sequence data to be detected and extracting sub-time sequence data with fixed time length through a preset sliding time window;
the 2D conversion module is used for calling a preset Graham angular field algorithm and carrying out 2D conversion on the sub-time sequence data to obtain a first time sequence angular field matrix;
the time sequence reconstruction module is used for inputting the first time sequence angle field matrix into a preset time sequence reconstruction model to reconstruct time sequence data to obtain a second time sequence angle field matrix;
the judging module is used for calling a preset loss function, calculating the loss values of the first time sequence angle field matrix and the second time sequence angle field matrix and judging whether the loss values exceed a preset loss threshold value or not;
and the output module is used for determining that the sub-time sequence data is abnormal if the loss value exceeds the loss threshold value.
9. A time-series data abnormality detection apparatus characterized by comprising: a memory and at least one processor, the memory having instructions stored therein;
the at least one processor invokes the instructions in the memory to cause the time series data anomaly detection apparatus to perform the time series data anomaly detection method of any one of claims 1-7.
10. A computer-readable storage medium having instructions stored thereon, wherein the instructions, when executed by a processor, implement the method of temporal data anomaly detection according to any one of claims 1-7.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011635793.6A CN112732785A (en) | 2020-12-31 | 2020-12-31 | Time series data abnormity detection method, device, equipment and storage medium |
| PCT/CN2021/084237 WO2022141871A1 (en) | 2020-12-31 | 2021-03-31 | Time sequence data anomaly detection method, apparatus and device, and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011635793.6A CN112732785A (en) | 2020-12-31 | 2020-12-31 | Time series data abnormity detection method, device, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112732785A true CN112732785A (en) | 2021-04-30 |
Family
ID=75608617
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011635793.6A Pending CN112732785A (en) | 2020-12-31 | 2020-12-31 | Time series data abnormity detection method, device, equipment and storage medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN112732785A (en) |
| WO (1) | WO2022141871A1 (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114067093A (en) * | 2021-09-23 | 2022-02-18 | 济南大学 | Time sequence and image-based scattered pollution user accurate capturing method and system |
| CN114494242A (en) * | 2022-02-21 | 2022-05-13 | 平安科技（深圳）有限公司 | Time series data detection method, device, equipment and computer storage medium |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115563095B (en) * | 2022-10-18 | 2023-11-24 | 上海宇佑船舶科技有限公司 | Data reconstruction method and system based on time sequence |
| CN116383747A (en) * | 2023-04-06 | 2023-07-04 | 中国科学院空间应用工程与技术中心 | Anomaly detection method for generating countermeasure network based on multi-time scale depth convolution |
| CN116107194B (en) * | 2023-04-13 | 2023-06-09 | 北京新科以仁科技发展有限公司 | Automatic switching control method, device and equipment of laser and storage medium |
| CN117113243B (en) * | 2023-06-28 | 2024-01-26 | 武汉盈风能源科技有限公司 | Photovoltaic equipment abnormality detection method |
| CN116865631B (en) * | 2023-07-20 | 2024-01-26 | 浙江联大科技有限公司 | Mute tubular motor of automatic curtain and control method thereof |
| CN116633705B (en) * | 2023-07-26 | 2023-10-13 | 山东省计算中心（国家超级计算济南中心） | Industrial control system abnormality detection method and system based on composite automatic encoder |
| CN117543791A (en) * | 2023-11-08 | 2024-02-09 | 深圳市瀚海星光科技有限公司 | Power supply detection method, device, equipment and storage medium for power supply |
| CN117724935B (en) * | 2024-02-06 | 2024-06-07 | 山东大学 | Multi-index abnormality detection method and system for software system |
| CN117760593B (en) * | 2024-02-22 | 2024-05-10 | 张家港江海粮油港务有限公司 | Abnormal monitoring method for storage temperature of grain crops |
| CN117786371B (en) * | 2024-02-27 | 2024-05-10 | 聊城市检验检测中心 | Temperature monitoring data optimization prediction analysis method and system |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109086876A (en) * | 2018-09-21 | 2018-12-25 | 广州发展集团股份有限公司 | Method for detecting operation state, device, computer equipment and the storage medium of equipment |
| WO2019036095A1 (en) * | 2017-08-17 | 2019-02-21 | Siemens Aktiengesellschaft | Deep convolutional neural network based anomaly detection for transactive energy systems |
| CN110490132A (en) * | 2019-08-16 | 2019-11-22 | 京东城市(北京)数字科技有限公司 | Data processing method and device |
| US20200004616A1 (en) * | 2018-07-02 | 2020-01-02 | Nec Laboratories America, Inc. | Failure prediction |
| CN111669373A (en) * | 2020-05-25 | 2020-09-15 | 山东理工大学 | Network anomaly detection method and system based on space-time convolutional network and topology perception |
| CN111695598A (en) * | 2020-05-11 | 2020-09-22 | 东南大学 | Monitoring data abnormity diagnosis method based on generation countermeasure network |
| CN111860785A (en) * | 2020-07-24 | 2020-10-30 | 中山大学 | Time sequence prediction method and system based on attention mechanism cyclic neural network |
| CN111880998A (en) * | 2020-07-30 | 2020-11-03 | 平安科技（深圳）有限公司 | Service system anomaly detection method and device, computer equipment and storage medium |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20200387829A1 (en) * | 2019-06-06 | 2020-12-10 | Retrace Labs | Systems And Methods For Dental Treatment Prediction From Cross- Institutional Time-Series Information |
| US11657269B2 (en) * | 2019-05-23 | 2023-05-23 | Salesforce.Com, Inc. | Systems and methods for verification of discriminative models |
| CN110909046B (en) * | 2019-12-02 | 2023-08-11 | 上海舵敏智能科技有限公司 | Time-series abnormality detection method and device, electronic equipment and storage medium |
| CN112000830A (en) * | 2020-08-26 | 2020-11-27 | 中国科学技术大学 | Time sequence data detection method and device |
| CN112016041B (en) * | 2020-08-27 | 2023-08-04 | 重庆大学 | Time sequence real-time classification method based on gram sum angle field imaging and Shortcut-CNN |
-
2020
- 2020-12-31 CN CN202011635793.6A patent/CN112732785A/en active Pending
-
2021
- 2021-03-31 WO PCT/CN2021/084237 patent/WO2022141871A1/en active Application Filing
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019036095A1 (en) * | 2017-08-17 | 2019-02-21 | Siemens Aktiengesellschaft | Deep convolutional neural network based anomaly detection for transactive energy systems |
| US20200004616A1 (en) * | 2018-07-02 | 2020-01-02 | Nec Laboratories America, Inc. | Failure prediction |
| CN109086876A (en) * | 2018-09-21 | 2018-12-25 | 广州发展集团股份有限公司 | Method for detecting operation state, device, computer equipment and the storage medium of equipment |
| CN110490132A (en) * | 2019-08-16 | 2019-11-22 | 京东城市(北京)数字科技有限公司 | Data processing method and device |
| CN111695598A (en) * | 2020-05-11 | 2020-09-22 | 东南大学 | Monitoring data abnormity diagnosis method based on generation countermeasure network |
| CN111669373A (en) * | 2020-05-25 | 2020-09-15 | 山东理工大学 | Network anomaly detection method and system based on space-time convolutional network and topology perception |
| CN111860785A (en) * | 2020-07-24 | 2020-10-30 | 中山大学 | Time sequence prediction method and system based on attention mechanism cyclic neural network |
| CN111880998A (en) * | 2020-07-30 | 2020-11-03 | 平安科技（深圳）有限公司 | Service system anomaly detection method and device, computer equipment and storage medium |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114067093A (en) * | 2021-09-23 | 2022-02-18 | 济南大学 | Time sequence and image-based scattered pollution user accurate capturing method and system |
| CN114494242A (en) * | 2022-02-21 | 2022-05-13 | 平安科技（深圳）有限公司 | Time series data detection method, device, equipment and computer storage medium |
| WO2023155296A1 (en) * | 2022-02-21 | 2023-08-24 | 平安科技（深圳）有限公司 | Time series data detection method and apparatus, device, and computer storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2022141871A1 (en) | 2022-07-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112732785A (en) | Time series data abnormity detection method, device, equipment and storage medium | |
| CN107086944B (en) | Anomaly detection method and device | |
| CN112380098B (en) | Timing sequence abnormality detection method and device, computer equipment and storage medium | |
| CN110378487B (en) | Method, device, equipment and medium for verifying model parameters in horizontal federal learning | |
| CN110928718A (en) | Exception handling method, system, terminal and medium based on correlation analysis | |
| CN112862127B (en) | Sensor data exception handling method and device, electronic equipment and medium | |
| US9003076B2 (en) | Identifying anomalies in original metrics of a system | |
| CN112257755A (en) | Method and device for analyzing operating state of spacecraft | |
| CN108829998A (en) | A kind of detection method of deflection of bridge span, detection device and terminal device | |
| CN115563180A (en) | Dynamic threshold generation method, device, equipment and storage medium | |
| CN110991504A (en) | MOTCN-AE-based multi-dimensional time sequence abnormal value detection method and system | |
| CN112990442A (en) | Data determination method and device based on spatial position and electronic equipment | |
| CN113723861A (en) | Abnormal electricity consumption behavior detection method and device, computer equipment and storage medium | |
| KR101960755B1 (en) | Method and apparatus of generating unacquired power data | |
| CN112380073B (en) | Fault position detection method and device and readable storage medium | |
| CN111259338B (en) | Component failure rate correction method and device, computer equipment and storage medium | |
| CN115705413A (en) | Method and device for determining abnormal log | |
| CN113495913A (en) | Air quality data missing value interpolation method and device | |
| CN117092980A (en) | Electrical fault detection control system based on big data | |
| CN112242929B (en) | Log detection method and device | |
| CN113220551A (en) | Index trend prediction and early warning method and device, electronic equipment and storage medium | |
| CN113515684A (en) | Abnormal data detection method and device | |
| CN116127149B (en) | Quantification method and system for health degree of graph database cluster | |
| CN117407264B (en) | Method, device, computer equipment and medium for predicting memory aging residual time | |
| RU2777616C2 (en) | System, device and method for compression of alarm log, and data carrier |
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 |