CN117692346A - Message blocking prediction method and device based on spectrum regularization variation self-encoder - Google Patents
Message blocking prediction method and device based on spectrum regularization variation self-encoder Download PDFInfo
- Publication number
- CN117692346A CN117692346A CN202410130793.2A CN202410130793A CN117692346A CN 117692346 A CN117692346 A CN 117692346A CN 202410130793 A CN202410130793 A CN 202410130793A CN 117692346 A CN117692346 A CN 117692346A
- Authority
- CN
- China
- Prior art keywords
- encoder
- message
- regularization
- variation
- low
- 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
- 230000000903 blocking effect Effects 0.000 title claims abstract description 49
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000001228 spectrum Methods 0.000 title claims abstract description 35
- 239000013598 vector Substances 0.000 claims abstract description 56
- 230000002159 abnormal effect Effects 0.000 claims abstract description 19
- 230000005540 biological transmission Effects 0.000 claims abstract description 14
- 238000012549 training Methods 0.000 claims abstract description 14
- 238000007781 pre-processing Methods 0.000 claims abstract description 5
- 230000003595 spectral effect Effects 0.000 claims description 27
- 238000012545 processing Methods 0.000 claims description 22
- 230000006870 function Effects 0.000 claims description 18
- 238000004590 computer program Methods 0.000 claims description 8
- 238000009825 accumulation Methods 0.000 claims description 4
- 238000000926 separation method Methods 0.000 claims description 3
- 238000012544 monitoring process Methods 0.000 abstract description 6
- 230000005856 abnormality Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a message blocking prediction method and a device based on a spectrum regularization variation self-encoder, which are characterized in that firstly, message data are collected, data preprocessing is carried out, and a data set is constructed; then constructing and training a spectrum regularization-based variation self-encoder; all local windows in the data set are generated into low-dimensional embedded vectors based on the variation self-encoder, and k-1 low-dimensional embedded vectors are predicted by a transform decoder through the first k-1 low-dimensional embedded vectors; k-1 partial windows are reconstructed from a decoder in the encoder through variation, whether message blocking exists in the time window is judged through accumulated errors and a given threshold value, if the accumulated errors exceed the threshold value, abnormal message transmission is judged, and the message blocking exists. The invention can early warn the abnormal scene of the message sending speed in time, obviously improves the recall rate of message blocking abnormal alarm, simultaneously maintains high accuracy, and can be effectively applied to the real-time message blocking monitoring scene.
Description
Technical Field
The present invention relates to the field of message blocking prediction, and in particular, to a method and apparatus for predicting message blocking based on spectrum regularization variation self-encoder.
Background
The message center serves as a centralized message pushing and processing center and comprises a series of messages for pushing client business related information, banking activities, bank notices and the like to clients. The blocking of messages once occurs results in a large area of messages not being sent, with serious consequences. While blocking of messages is often not a momentary consequence, early message blocking often appears as a gradual slow message transmission, and over time, the gradual accumulation of messages to be transmitted may render the overall message service unusable. Since the abnormality of the feature guide system is not easy to be found, and the actual loss is caused by the fact that the abnormality is too late when the abnormality is found, the timely monitoring and alarming are particularly important.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a method and a device for predicting message blocking based on a spectrum regularization variation self-encoder.
The aim of the invention is realized by the following technical scheme: in a first aspect, the present invention provides a method for message blocking prediction based on a spectral regularization variation self-encoder, the method comprising the steps of:
s1: collecting message data, performing data preprocessing and characteristic engineering, and constructing a data set;
s2: constructing and training a spectrum regularization-based variation self-encoder, deducing a variation self-encoder loss function based on KL divergence and variation, adding a spectrum regularization term, and expressing the accumulation of each layer on the square of a characteristic spectrum function;
s3: generating all local windows in the data set into low-dimensional embedded vectors based on the trained variation self-encoder, and predicting k-1 low-dimensional embedded vectors by using the first k-1 low-dimensional embedded vectors through a transform decoder;
s4: k-1 partial windows are reconstructed from the decoder in the encoder through variation based on the k-1 low-dimensional embedded vectors predicted by the transducer decoder, whether message blocking exists in the time window or not is judged through accumulated errors and a given threshold value, if the accumulated errors exceed the threshold value, abnormal message transmission is judged, and message blocking exists.
Further, in step S1, the collected message data features include: message processing time, message push time consumption, message push success rate, single user message volume, user volume, message total volume, queue message processing speed and network speed.
Further, in step S1, missing values and abnormal values of the message data are processed, and standardized processing is performed, so that the values conform to normal distribution.
Further, in step S1, the feature engineering removes redundant features according to the correlation between features, and retains key features.
Further, in step S2, the variable self-encoder includes two parts of an encoder and a decoder; the encoder takes a lightweight convolution network of continuous 3*3 convolution, 3*3 depth separation convolution and SE structure as a basic structure and is recorded as a block; the encoder network structure sequentially comprises three continuous blocks and three continuous full-connection layers, and hidden variable parameters are generated; the decoder is a continuous three 3*3 deconvolution network for restoring the original features with hidden variables.
Further, in step S2, the spectral regularization-based variance self-encoder training procedure is as follows:
1) Selecting message data of N time periods with stable continuous transmission speed in a data set, wherein the input of an encoder is a local window of a plurality of continuous readings;
2) Generating means and standard deviations by the encoder based on all local windows;
3) The generated and low-dimensional embedded vector, namely the coding vector, is obtained in a re-parameterized mode;
4) Inputting the coded vector, and generating a reconstruction vector through a decoder;
5) The network parameters are optimized by back-propagation of the loss function f defined as follows:
wherein,for the feature map of the first layer of the model, L is the total layer number of the model, z represents the low-dimensional potential coding vector generated by the encoder,/for the model>For a priori distribution of potential encoding vectors, +.>Distribution of posterior->Is a function of the approximate estimate of (a),representing the likelihood of reconstructing data x from potentially encoded vectors, E operating for the desired value, D KL K-L divergence; the last term is a spectrum regularization term, which represents the summation of each layer to the square of the characteristic spectrum function, ++>Represents the spectral norms, λ is the spectral regularization parameter.
Further, in step S4, the specific process of determining whether there is a message blocking is:
1) Generating a plurality of low-dimensional embedded vector sequences corresponding to the non-overlapping local window sequences through an encoder;
2) Predicting k-1 low-dimensional embedded vectors by the first k-1 low-dimensional embedded vectors based on a transducer decoder;
3) K-1 partial windows are reconstructed from the low-dimensional embedded vectors obtained by the k-1 prediction after the general use through the variation from a decoder in the encoder;
4) And judging whether the time window is abnormal or not according to the accumulated L1 error and a given threshold value, if the time window is over the threshold value, judging that the message transmission is abnormal, otherwise, judging that the message transmission is normal.
In a second aspect, the present invention further provides a method apparatus for predicting message blocking based on a spectrum regularization variation self-encoder, including a memory and one or more processors, where the memory stores executable codes, and the processors implement the method for predicting message blocking based on a spectrum regularization variation self-encoder when executing the executable codes.
In a third aspect, the present invention also provides a computer readable storage medium having stored thereon a program which, when executed by a processor, implements the method for message blocking prediction based on a spectral regularization variation self-encoder.
In a fourth aspect, the present invention also provides a computer program product comprising computer programs/instructions which, when executed by a processor, implement the method for message blocking prediction based on a spectral regularization variation self-encoder.
The invention has the beneficial effects that: the spectral regularization-based variation self-encoder provided by the invention can enable the model to train more stably and be easy to converge under the low-dimensional characteristic, the model relies on the model to extract the local characteristic of the short window, and the time cyclic neural network is utilized to estimate the long-term trend, so that the abnormal scene within a certain duration range is early-warned in time. Meanwhile, a differential monitoring model is formed by combining characteristics such as classification, importance degree and the like of the information, and risk monitoring accuracy is improved. The scheme of the invention obviously improves the recall rate of message blocking abnormal alarm, simultaneously maintains high accuracy, and can be effectively applied to a real-time message blocking monitoring scene.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of a method for predicting message blocking based on a spectral regularization variation self-encoder.
Fig. 2 is a diagram showing a structure of a model in the present invention.
Fig. 3 is a block diagram of a message blocking prediction device based on a spectrum regularization variation self-encoder.
Detailed Description
The following describes the embodiments of the present invention in further detail with reference to the drawings.
Based on indexes such as daily sending data of each message scene, message processing time, sending quantity, message pushing time consumption, message pushing success rate, single-user message quantity, queue processing speed and the like, the invention provides a message blocking prediction method based on a spectrum regularization variation self-encoder, which is used for monitoring message sending abnormality. As shown in fig. 1 and 2, the method comprises a self-encoder based on spectrum regularization variation and a time-cycled neural network model, and the self-encoder based on spectrum regularization variation captures the structural rule of the time sequence on the local window. The encoder sums the local information of the short window into a low-dimensional embedding from which the decoder can generate the original window features. The model is trained only on normal message data, and an abnormal short window is identified by utilizing the characteristic that the abnormal message data and the normal message data are different in characteristics, namely, the abnormal message data cannot be generated by utilizing a normal sequence. The time-cycled neural network acts on the low-dimensional embedding generated by the encoder to model long-term trends, so that the detection algorithm can identify anomalies spanning multiple time scales in real time. The method comprises the following steps:
s1: and (5) collecting message data. The collected message data features include: message processing time, message push time consumption, message push success rate, single user message volume, user volume, message total volume, queue message processing speed, network speed and the like.
S2: data partitioning, data preprocessing and feature screening.
S201: the training set and the test set are partitioned. The training set is a continuous sequence containing normal message data, and the test set is a continuous sequence containing normal message data and abnormal message data.
S202: and (5) preprocessing data. And carrying out missing value and abnormal value processing on the message data, and carrying out standardization processing so that the numerical value accords with normal distribution.
S203: and (5) feature engineering. And removing redundant features according to the correlation among the features, and reserving key features.
S3: a spectral regularization-based variational self-encoder is constructed and trained.
S301: and (5) constructing a model. The variable self-encoder comprises an encoder and a decoder. The encoder takes a lightweight convolutional network (SENet) of continuous normal 3*3 convolution, 3*3 depth separation convolution and SE structure as a block. The encoder network structure is sequentially three consecutive blocks and three consecutive fc layers (fully-connected layers), and generates the hidden variable parameters. The decoder is a continuous three 3*3 deconvolution network for restoring the original features with hidden variables.
S302: a loss function is constructed. The variation is inferred from the encoder loss function based on the KL divergence and the variation, and has unbounded properties, so that the problems of unstable training and easy collapse of the model exist in the early stage of training, especially under the low-dimensional characteristics. Spectral regularization can make the convolution features conform to the lipschz constraint, which suppresses the spectral norms without reducing model complexity, making model training more robust. For any matrix a, the spectral norms are:
wherein,for a disturbance vector with a smaller L2 norm, the model needs to minimize the following loss function after adding spectral regularization:
wherein,for the feature map of the first layer of the model, L is the total layer number of the model, z represents the low-dimensional potential coding vector generated by the encoder,/for the model>For a priori distribution of potential encoding vectors, +.>Distribution of posterior->Is a function of the approximate estimate of (a),representing the likelihood of reconstructing x from the potentially encoded vector, E is the desired value operation, D KL Is K-L divergence. The last term is a spectrum regularization term, which represents the accumulation of each layer on the square of the characteristic spectrum function, and lambda is a spectrum regularization parameter.
S303: and training a model.
1) Selecting message data of N time periods with stable continuous transmission speedWherein->The m-dimensional feature vector representing the i-th timestamp. In which the encoder modelA local window of p consecutive readings is entered, < ->;
2) Input deviceGenerating a mean μ and a standard deviation σ by an encoder;
3) The μ and σ generated are used to obtain low-dimensional embedded vectors, i.e. encoded vectors, by means of re-parameterization;
4) Input encoded vector, and generate reconstructed vector by decoderThe decoder forms a feature matrix by reshape, and then forms a reconstruction vector by the decoder>;
5) The network parameters are optimized by back-propagation of defined loss functions.
S4: the transducer decoder is trained. After optimizing the self-encoder model based on the spectral regularization variation, all local windows in the training set are generated into low-dimensional embedded vectors by an encoder, and a transform decoder is trained by predicting k-1 low-dimensional embedded vectors by using the first k-1 low-dimensional embedded vectors.
The specific process of model training is as follows:
1) The transform decoder input corresponds to a low-dimensional embedded vector, i.e., an encoded vector, generated by the encoderCorresponding to k non-overlapping window sequences +.>,Representation ofA kth low-dimensional embedded vector;
2) Training the transducer decoder by predicting the k-1 low-dimensional embedded vectors after the k-1 low-dimensional embedded vectors, and optimizing the transducer decoder by minimizing the k-th low-dimensional embedded vector error, i.e。/>Representing the low-dimensional embedded vector of the kth prediction.
S5: message blocking anomaly prediction is performed from the encoder and the transducer decoder based on spectral regularization variation. The specific flow is as follows:
s501: generated by an encoderIs->;
S502: predicting k-1 low-dimensional embedded vectors by the first k-1 low-dimensional embedded vectors based on a transducer decoder;
s503: the low-dimensional embedded vector obtained by k-1 prediction is reconstructed from the decoder in the encoder through variation, namely k-1 partial windows;
S504: determining time by accumulated L1 error and given thresholdIf the message transmission is abnormal, judging that the message transmission is abnormal, otherwise, judging that the message transmission is normal. The threshold is set according to the scene, for example, the threshold of the client business related information is higher than the threshold of the scenes such as banking activity, banking announcement and the like.
Corresponding to the embodiment of the message blocking prediction method based on the spectrum regularization variation self-encoder, the invention also provides an embodiment of a message blocking prediction device based on the spectrum regularization variation self-encoder.
Referring to fig. 3, a message blocking prediction device based on a spectrum regularization variation self-encoder provided by an embodiment of the present invention includes a memory and one or more processors, where the memory stores executable codes, and the processors are configured to implement a message blocking prediction method based on a spectrum regularization variation self-encoder in the above embodiment when executing the executable codes.
The embodiment of the message blocking prediction device based on the spectrum regularization variation self-encoder can be applied to any device with data processing capability, and the device with data processing capability can be a device or a device such as a computer. The apparatus embodiments may be implemented by software, or may be implemented by hardware or a combination of hardware and software. Taking software implementation as an example, the device in a logic sense is formed by reading corresponding computer program instructions in a nonvolatile memory into a memory by a processor of any device with data processing capability. In terms of hardware, as shown in fig. 3, a hardware structure diagram of an apparatus with optional data processing capability, where a message blocking prediction device based on a spectrum regularization variation self-encoder provided by the present invention is shown, except for a processor, a memory, a network interface, and a nonvolatile memory shown in fig. 3, where an apparatus with optional data processing capability in an embodiment is generally according to an actual function of the apparatus with optional data processing capability, and other hardware may also be included, which is not described herein.
The implementation process of the functions and roles of each unit in the above device is specifically shown in the implementation process of the corresponding steps in the above method, and will not be described herein again.
For the device embodiments, reference is made to the description of the method embodiments for the relevant points, since they essentially correspond to the method embodiments. The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purposes of the present invention. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
The embodiment of the invention also provides a computer readable storage medium, on which a program is stored, which when executed by a processor, implements a method for predicting message blocking based on a spectral regularization variation self-encoder in the above embodiment.
The computer readable storage medium may be an internal storage unit, such as a hard disk or a memory, of any of the data processing enabled devices described in any of the previous embodiments. The computer readable storage medium may be any external storage device that has data processing capability, such as a plug-in hard disk, a Smart Media Card (SMC), an SD Card, a Flash memory Card (Flash Card), or the like, which are provided on the device. Further, the computer readable storage medium may include both internal storage units and external storage devices of any data processing device. The computer readable storage medium is used for storing the computer program and other programs and data required by the arbitrary data processing apparatus, and may also be used for temporarily storing data that has been output or is to be output.
The invention also provides a computer program product comprising computer programs/instructions which, when executed by a processor, implement the method for message blocking prediction based on a spectrum regularization variation self-encoder.
The above-described embodiments are intended to illustrate the present invention, not to limit it, and any modifications and variations made thereto are within the spirit of the invention and the scope of the appended claims.
Claims (10)
1. A method for message blocking prediction based on spectral regularization variation self-encoder, the method comprising the steps of:
s1: collecting message data, performing data preprocessing and characteristic engineering, and constructing a data set;
s2: constructing and training a spectrum regularization-based variation self-encoder, deducing a variation self-encoder loss function based on KL divergence and variation, adding a spectrum regularization term, and expressing the accumulation of each layer on the square of a characteristic spectrum function;
s3: generating all local windows in the data set into low-dimensional embedded vectors based on the trained variation self-encoder, and predicting k-1 low-dimensional embedded vectors by using the first k-1 low-dimensional embedded vectors through a transform decoder;
s4: k-1 partial windows are reconstructed from the decoder in the encoder through variation based on the k-1 low-dimensional embedded vectors predicted by the transducer decoder, whether message blocking exists in the time window or not is judged through accumulated errors and a given threshold value, if the accumulated errors exceed the threshold value, abnormal message transmission is judged, and message blocking exists.
2. The method for message blocking prediction based on spectral regularization variation self-encoder of claim 1, wherein in step S1, the collected message data features include: message processing time, message push time consumption, message push success rate, single user message volume, user volume, message total volume, queue message processing speed and network speed.
3. The method for predicting message blocking based on spectral regularization variation self-encoder according to claim 1, wherein in step S1, missing values and outliers are processed on the message data, and standardized processing is performed so that the values conform to normal distribution.
4. The method for message blocking prediction based on spectral regularization variation self-encoder of claim 1, wherein in step S1, feature engineering is to remove redundant features according to correlation between features, and preserve key features.
5. The method for message blocking prediction based on a spectral regularized variable self-encoder as claimed in claim 1, wherein in step S2, the variable self-encoder comprises two parts of encoder and decoder; the encoder takes a lightweight convolution network of continuous 3*3 convolution, 3*3 depth separation convolution and SE structure as a basic structure and is recorded as a block; the encoder network structure sequentially comprises three continuous blocks and three continuous full-connection layers, and hidden variable parameters are generated; the decoder is a continuous three 3*3 deconvolution network for restoring the original features with hidden variables.
6. The method for message blocking prediction based on spectrum regularized variation self-encoder according to claim 1, wherein in step S2, the spectrum regularized variation self-encoder training process is as follows:
1) Selecting message data of N time periods with stable continuous transmission speed in a data set, wherein the input of an encoder is a local window of a plurality of continuous readings;
2) Generating means and standard deviations by the encoder based on all local windows;
3) The generated and low-dimensional embedded vector, namely the coding vector, is obtained in a re-parameterized mode;
4) Inputting the coded vector, and generating a reconstruction vector through a decoder;
5) The network parameters are optimized by back-propagation of the loss function f defined as follows:
;
wherein,for the feature map of the first layer of the model, L is the total layer number of the model, z represents the low-dimensional potential coding vector generated by the encoder,/for the model>For a priori distribution of potential encoding vectors, +.>Distribution of posterior->Is a function of the approximate estimate of (a),representing the likelihood of reconstructing data x from potentially encoded vectors, E operating for the desired value, D KL K-L divergence; the last term is a spectrum regularization term, which represents the summation of each layer to the square of the characteristic spectrum function, ++>Represents the spectral norms, λ is the spectral regularization parameter.
7. The method for predicting message blocking based on spectrum regularization variation self-encoder according to claim 1, wherein in step S4, the specific process of determining whether there is a message blocking is as follows:
1) Generating a plurality of low-dimensional embedded vector sequences corresponding to the non-overlapping local window sequences through an encoder;
2) Predicting k-1 low-dimensional embedded vectors by the first k-1 low-dimensional embedded vectors based on a transducer decoder;
3) K-1 partial windows are reconstructed from the low-dimensional embedded vectors obtained by the k-1 prediction after the general use through the variation from a decoder in the encoder;
4) And judging whether the time window is abnormal or not according to the accumulated L1 error and a given threshold value, if the time window is over the threshold value, judging that the message transmission is abnormal, otherwise, judging that the message transmission is normal.
8. A method and apparatus for message blocking prediction based on spectral regularization variation self-encoder, comprising a memory and one or more processors, the memory having executable code stored therein, wherein the processor, when executing the executable code, implements a method for message blocking prediction based on spectral regularization variation self-encoder as claimed in any one of claims 1-7.
9. A computer readable storage medium having stored thereon a program, which when executed by a processor, implements a method of spectral regularization variation self-encoder based message blocking prediction as claimed in any of claims 1-7.
10. A computer program product comprising computer program/instructions which, when executed by a processor, implements a method of spectral regularization variation self-encoder based message blocking prediction according to any of claims 1-7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410130793.2A CN117692346A (en) | 2024-01-31 | 2024-01-31 | Message blocking prediction method and device based on spectrum regularization variation self-encoder |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202410130793.2A CN117692346A (en) | 2024-01-31 | 2024-01-31 | Message blocking prediction method and device based on spectrum regularization variation self-encoder |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117692346A true CN117692346A (en) | 2024-03-12 |
Family
ID=90139327
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202410130793.2A Pending CN117692346A (en) | 2024-01-31 | 2024-01-31 | Message blocking prediction method and device based on spectrum regularization variation self-encoder |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117692346A (en) |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108985330A (en) * | 2018-06-13 | 2018-12-11 | 华中科技大学 | A kind of autoencoder network and its training method abnormal use electric detection method and system |
CN109614992A (en) * | 2018-11-26 | 2019-04-12 | 北京工业大学 | A kind of batch process method for diagnosing faults based on DSSAE |
CN109948117A (en) * | 2019-03-13 | 2019-06-28 | 南京航空航天大学 | A kind of satellite method for detecting abnormality fighting network self-encoding encoder |
CN110444257A (en) * | 2019-08-05 | 2019-11-12 | 燕山大学 | It is a kind of based on unsupervised and supervised learning cement free calcium flexible measurement method |
CN110970099A (en) * | 2019-12-10 | 2020-04-07 | 北京大学 | Medicine molecule generation method based on regularization variational automatic encoder |
CN112633386A (en) * | 2020-12-26 | 2021-04-09 | 北京工业大学 | SACVAEGAN-based hyperspectral image classification method |
CN113180659A (en) * | 2021-01-11 | 2021-07-30 | 华东理工大学 | Electroencephalogram emotion recognition system based on three-dimensional features and cavity full convolution network |
CN114676622A (en) * | 2022-02-17 | 2022-06-28 | 国网冀北电力有限公司承德供电公司 | Short-term photovoltaic power prediction method based on self-encoder deep learning model |
WO2023138560A1 (en) * | 2022-01-24 | 2023-07-27 | 北京字跳网络技术有限公司 | Stylized image generation method and apparatus, electronic device, and storage medium |
CN117074961A (en) * | 2023-08-21 | 2023-11-17 | 广东恒翼能科技股份有限公司 | Variation self-encoder power battery abnormality detection method, system, device and storage medium |
CN117354652A (en) * | 2023-12-04 | 2024-01-05 | 四川省光为通信有限公司 | High-speed multimode optical module performance optimization method based on variation self-encoder |
-
2024
- 2024-01-31 CN CN202410130793.2A patent/CN117692346A/en active Pending
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108985330A (en) * | 2018-06-13 | 2018-12-11 | 华中科技大学 | A kind of autoencoder network and its training method abnormal use electric detection method and system |
CN109614992A (en) * | 2018-11-26 | 2019-04-12 | 北京工业大学 | A kind of batch process method for diagnosing faults based on DSSAE |
CN109948117A (en) * | 2019-03-13 | 2019-06-28 | 南京航空航天大学 | A kind of satellite method for detecting abnormality fighting network self-encoding encoder |
CN110444257A (en) * | 2019-08-05 | 2019-11-12 | 燕山大学 | It is a kind of based on unsupervised and supervised learning cement free calcium flexible measurement method |
CN110970099A (en) * | 2019-12-10 | 2020-04-07 | 北京大学 | Medicine molecule generation method based on regularization variational automatic encoder |
CN112633386A (en) * | 2020-12-26 | 2021-04-09 | 北京工业大学 | SACVAEGAN-based hyperspectral image classification method |
CN113180659A (en) * | 2021-01-11 | 2021-07-30 | 华东理工大学 | Electroencephalogram emotion recognition system based on three-dimensional features and cavity full convolution network |
WO2023138560A1 (en) * | 2022-01-24 | 2023-07-27 | 北京字跳网络技术有限公司 | Stylized image generation method and apparatus, electronic device, and storage medium |
CN114676622A (en) * | 2022-02-17 | 2022-06-28 | 国网冀北电力有限公司承德供电公司 | Short-term photovoltaic power prediction method based on self-encoder deep learning model |
CN117074961A (en) * | 2023-08-21 | 2023-11-17 | 广东恒翼能科技股份有限公司 | Variation self-encoder power battery abnormality detection method, system, device and storage medium |
CN117354652A (en) * | 2023-12-04 | 2024-01-05 | 四川省光为通信有限公司 | High-speed multimode optical module performance optimization method based on variation self-encoder |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112131272B (en) | Method, device, equipment and storage medium for detecting multi-element KPI time sequence | |
CN111177792B (en) | Method and device for determining target business model based on privacy protection | |
WO2023123941A1 (en) | Data anomaly detection method and apparatus | |
CN111880998B (en) | Service system anomaly detection method and device, computer equipment and storage medium | |
EP3862927A1 (en) | Anomaly detector, method of anomaly detection and method of training an anomaly detector | |
CN111708739B (en) | Abnormality detection method and device for time sequence data, electronic equipment and storage medium | |
CN111914873A (en) | Two-stage cloud server unsupervised anomaly prediction method | |
CN111967571A (en) | MHMA-based anomaly detection method and equipment | |
CN111178523A (en) | Behavior detection method and device, electronic equipment and storage medium | |
CN113114529B (en) | KPI (Key Performance indicator) anomaly detection method and device based on condition variation automatic encoder and computer storage medium | |
JP2022549031A (en) | Transformation of data samples into normal data | |
CN111782491A (en) | Disk failure prediction method, device, equipment and storage medium | |
CN116993537A (en) | Power load abnormality detection method and system based on serial GRU (generic routing unit) self-encoder | |
CN116821646A (en) | Data processing chain construction method, data reduction method, device, equipment and medium | |
CN117829209A (en) | Abnormal operation detection method, computing device and computer program for process equipment | |
US20230299788A1 (en) | Systems and Methods for Improved Machine-Learned Compression | |
CN114357858A (en) | Equipment deterioration analysis method and system based on multi-task learning model | |
CN117195044A (en) | Fault analysis method, device, computer equipment, storage medium and program product | |
CN117692346A (en) | Message blocking prediction method and device based on spectrum regularization variation self-encoder | |
CN115934484B (en) | Diffusion model data enhancement-based anomaly detection method, storage medium and apparatus | |
CN116522070A (en) | Non-supervision intelligent fault diagnosis method and system for mechanical parts | |
CN113111096B (en) | Abnormality detection method for high-dimensional time sequence working condition data of power generation equipment | |
CN114501029A (en) | Image encoding method, image decoding method, image encoding apparatus, image decoding apparatus, computer device, and storage medium | |
KR102184655B1 (en) | Improvement Of Regression Performance Using Asymmetric tanh Activation Function | |
CN112798949A (en) | Pumped storage unit generator temperature early warning method and system |
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 |