CN108287782A - A kind of multidimensional data method for detecting abnormality and device - Google Patents
A kind of multidimensional data method for detecting abnormality and device Download PDFInfo
- Publication number
- CN108287782A CN108287782A CN201710411852.3A CN201710411852A CN108287782A CN 108287782 A CN108287782 A CN 108287782A CN 201710411852 A CN201710411852 A CN 201710411852A CN 108287782 A CN108287782 A CN 108287782A
- Authority
- CN
- China
- Prior art keywords
- data
- multidimensional
- detected
- housebroken
- reconstruction model
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F11/00—Error detection; Error correction; Monitoring
- G06F11/30—Monitoring
- G06F11/34—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment
- G06F11/3447—Performance evaluation by modeling
Abstract
The invention discloses a kind of multidimensional data method for detecting abnormality and devices, are related to mobile communication big data field, the method includes:Multidimensional data to be detected is inputted into housebroken data reconstruction model;Using the housebroken data reconstruction model, data reconstruction is carried out to the multidimensional data to be detected, obtains multidimensional reconstruct data;Data are reconstructed according to the multidimensional, determine whether the multidimensional data to be detected is abnormal.The present invention be capable of detecting when the relationship between data characteristics extremely caused by data exception.
Description
Technical field
The present invention relates to mobile communication big data field, more particularly to a kind of multidimensional data method for detecting abnormality and device.
Background technology
The fast development and popularization and application of information technology and internet, data have penetrated into each current industry and industry
The functional field of business, becomes the important factor of production.It is high-quality there is directly contacting between the quality of data and Professional performance of enterprise
The data of amount can make company keep competitiveness and establish oneself in an unassailable position in economic turmoil period.There is pervasive data matter
Amount, enterprise can trust all data for meeting all demands at any time.In routine maintenance procedure, monitoring is very spine
The problem of hand, for many companies, service module is large number of, for each service or component, has various
Performance indicator, and when these achievement data scales grow to million, ten million rank, for there was only the engineering of tens or hundreds of people
For team of division, even if there is the diagrammatic representation of various beauties, it is also possible to spend longer time.
The appearance of conventional monitoring systems solves the problems, such as a part of, and engineer or expert can go to configure the different of some indexs
Normal threshold value, when data exceed threshold value, system will trigger alarm.But it is covered (no if some index is not monitored
Pipe is system or people), then it show exception after may nobody will appreciate that, cause not expected problem or failure.
In addition, many problems can be solved not manually given threshold, especially to ultra-large performance indicator progress
When monitoring, it is difficult to complete to monitor by human configuration.
The key problem of abnormality detection is how quickly and accurately to establish a detection model, by being built to the study of data
Vertical subjectivity detection model, some new theories can establish objective detection model by disclosing the rule of data itself.By study side
Method is applied to new theory in abnormality detection, for solving abnormality detection problems faced, improves accuracy in detection and has with speed
There is important meaning.
Common Abnormity detection method, such as statistical method, the outlier detection based on proximity, the outlier based on density
Detection, the technology etc. based on cluster all cannot have certain more stable between each dimension or feature of detection data well
Relationship, and when this relationship is abnormal caused data exception.The method of the present invention can be good at solving the problems, such as this.
Invention content
A kind of multidimensional data method for detecting abnormality and device provided in an embodiment of the present invention, can be in the multidimensional characteristic of data
Between have certain unknown relation when, solve the problems, such as the abnormality detection of data.
A kind of multidimensional data method for detecting abnormality provided according to embodiments of the present invention, including:
Multidimensional data to be detected is inputted into housebroken data reconstruction model;
Using the housebroken data reconstruction model, data reconstruction is carried out to the multidimensional data to be detected, is obtained more
Dimension reconstruct data;
Data are reconstructed according to the multidimensional, determine whether the multidimensional data to be detected is abnormal.
Preferably, before the housebroken data reconstruction model by multidimensional data to be detected input, further include:
Data reconstruction model is trained using multidimensional training data, determines one group of weight for making the data reconstruction model
Parameter of the parameter of structure loss function value minimum as the data reconstruction model, to obtain housebroken data reconstruction mould
Type.
Preferably, after obtaining the housebroken data reconstruction model, further include:
Each multidimensional training data is concentrated to determine error threshold with the error of its multidimensional reconstruct training data using training data
It is worth section.
Preferably, described to utilize the housebroken data reconstruction model, data are carried out to the multidimensional data to be detected
Reconstruct, obtaining multidimensional reconstruct data includes:
Using the housebroken data reconstruction model, coded treatment is carried out to the multidimensional data to be detected, is tieed up
The coded data that number declines, and the coded data declined to the dimension is decoded processing, obtains and the multidimensional to be detected
The identical multidimensional of data dimension reconstructs data.
Preferably, it is described according to the multidimensional reconstruct data, determine the multidimensional data to be detected whether include extremely:
If the multidimensional reconstruct data and the error of the multidimensional data to be detected be not in error threshold section, it is determined that
The multidimensional data to be detected is abnormal.
Preferably, the housebroken data reconstruction model is the neural network at least contained there are one hidden layer.
The storage medium provided according to embodiments of the present invention stores for realizing above-mentioned multidimensional data method for detecting abnormality
Program.
A kind of multidimensional data abnormal detector provided according to embodiments of the present invention, including:
Input module, for multidimensional data to be detected to be inputted housebroken data reconstruction model;
Reconstructed module, for utilizing the housebroken data reconstruction model, to the multidimensional data to be detected into line number
According to reconstruct, multidimensional reconstruct data are obtained;
Judging module determines whether the multidimensional data to be detected is abnormal for reconstructing data according to the multidimensional.
Preferably, the reconstructed module is specifically used for utilizing the housebroken data reconstruction model, to described to be detected
Multidimensional data carries out coded treatment, obtains the coded data of dimension decline, and the coded data declined to the dimension solves
Code processing obtains multidimensional identical with the multidimensional data dimension to be detected and reconstructs data.
Preferably, the judging module is not being missed in multidimensional reconstruct data and the error of the multidimensional data to be detected
When in poor threshold interval, determine that the multidimensional data to be detected is abnormal.
Preferably, the housebroken data reconstruction model is the neural network at least contained there are one hidden layer.
Technical solution provided in an embodiment of the present invention has the advantages that:
The embodiment of the present invention is capable of detecting when the extremely caused data exception problem of the relationship between the feature due to data,
Suitable for high dimensional data abnormality detection.
Description of the drawings
Fig. 1 is multidimensional data method for detecting abnormality block diagram provided in an embodiment of the present invention;
Fig. 2 is multidimensional data abnormal detector block diagram provided in an embodiment of the present invention;
Fig. 3 is the flow chart illustration for carrying out data exception detection;
Fig. 4 is the basic block diagram of sparse self-encoding encoder;
Fig. 5 is the structural schematic diagram of training data abnormality detection model.
Specific implementation mode
Below in conjunction with attached drawing to a preferred embodiment of the present invention will be described in detail, it should be understood that described below is excellent
Select embodiment only for the purpose of illustrating and explaining the present invention and is not intended to limit the present invention.
Fig. 1 is multidimensional data method for detecting abnormality block diagram provided in an embodiment of the present invention, as shown in Figure 1, step includes:
Step S101:Multidimensional data to be detected is inputted into housebroken data reconstruction model.
The housebroken data reconstruction model of step S101 is the neural network at least contained there are one hidden layer, implies the number of plies
Mesh is related with the dimension of input data.
Before executing step S101, data reconstruction model is trained using multidimensional training data, determines that one group makes
Parameter of the parameter of the reconstruct loss function value minimum of the data reconstruction model as the data reconstruction model, to obtain
Housebroken data reconstruction model.In other words, trained purpose is to obtain one group of parameter of neural network, under this group of parameter
When carrying out data reconstruction, reconstruct loss function value can be made minimum.
After obtaining housebroken data reconstruction model, concentrate each multidimensional training data more with it using training data
The error of dimension reconstruct training data, determines error threshold section.
The situation that the embodiment of the present invention contains a small amount of abnormal data to training dataset has robustness.
Step S102:Using the housebroken data reconstruction model, data weight is carried out to the multidimensional data to be detected
Structure obtains multidimensional reconstruct data.
Step S102 includes:Using the housebroken data reconstruction model, the multidimensional data to be detected is compiled
Code processing obtains the coded data of dimension decline, and the coded data declined to the dimension is decoded processing, obtains and institute
State the identical multidimensional reconstruct data of multidimensional data dimension to be detected.
Step S103:Data are reconstructed according to the multidimensional, determine whether the multidimensional data to be detected is abnormal.
Step S103 includes:If the multidimensional reconstruct data and the error of the multidimensional data to be detected be not in error threshold
In section, it is determined that the multidimensional data to be detected is abnormal.
The embodiment of the present invention can be detected and judge to the exception of data, and especially data have multidimensional characteristic, and
Data exception detection when between certain features with certain unknown relationship.Such as:Singly from the point of view of each dimension of data,
There is no problem for data, but data may have abnormal situation, the age such as people and height together Multidimensional Comprehensive,
Age from 0 to 100 between any one integer be likely to, height from 50 centimetres to 220 centimetre in any data have
May, but if the age of a people is 1 years old, it is exactly abnormal data that height, which is 210 centimetres,.This method is concentrated in training data
Containing a small amount of abnormal data, but which does not know when being abnormal data, can carry out the abnormality detection of data, i.e. this hair well yet
The method of bright embodiment has robustness.
It will appreciated by the skilled person that implement the method for the above embodiments be can be with
Relevant hardware is instructed to complete by program, the program can be stored in computer read/write memory medium, should
Program when being executed, including step S101 to step S103.Furtherly, the present invention can also provide a kind of storage medium,
On be stored with computer program, which at least realizes following steps when being executed by processor:Multidimensional data to be detected is inputted
Housebroken data reconstruction model;Using the housebroken data reconstruction model, to the multidimensional data to be detected into line number
According to reconstruct, multidimensional reconstruct data are obtained;Data are reconstructed according to the multidimensional, determine whether the multidimensional data to be detected is abnormal.
Wherein, the storage medium may include ROM/RAM, magnetic disc, CD, USB flash disk.
Fig. 2 is multidimensional data abnormal detector block diagram provided in an embodiment of the present invention, as shown in Fig. 2, including:
Input module, for multidimensional data to be detected to be inputted housebroken data reconstruction model.
Reconstructed module, for utilizing the housebroken data reconstruction model, to the multidimensional data to be detected into line number
According to reconstruct, multidimensional reconstruct data are obtained.
Judging module determines whether the multidimensional data to be detected is abnormal for reconstructing data according to the multidimensional.
When whether detection data is abnormal, input module inputs multidimensional data to be detected trained the device of the present embodiment
Data reconstruction model;Reconstructed module utilizes the housebroken data reconstruction model, successively to the multidimensional data to be detected
It is encoded, decoding process, obtains multidimensional reconstruct data;Judging module is in multidimensional reconstruct data and the multidimensional to be detected
When the error of data is not in error threshold section, determine that the multidimensional data to be detected is abnormal.
The data reconstruction model of the present embodiment is sparse self-encoding encoder, specifically at least contains the nerve net there are one hidden layer
Network.
Described device can further include:
Training module, for being trained to data reconstruction model using multidimensional training data, in other words, trained mesh
Be to obtain one group of parameter of neural network, under this group of parameter carry out data reconstruction when, reconstruct loss function value can be made most
It is small.
Threshold determination module, for after obtaining housebroken data reconstruction model, being concentrated using training data each
The error of multidimensional training data and its multidimensional reconstruct training data, determines error threshold section.
Described device before detection data exception, instruct data reconstruction model using multidimensional training data by training module
Practice, determines the parameter for the reconstruct loss function value minimum that one group makes the data reconstruction model as the data reconstruction model
Parameter, to obtain housebroken data reconstruction model;The threshold determination module utilizes housebroken data reconstruction model, obtains
The multidimensional of each multidimensional training data is concentrated to reconstruct training data to training data, and according to each multidimensional training data and its more
The error of dimension reconstruct training data, determines error threshold section.
The embodiment of the present invention also provides a kind of multidimensional data abnormal detector, including memory and processor, wherein institute
Processor is stated in program or the instruction for executing the memory storage, at least realizes the following contents:By multidimensional data to be detected
Input housebroken data reconstruction model;Using the housebroken data reconstruction model, to the multidimensional data to be detected into
Row data reconstruction obtains multidimensional reconstruct data;Data are reconstructed according to the multidimensional, determine whether the multidimensional data to be detected is different
Often.
Fig. 3 is the flow chart illustration for carrying out data exception detection, as shown in Figure 3.
It before detection data exception, needs to be trained data reconstruction model, then determines error threshold section, specifically
Calculating process includes:
1, data prediction pre-processes training dataset, including removal invalid value, is standardized to data
Deng.
2, model training trains sparse own coding model (i.e. data reconstruction model, sparse using pretreated data
Self-encoding encoder), specifically, being the parameter of trained sparse own coding model.
3, the calculating in error threshold section utilizes the data of trained model reconstruction training dataset, calculate training
Error between the corresponding reconstruct data of every data in data set, is then determined using box traction substation or confidence interval
The threshold interval of error.
When whether detection data is abnormal, it is divided into data prediction, data reconstruction, anomalous discrimination three phases, it is specific to count
Calculation process includes:
1, data prediction pre-processes data to be detected, including removal invalid value, data are standardized
Deng.
2, data reconstruction enters data into sparse self-encoding encoder the decoding of the coding and data that carry out data.Its
In, sparse self-encoding encoder is the neural network at least contained there are one hidden layer, and the parameters of network can be obtained by training data.
3, anomalous discrimination, judge data to be discriminated (data i.e. to be detected) and its reconstruct the error between data whether
In error threshold section, if in section, the data are normal, conversely, the data exception.
That is, being first standardized data to be detected, is then encoded, solved by sparse self-encoding encoder
Code obtains reconstruct data, and whether the error of data and reconstruct data after normalized finally compares the error in threshold zone
In, if if the data it is normal, it is otherwise abnormal.
The present embodiment is suitable for the data exception detection of multiple features, usually three or three or more features;It is dilute
The half of input feature vector can be generally set as by dredging the number of hidden layer in self-encoding encoder.
Fig. 4 is the basic block diagram of sparse self-encoding encoder, as shown in figure 4, it is by input layer, hidden layer and output layer three
Layer composition.
N feature of input layer corresponds to a n-dimensional vector x=(x1,x2,…,xn), that is, the dimension of feature and input to
The dimension of amount is consistent, and the circle for putting on+1 is referred to as bias node.
The feature of the Feature Mapping of input layer to hidden layer, hidden layer is denoted as a k dimensional vector h=(h1,h2,…,
hk), then the mapping can be expressed as h=f (w1x+b1).In formula, w1,b1Input layer is expressed as between the unit of hidden layer
Connection parameter and bias term parameter, can be trained by training set data come, f be activation letter (usually using Sigmoid or
Tanh functions), the dimension of hidden layer is traditionally arranged to be the half of input layer feature (if decimal, then in Outlier Detection Algorithm
It rounds up).Hidden layer expression h obtains reconstructing vectorial y=(y further through mapping1,y2,…,yn), which is expressed as y=g
(w2h+b2).W in formula2,b2Hidden layer is expressed as to connection parameter and bias term parameter between the unit of output layer, it can
It is trained by training set data come g is activation primitive.
Fig. 5 is the structural schematic diagram of training data abnormality detection model, as shown in Figure 5, it is necessary first to by the data of input
It is standardized.Autoencoder network (i.e. sparse self-encoding encoder) can be regarded as and compress data, then need again
When by lose it is small as possible in the way of data are recovered.
Without loss of generality, if training sample number is m,WithIt indicates respectively
The i-th sample after standardization and xiValue after reconstruct, hjFor the output of j-th of neuron of hidden layer, exported for expression
It is related to some input, it is denoted as hj(xi), it enablesIt is j-th of unit of hidden layer to the defeated of all input samples
Go out average value, then can be lost with the reconstruct of following function representation sample data:
Wherein, penalty isM is training sample number, ρ
For Sparse parameter, usually replaced (such as ρ=0.05) close to 0 constant with one, W=(w1,b1,w2,b2), λ is non-negative normal
Number, referred to as regularization parameter.Then pretreated data are input to autoencoder network, train one group of parameter w1,b1,w2,
b2, keep loss function value above minimum, to learn a mapping from input layer to output layer, that is, it is of the invention different
Reconfiguration system (i.e. data reconstruction model) in normal detecting system.
Loss function by training data and above is learnt by the method that gradient declines to parameter w1,b1,w2,b2It
Afterwards, the design in error threshold section is carried out.The error for calculating each training sample and the sample after reconstruct first, then utilizes
Confidence interval or box traction substation calculate the threshold interval of the data set of m error composition.For example, calculate each training sample with
Mean square error (the y of sample after reconstructi-xi)2, then calculate mean μ and the side of the data set being made of this m mean square error
Poor σ can thus obtain the mean square error threshold interval [+3 σ of μ -3 σ, μ] of normal data.
In conclusion the embodiment of the present invention has the following technical effects:
The multidimensional data of the embodiment of the present invention is the data for having multiple features, through the above steps, be capable of detecting when by
Relationship between data characteristics is abnormal (for example, the ratio between data characteristics is more stable, and between the feature of some data
Ratio be very different with what overall data was showed, then the data just be relationship abnormal data) caused by data it is different
Chang Wenti, and need not understand and know in it physical relationship.
Although describing the invention in detail above, but the invention is not restricted to this, those skilled in the art of the present technique
It can be carry out various modifications with principle according to the present invention.Therefore, all to be changed according to made by the principle of the invention, all it should be understood as
Fall into protection scope of the present invention.
Claims (10)
1. a kind of multidimensional data method for detecting abnormality, including:
Multidimensional data to be detected is inputted into housebroken data reconstruction model;
Using the housebroken data reconstruction model, data reconstruction is carried out to the multidimensional data to be detected, obtains multidimensional weight
Structure data;
Data are reconstructed according to the multidimensional, determine whether the multidimensional data to be detected is abnormal.
2. according to the method described in claim 1, multidimensional data to be detected is inputted housebroken data reconstruction model described
Before, further include:
Data reconstruction model is trained using multidimensional training data, determines that one group makes the reconstruct of the data reconstruction model damage
Parameter of the parameter of functional value minimum as the data reconstruction model is lost, to obtain housebroken data reconstruction model.
3. according to the method described in claim 2, after obtaining the housebroken data reconstruction model, further include:
Each multidimensional training data is concentrated to determine error threshold area with the error of its multidimensional reconstruct training data using training data
Between.
4. according to the method described in claim 1, described utilize the housebroken data reconstruction model, to described to be detected more
Dimension data carries out data reconstruction, obtains multidimensional reconstruct data and includes:
Using the housebroken data reconstruction model, coded treatment is carried out to the multidimensional data to be detected, is obtained under dimension
The coded data of drop, and the coded data declined to the dimension is decoded processing, obtains and the multidimensional data to be detected
The identical multidimensional of dimension reconstructs data.
5. according to the method described in claim 3, described reconstruct data according to the multidimensional, the multidimensional data to be detected is determined
Whether exception includes:
If the multidimensional reconstruct data and the error of the multidimensional data to be detected be not in error threshold section, it is determined that described
Multidimensional data to be detected is abnormal.
6. according to the method described in claim 1-5 any one, the housebroken data reconstruction model is at least to contain one
The neural network of a hidden layer.
7. a kind of multidimensional data abnormal detector, including:
Input module, for multidimensional data to be detected to be inputted housebroken data reconstruction model;
Reconstructed module carries out data weight for utilizing the housebroken data reconstruction model to the multidimensional data to be detected
Structure obtains multidimensional reconstruct data;
Judging module determines whether the multidimensional data to be detected is abnormal for reconstructing data according to the multidimensional.
8. device according to claim 7, the reconstructed module is specifically used for utilizing the housebroken data reconstruction mould
Type carries out coded treatment to the multidimensional data to be detected, obtains the coded data of dimension decline, and the dimension is declined
Coded data is decoded processing, obtains multidimensional identical with the multidimensional data dimension to be detected and reconstructs data.
9. device according to claim 7, the judging module is in multidimensional reconstruct data and the multidimensional to be detected
When the error of data is not in error threshold section, determine that the multidimensional data to be detected is abnormal.
10. according to the device described in claim 7-9 any one, the housebroken data reconstruction model is at least to contain one
The neural network of a hidden layer.
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710411852.3A CN108287782A (en) | 2017-06-05 | 2017-06-05 | A kind of multidimensional data method for detecting abnormality and device |
PCT/CN2018/098540 WO2018224055A2 (en) | 2017-06-05 | 2018-08-03 | Multi-dimensional data abnormality detection method and apparatus |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710411852.3A CN108287782A (en) | 2017-06-05 | 2017-06-05 | A kind of multidimensional data method for detecting abnormality and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108287782A true CN108287782A (en) | 2018-07-17 |
Family
ID=62831591
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710411852.3A Pending CN108287782A (en) | 2017-06-05 | 2017-06-05 | A kind of multidimensional data method for detecting abnormality and device |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN108287782A (en) |
WO (1) | WO2018224055A2 (en) |
Cited By (23)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109145595A (en) * | 2018-07-31 | 2019-01-04 | 顺丰科技有限公司 | A kind of user's unusual checking system, method, equipment and storage medium |
WO2018224055A3 (en) * | 2017-06-05 | 2019-01-24 | 中兴通讯股份有限公司 | Multi-dimensional data abnormality detection method and apparatus |
CN109492193A (en) * | 2018-12-28 | 2019-03-19 | 同济大学 | Abnormal network data based on depth machine learning model generate and prediction technique |
CN109657947A (en) * | 2018-12-06 | 2019-04-19 | 西安交通大学 | A kind of method for detecting abnormality towards enterprises ' industry classification |
CN109714340A (en) * | 2018-12-28 | 2019-05-03 | 厦门服云信息科技有限公司 | The Network Abnormal of a kind of sequence to sequence requests recognition methods and device |
CN109753372A (en) * | 2018-12-20 | 2019-05-14 | 东软集团股份有限公司 | Multidimensional data method for detecting abnormality, device, readable storage medium storing program for executing and electronic equipment |
CN110033014A (en) * | 2019-01-08 | 2019-07-19 | 阿里巴巴集团控股有限公司 | The detection method and its system of abnormal data |
CN110503190A (en) * | 2019-08-13 | 2019-11-26 | 上海华力集成电路制造有限公司 | The method for detecting abnormality of various dimensions process data in semiconductor board process |
CN110798330A (en) * | 2018-08-01 | 2020-02-14 | 中国移动通信集团浙江有限公司 | Telecommunication fraud library updating processing method and device |
CN111104880A (en) * | 2019-12-09 | 2020-05-05 | 北京国网富达科技发展有限责任公司 | Method, device and system for processing cable tunnel state data |
CN111241688A (en) * | 2020-01-15 | 2020-06-05 | 北京百度网讯科技有限公司 | Method and device for monitoring composite production process |
CN111275288A (en) * | 2019-12-31 | 2020-06-12 | 华电国际电力股份有限公司十里泉发电厂 | XGboost-based multi-dimensional data anomaly detection method and device |
CN111444233A (en) * | 2020-02-15 | 2020-07-24 | 中国环境监测总站 | Method for discovering environmental monitoring abnormal data based on duplicator neural network model |
CN111767275A (en) * | 2020-06-28 | 2020-10-13 | 北京林克富华技术开发有限公司 | Data processing method and device and data processing system |
CN112203272A (en) * | 2019-07-08 | 2021-01-08 | 中国移动通信集团浙江有限公司 | HSS user relocation abnormity diagnosis method and device and computing equipment |
CN112202625A (en) * | 2019-07-08 | 2021-01-08 | 中国移动通信集团浙江有限公司 | Network element abnormity diagnosis method and device, computing equipment and computer storage medium |
CN112261018A (en) * | 2020-10-13 | 2021-01-22 | 中国光大银行股份有限公司 | Abnormal object detection method and device, storage medium and electronic device |
CN112308104A (en) * | 2019-08-02 | 2021-02-02 | 杭州海康威视数字技术股份有限公司 | Abnormity identification method and device and computer storage medium |
CN112306835A (en) * | 2020-11-02 | 2021-02-02 | 平安科技(深圳)有限公司 | User data monitoring and analyzing method, device, equipment and medium |
CN112801497A (en) * | 2021-01-26 | 2021-05-14 | 上海华力微电子有限公司 | Anomaly detection method and device |
CN113657516A (en) * | 2021-08-20 | 2021-11-16 | 泰康保险集团股份有限公司 | Method and device for processing medical transaction data, electronic equipment and storage medium |
TWI749586B (en) * | 2020-06-11 | 2021-12-11 | 華碩電腦股份有限公司 | Signal detection method and electronic device using the same |
CN112801497B (en) * | 2021-01-26 | 2024-04-30 | 上海华力微电子有限公司 | Abnormality detection method and device |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104914851A (en) * | 2015-05-21 | 2015-09-16 | 北京航空航天大学 | Adaptive fault detection method for airplane rotation actuator driving device based on deep learning |
CN106555788A (en) * | 2016-11-11 | 2017-04-05 | 河北工业大学 | Application of the deep learning based on Fuzzy Processing in hydraulic equipment fault diagnosis |
CN106779069A (en) * | 2016-12-08 | 2017-05-31 | 国家电网公司 | A kind of abnormal electricity consumption detection method based on neutral net |
Family Cites Families (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8306931B1 (en) * | 2009-08-06 | 2012-11-06 | Data Fusion & Neural Networks, LLC | Detecting, classifying, and tracking abnormal data in a data stream |
CN102821002B (en) * | 2011-06-09 | 2015-08-26 | 中国移动通信集团河南有限公司信阳分公司 | Network flow abnormal detecting method and system |
CN108287782A (en) * | 2017-06-05 | 2018-07-17 | 中兴通讯股份有限公司 | A kind of multidimensional data method for detecting abnormality and device |
-
2017
- 2017-06-05 CN CN201710411852.3A patent/CN108287782A/en active Pending
-
2018
- 2018-08-03 WO PCT/CN2018/098540 patent/WO2018224055A2/en active Application Filing
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104914851A (en) * | 2015-05-21 | 2015-09-16 | 北京航空航天大学 | Adaptive fault detection method for airplane rotation actuator driving device based on deep learning |
CN106555788A (en) * | 2016-11-11 | 2017-04-05 | 河北工业大学 | Application of the deep learning based on Fuzzy Processing in hydraulic equipment fault diagnosis |
CN106779069A (en) * | 2016-12-08 | 2017-05-31 | 国家电网公司 | A kind of abnormal electricity consumption detection method based on neutral net |
Cited By (31)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2018224055A3 (en) * | 2017-06-05 | 2019-01-24 | 中兴通讯股份有限公司 | Multi-dimensional data abnormality detection method and apparatus |
CN109145595A (en) * | 2018-07-31 | 2019-01-04 | 顺丰科技有限公司 | A kind of user's unusual checking system, method, equipment and storage medium |
CN110798330A (en) * | 2018-08-01 | 2020-02-14 | 中国移动通信集团浙江有限公司 | Telecommunication fraud library updating processing method and device |
CN109657947A (en) * | 2018-12-06 | 2019-04-19 | 西安交通大学 | A kind of method for detecting abnormality towards enterprises ' industry classification |
CN109753372A (en) * | 2018-12-20 | 2019-05-14 | 东软集团股份有限公司 | Multidimensional data method for detecting abnormality, device, readable storage medium storing program for executing and electronic equipment |
CN109492193B (en) * | 2018-12-28 | 2020-11-27 | 同济大学 | Abnormal network data generation and prediction method based on deep machine learning model |
CN109714340A (en) * | 2018-12-28 | 2019-05-03 | 厦门服云信息科技有限公司 | The Network Abnormal of a kind of sequence to sequence requests recognition methods and device |
CN109492193A (en) * | 2018-12-28 | 2019-03-19 | 同济大学 | Abnormal network data based on depth machine learning model generate and prediction technique |
WO2020143379A1 (en) * | 2019-01-08 | 2020-07-16 | 阿里巴巴集团控股有限公司 | Abnormal data detection method and system |
CN110033014A (en) * | 2019-01-08 | 2019-07-19 | 阿里巴巴集团控股有限公司 | The detection method and its system of abnormal data |
CN112202625A (en) * | 2019-07-08 | 2021-01-08 | 中国移动通信集团浙江有限公司 | Network element abnormity diagnosis method and device, computing equipment and computer storage medium |
CN112202625B (en) * | 2019-07-08 | 2023-08-15 | 中国移动通信集团浙江有限公司 | Network element abnormality diagnosis method, device, computing equipment and computer storage medium |
CN112203272B (en) * | 2019-07-08 | 2023-09-08 | 中国移动通信集团浙江有限公司 | Abnormality diagnosis method, device and computing equipment for moving HSS (home subscriber server) user |
CN112203272A (en) * | 2019-07-08 | 2021-01-08 | 中国移动通信集团浙江有限公司 | HSS user relocation abnormity diagnosis method and device and computing equipment |
CN112308104A (en) * | 2019-08-02 | 2021-02-02 | 杭州海康威视数字技术股份有限公司 | Abnormity identification method and device and computer storage medium |
CN110503190A (en) * | 2019-08-13 | 2019-11-26 | 上海华力集成电路制造有限公司 | The method for detecting abnormality of various dimensions process data in semiconductor board process |
CN111104880A (en) * | 2019-12-09 | 2020-05-05 | 北京国网富达科技发展有限责任公司 | Method, device and system for processing cable tunnel state data |
CN111275288B (en) * | 2019-12-31 | 2023-12-26 | 华电国际电力股份有限公司十里泉发电厂 | XGBoost-based multidimensional data anomaly detection method and device |
CN111275288A (en) * | 2019-12-31 | 2020-06-12 | 华电国际电力股份有限公司十里泉发电厂 | XGboost-based multi-dimensional data anomaly detection method and device |
CN111241688B (en) * | 2020-01-15 | 2023-08-25 | 北京百度网讯科技有限公司 | Method and device for monitoring composite production process |
CN111241688A (en) * | 2020-01-15 | 2020-06-05 | 北京百度网讯科技有限公司 | Method and device for monitoring composite production process |
CN111444233A (en) * | 2020-02-15 | 2020-07-24 | 中国环境监测总站 | Method for discovering environmental monitoring abnormal data based on duplicator neural network model |
TWI749586B (en) * | 2020-06-11 | 2021-12-11 | 華碩電腦股份有限公司 | Signal detection method and electronic device using the same |
CN111767275A (en) * | 2020-06-28 | 2020-10-13 | 北京林克富华技术开发有限公司 | Data processing method and device and data processing system |
CN111767275B (en) * | 2020-06-28 | 2024-04-19 | 北京林克富华技术开发有限公司 | Data processing method and device and data processing system |
CN112261018B (en) * | 2020-10-13 | 2023-01-31 | 中国光大银行股份有限公司 | Abnormal object detection method and device, storage medium and electronic device |
CN112261018A (en) * | 2020-10-13 | 2021-01-22 | 中国光大银行股份有限公司 | Abnormal object detection method and device, storage medium and electronic device |
CN112306835A (en) * | 2020-11-02 | 2021-02-02 | 平安科技(深圳)有限公司 | User data monitoring and analyzing method, device, equipment and medium |
CN112801497A (en) * | 2021-01-26 | 2021-05-14 | 上海华力微电子有限公司 | Anomaly detection method and device |
CN112801497B (en) * | 2021-01-26 | 2024-04-30 | 上海华力微电子有限公司 | Abnormality detection method and device |
CN113657516A (en) * | 2021-08-20 | 2021-11-16 | 泰康保险集团股份有限公司 | Method and device for processing medical transaction data, electronic equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
WO2018224055A2 (en) | 2018-12-13 |
WO2018224055A3 (en) | 2019-01-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108287782A (en) | A kind of multidimensional data method for detecting abnormality and device | |
Izakian et al. | Anomaly detection and characterization in spatial time series data: A cluster-centric approach | |
CN109086805B (en) | Clustering method based on deep neural network and pairwise constraints | |
CN112699960B (en) | Semi-supervised classification method, equipment and storage medium based on deep learning | |
CN109657947A (en) | A kind of method for detecting abnormality towards enterprises ' industry classification | |
CN111401433A (en) | User information acquisition method and device, electronic equipment and storage medium | |
CN107730040A (en) | Power information system log information comprehensive characteristics extracting method and device based on RBM | |
Mustika et al. | Analysis accuracy of xgboost model for multiclass classification-a case study of applicant level risk prediction for life insurance | |
Long et al. | A new approach for construction of geodemographic segmentation model and prediction analysis | |
CN114399321A (en) | Business system stability analysis method, device and equipment | |
Sina Mirabdolbaghi et al. | Model optimization analysis of customer churn prediction using machine learning algorithms with focus on feature reductions | |
CN110766100B (en) | Bearing fault diagnosis model construction method, bearing fault diagnosis method and electronic equipment | |
CN113283546B (en) | Furnace condition abnormity alarm method and system of heating furnace integrity management centralized control device | |
CN109214401A (en) | SAR image classification method and device based on stratification autocoder | |
CN114298299A (en) | Model training method, device, equipment and storage medium based on course learning | |
US11928017B2 (en) | Point anomaly detection | |
KR102653259B1 (en) | Method for extracting heart rate variability feature value | |
CN114880538A (en) | Attribute graph community detection method based on self-supervision | |
Sindhu et al. | Disaster management from social media using machine learning | |
CN114418189A (en) | Water quality grade prediction method, system, terminal device and storage medium | |
Huo et al. | Sparse embedding for interpretable hospital admission prediction | |
CN113159419A (en) | Group feature portrait analysis method, device and equipment and readable storage medium | |
Nsofor | Comparative analysis of predictive data-mining techniques | |
Rianto et al. | Determining the Eligibility of Providing Motorized Vehicle Loans by Using the Logistic Regression, Naive Bayes and Decission Tree (C4. 5) | |
Wang | Financial distress prediction for listed enterprises using fuzzy C-means |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180717 |
|
RJ01 | Rejection of invention patent application after publication |