CN109543765A - A kind of industrial data denoising method based on improvement IForest - Google Patents
A kind of industrial data denoising method based on improvement IForest Download PDFInfo
- Publication number
- CN109543765A CN109543765A CN201811439128.2A CN201811439128A CN109543765A CN 109543765 A CN109543765 A CN 109543765A CN 201811439128 A CN201811439128 A CN 201811439128A CN 109543765 A CN109543765 A CN 109543765A
- Authority
- CN
- China
- Prior art keywords
- data
- iforest
- abnormal
- value
- buffer area
- 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
- 238000000034 method Methods 0.000 title claims abstract description 56
- 230000006872 improvement Effects 0.000 title description 3
- 230000002159 abnormal effect Effects 0.000 claims abstract description 39
- 238000001514 detection method Methods 0.000 claims abstract description 23
- 230000005856 abnormality Effects 0.000 claims abstract description 14
- 239000000523 sample Substances 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 9
- 241001269238 Data Species 0.000 claims description 6
- 238000004422 calculation algorithm Methods 0.000 claims description 6
- 239000012723 sample buffer Substances 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 2
- 238000010276 construction Methods 0.000 claims description 2
- 238000012360 testing method Methods 0.000 claims description 2
- 238000012549 training Methods 0.000 abstract description 6
- 230000000694 effects Effects 0.000 abstract description 5
- 238000012545 processing Methods 0.000 abstract description 4
- 238000010801 machine learning Methods 0.000 abstract description 3
- 230000008901 benefit Effects 0.000 abstract description 2
- 238000013135 deep learning Methods 0.000 abstract description 2
- 238000005516 engineering process Methods 0.000 abstract description 2
- 238000013450 outlier detection Methods 0.000 abstract description 2
- 238000002955 isolation Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000004140 cleaning Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 239000000470 constituent Substances 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000012067 mathematical method Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000001932 seasonal effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Probability & Statistics with Applications (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention is a kind of to be related to the industrial data denoising method based on deep learning based on the industrial data denoising method for improving IForest, belongs to the application fields such as big data processing and machine learning.Include the following steps: 1) building initially isolated forest IForest;2) online abnormality detection;3) judge whether to need to update detector.Compared to traditional IForset abnormal point detecting method, the present invention has the advantage that first is that not using the threshold value of single numeric type using the threshold setting method arrived when proposing outlier detection based on global isolated forest;Second is that using buffer technology to store sampled data, isolated forest is updated with timing, improves the accuracy of training pattern.Third is that the set time is updated the training data of buffer area, to eliminate excessively outmoded data, new data is stored, to improve the abnormality detection effect to data at hand.
Description
Technical field
A kind of industrial data being related to based on the industrial data denoising method for improving IForest based on deep learning of the present invention
Denoising method belongs to the application fields such as big data processing and machine learning.
Background technique
Big data and machine learning have become the new trend of computer nowadays research industry, and major IT enterprises are released one after another
Commercialized platform and its big data processing scheme.In numerous data, due to data acquisition or misregistration, data itself
Intrinsic variation, inevitable accidentalia, all may cause the generation of abnormal data.There are the obvious spies of noise in data
Sign is that wherein there is abnormal datas.Because of the particularity of this data, it is believed that it is not random deviation, but with it is other
Data are not belonging to same mechanism.By being detected to this data, can effectively find present in current industrial data
Problem carries out data cleansing or timely early warning, reduces the risk or accident of industrial equipment operation, can also set to currently used
The fields such as the standby following operating condition, anti-cheating, pseudo-base station and financial swindling provide directive significance.It must in data prediction
These abnormal datas must be handled, just can ensure that the processing accuracy of next step data, therefore abnormality detection is also data
The crucial research direction of one excavated.
There is currently the algorithms of a variety of data de-noisings, mainly include Statistics-Based Method, the method based on propinquity, base
Method in cluster and the method based on time series etc..Various methods have the scope of application of oneself.Based on statistical
Method be by training one data-oriented collection probabilistic model, using in the model be lower than probability tend in object as extremely
Value.Statistical method can be subdivided into parametric technique and nonparametric technique, and common parametric technique needs first to assume normal parameter
It is to obey certain distribution according to some parameter to generate, common parametric technique has method based on Gauss model, based on returning
The method etc. of model.And nonparametric technique does not need to assume that data meet certain probabilistic model in advance, but according to the number of input
According to final probabilistic model is established, wherein common nonparametric technique has the method based on histogram and the side based on kernel estimates
Method etc..Neighborhood density where method based on propinquity assumes normal value object is higher, and abnormal data is often in density
Lower neighborhood, there are significantly different with normal data.Such algorithm needs the neighborhood to each data in the detection process
It is searched, so its corresponding time complexity is very high.Method based on cluster is according to data similarity, by set of metadata of similar data
It is divided into the same cluster, the similarity for needing to meet data in cluster is big as far as possible, and the similarity of data is as far as possible between cluster
It is small.Usually assume that normal data belongs to the high cluster of the big degree of approximation of range, and abnormal data then belongs to the low cluster of the small degree of approximation of range.
When in this way, usually the degree of peeling off of each data is calculated, and this degree of peeling off tends to rely on data point
The size and cluster central point distance of affiliated cluster.Method based on time series is to arrange the data obtained with time change
Column combine, and are nursed using mathematical method or the method for data mining to it, right to find out wherein implicit meaning
Follow-up work provides safeguard.But time data are often with apparent time response, constituent also different from general data,
It usually will appear seasonal variations, circulation change or irregular variation etc..
And the IForest based on statistical modeling has linear time complexity, can be used in the data containing mass data
Collection is above.The quantity usually set is more, and algorithm is more stable.Since each tree all independently generates mutually, can dispose
Accelerate operation on large scale distributed system.Traditional IForest method is fewer in terms of industrial data cleaning,
It is simple threshold values to be set, to judge number currently entered on the basis of establishing isolated forest when handling other abnormal datas
According to whether in abnormal.It has ignored as time goes by, the continuous arrival of new data, the isolated forest originally established can also occur
Variation will lead to this wrong feature of Result.
Although it is sentenced in conclusion currently used traditional IForest method can be used in industrial data cleaning
Fixed abnormal threshold value is excessively single, and does not account for as the model that new data arrival was established originally should be updated,
To ensure this characteristic of the accuracy of abnormal data result of lookup.
In the application of industrial equipment, the data acquired in real time would generally do Feature Engineering comprising some abnormal datas
It needs to analyze the data being related to before with model training, removes abnormal data therein.
Therefore the industrial data denoising method that a kind of effect is good, high-efficient is needed.
Summary of the invention
It is a kind of based on the industrial data denoising for improving IForest it is an object of the invention to provide in view of the above shortcomings
Method, setting are updated original IForest, with the isolated relevant threshold value of forest of the overall situation, and as time goes by mention
The accuracy that high abnormal data is searched, convenient for quickly and effectively being denoised to large-scale industry data.In solving industrial data
There are when abnormal data, abnormal data therein can be judged and be searched, while guaranteeing to search accuracy, be improved
The speed of lookup.
The present invention adopts the following technical solutions to achieve:
Based on the industrial data denoising method for improving IForest, include the following steps,
1) forest IForest is initially isolated in building;
1-1) building tree iTree;
Isolated forest IForest(Isolation Forest) it is to be constituted by largely setting iTree;ITree is one kind random two
Fork tree, each node or there are two children or be exactly leaf node;All properties are all in given history data set D, D
The composition process of the variable of continuous type, iTree is as follows:
An attribute Qi 1-1-1) is randomly choosed in given history data set D;
A value q of the attribute 1-1-2) is randomly choosed, the value is between maximum and minimum value;
1-1-3) the attribute Qi according to step 1-1-1) classifies to every record, attribute Qi value is less than the note of q
Recording playback is placed on right child in left child, the record attribute Qi value more than or equal to q;
1-1-4) the left child of recursive construction and right child only have a record or a plurality of one until meeting incoming data set
The record of sample or the height of tree have reached restriction height l;Modify the number of path N of isolated forest, initial value 0.
1-2) data generated because of industrial equipment are diversified, and every kind of data can all generate an iTree, by institute
Some trees iTree has combined to form initial isolated forest IForest.The isolated forest can be used as initial abnormal inspection
Survey device.
2) online abnormality detection;
To the data of each arrival, its corresponding different types of data is put into the isolated forest built up in step 1)
In, unusual condition is judged, if being higher than preset threshold value, explanation according to the abnormal score that input data mean depth obtains
The arrival data do not reach general level, are abnormal datas;
3) judge whether to need to update detector;
Calculated according to user's application scale predetermined, if storage sample buffer area expired, to detector into
Row updates.According to principle of locality, for upcoming data, new data is more more valuable than legacy data, and therefore, setting is fixed
Time legacy data in buffer area is carried out it is superseded, so as to new data arrival when facilitate storage, improve the accuracy of abnormality detection.
In step 2, when the data of the arrival reach, it to be also based on Poisson distribution, judges whether the sample is used as more
New samples are added to the buffer area for setting storage sample data, and in the buffer area, sample data is suitable according to time order and function
Sequence arrangement.
Compared to traditional IForset abnormal point detecting method, the present invention has the advantage that first is that being isolated based on global
Using the threshold setting method arrived when forest proposes outlier detection, the threshold value of single numeric type is not used;Second is that using
Buffer technology stores sampled data, updates isolated forest with timing, improves the accuracy of training pattern.Third is that the set time
The training data of buffer area is updated, to eliminate excessively outmoded data, stores new data, to improve at hand
The abnormality detection effect of data.
Detailed description of the invention
Below with reference to attached drawing, the invention will be further described:
Fig. 1 is the step flow chart of the method for the present invention;
Fig. 2 is transformer data exception detection process flow chart in embodiment of the present invention method.
Specific embodiment
Below by embodiment, in conjunction with the realization process of attached drawing present invention be described in more detail method.
Fig. 1 shows the step process of the method for the present invention, based on the industrial data denoising method for improving IForest, including
Following steps:
1) forest IForest is initially isolated in building;
2) online abnormality detection;
To the data of each arrival, its corresponding different types of data is put into the isolated forest built up, is judged different
Normal situation illustrates the arrival number if being higher than preset threshold value according to the abnormal score that input data mean depth obtains
It is abnormal data according to general level is not reached;
The main task of the step is to calculate the abnormal score of data record x, and abnormal scoring function is as follows,
, in which:
Wherein N is the number of passes in IForest from root to leaf node, and M is sample set scale, and avgdepth (x) is forest
Average isolation depth (isolation depth smaller, easier to be isolated, i.e. x is that the probability of abnormal data is bigger), depth (x,
Itree data x) is indicated from tree root to the traversal depth for being isolated node, and c (M) is the isolation for the data set being made of M sample
The desired value of depth is set, H (M) is harmonic number, can estimate to obtain with Euler's constant, i.e. H (M)=ln (M)+0.572156649.
It can be seen that working asWhen (i.e. the average isolation depth of forest level off to 0),(i.e. the abnormal score of data record x level off to 1), data x is abnormal data at this time;Work as avgdepth
(x) when>c (M), score (x)<0.5 thinks that data x is normal data at this time;Threshold value u can be determined by experiment 0.5.
Wherein it should be noted that needing to calculate whether it meets Poisson distribution when input data arrives, calculate public
Formula is:
Wherein X is stochastic variable, only takes nonnegative integral value 0,1,2 ..., parameter It is that random time is averaged in the unit time
Incidence;E is constant, and normal value is e=2.718281828459;K=0,1,2 ....When Poisson distribution effect is description unit
Between random time occur number.
When the arrival data reach, it is based on Poisson distribution, judges whether the sample is added to as more new samples and sets
Set the buffer area of storage sample data.
3) judge whether to need to update detector;
According to principle of locality, for upcoming data, new data is more more valuable than legacy data, therefore, when setting is fixed
Between legacy data in buffer area is carried out it is superseded, so as to new data arrival when facilitate storage;
Abnormal rate 3-1) is calculated according to user's application scale predetermined, if having arrived fixed renewal time or Sample Buffer
Area has expired, then is updated to detector;
It 3-2) is calculated according to user's application scale predetermined, if the buffer area of storage sample has been expired, to detection
Device is updated;
Finally return that the anomaly detector of update, that is, isolated forest IForest '.
It is the detection process flow chart of transformer data exception in step 2 shown in Fig. 2, process includes the following steps:
2-1) first collect transformer historical data, wherein transformer historical data include voltage ratio, working frequency, inductance,
Degree of protection and fully loaded performance etc.;These different types of data are arranged, are indicated using numerical value;
Above-mentioned historical data is established into isolated forest according to IForest algorithm, as the model carried out abnormality detection;
The threshold value for test determining abnormal data score 2-2) is set as 0.5;When equipment operating data arrives, to wherein each
The data of seed type carry out depth calculation in isolated forest, seek its mean value, and calculate it according to abnormal scoring function and obtain extremely
Point, if it exceeds threshold value 0.5, then carry out early warning, it is abnormal to illustrate that current arrival data exist;If being not above threshold value 0.5,
Wait the arrival of equipment operating data next time.
The distribution situation that it is obeyed should also be calculated when operation data reaches, if obeying Poisson distribution, and
And buffer area it is discontented when, which should be added buffer area.
2-3) when buffer area has been expired, or time is up for fixed update abnormal detection model, then needs to abnormality detection
Model, that is, IForest are updated, to ensure the accuracy of abnormality detection.
The present invention by being improved to traditional IForest method, using improve abnormal data lookup accuracy as target,
A kind of industrial data denoising method based on improvement IForest is devised, it, should to carry out quickly and effectively anomaly data detection
Algorithm effect is good, and time efficiency is high, and higher-dimension mass data can be effectively treated.
Claims (5)
1. a kind of based on the industrial data denoising method for improving IForest, which comprises the steps of:
1) forest IForest is initially isolated in building;
1-1) building tree iTree;
All tree iTree 1-2) have been combined to form to initial isolated forest IForest;The isolated forest is as initial
Anomaly detector;
2) online abnormality detection;
To the data of each arrival, its corresponding different types of data is put into the isolated forest built up in step 1)
In, unusual condition is judged, if being higher than preset threshold value, explanation according to the abnormal score that input data mean depth obtains
The arrival data do not reach general level, are abnormal datas;
3) judge whether to need to update detector;
Calculated according to user's application scale predetermined, if storage sample buffer area expired, to detector into
Row updates.
2. according to claim 1 based on the industrial data denoising method for improving IForest, which is characterized in that isolate gloomy
Woods IForest is constituted by largely setting iTree;ITree is a kind of random binary tree, each node or there are two child,
It is exactly leaf node;All properties are all the variable of continuous type, the composition process of iTree in given history data set D, D
It is as follows:
An attribute Qi 1-1-1) is randomly choosed in given history data set D;
A value q of the attribute 1-1-2) is randomly choosed, the value is between maximum and minimum value;
1-1-3) the attribute Qi according to step 1-1-1) classifies to every record, attribute Qi value is less than the note of q
Recording playback is placed on right child in left child, the record attribute Qi value more than or equal to q;
1-1-4) the left child of recursive construction and right child only have a record or a plurality of one until meeting incoming data set
The record of sample or the height of tree have reached restriction height l;Modify the number of path N of isolated forest, initial value 0.
3. according to claim 1 based on the industrial data denoising method for improving IForest, which is characterized in that step 2
In, when the data of the arrival reach, it to be based on Poisson distribution, judges whether the sample is used as more new samples to be added to setting
The buffer area of sample data is stored well, and in the buffer area, sample data is arranged according to chronological order.
4. according to claim 1 based on the industrial data denoising method for improving IForest, which is characterized in that step 2
The detection process of middle transformer data exception, includes the following steps:
2-1) first collect transformer historical data, wherein transformer historical data include voltage ratio, working frequency, inductance,
Degree of protection and fully loaded performance;These different types of data are arranged, are indicated using numerical value;
Above-mentioned historical data is established into isolated forest according to IForest algorithm, as the model carried out abnormality detection;
The threshold value for test determining abnormal data score 2-2) is set as 0.5;When equipment operating data arrives, to wherein each
The data of seed type carry out depth calculation in isolated forest, seek its mean value, and calculate it according to abnormal scoring function and obtain extremely
Point, if it exceeds threshold value 0.5, then carry out early warning, it is abnormal to illustrate that current arrival data exist;If being not above threshold value 0.5,
Wait the arrival of equipment operating data next time;
The distribution situation that it is obeyed should also be calculated when operation data reaches, if obeying Poisson distribution, Er Qiehuan
Rush area it is discontented when, which should be added buffer area;
2-3) when buffer area has been expired, or time is up for fixed update abnormal detection model, then needs to abnormality detection model,
Namely IForest is updated, to ensure the accuracy of abnormality detection.
5. according to claim 1 based on the industrial data denoising method for improving IForest, which is characterized in that step (3)
Judge whether to need to update detector;According to principle of locality, for upcoming data, new data is more valuable than legacy data
Value, therefore, setting the set time legacy data in buffer area is carried out it is superseded, so as to new data arrival when facilitate storage;It is specific
Steps are as follows:
Abnormal rate 3-1) is calculated according to user's application scale predetermined, if having arrived fixed renewal time or Sample Buffer
Area has expired, then is updated to detector;
It 3-2) is calculated according to user's application scale predetermined, if the buffer area of storage sample has been expired, to detection
Device is updated;
Finally return that the anomaly detector of update, that is, isolated forest IForest '.
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2018109655990 | 2018-08-23 | ||
CN201810965599 | 2018-08-23 |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109543765A true CN109543765A (en) | 2019-03-29 |
Family
ID=65850765
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811439128.2A Pending CN109543765A (en) | 2018-08-23 | 2018-11-29 | A kind of industrial data denoising method based on improvement IForest |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109543765A (en) |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109948738A (en) * | 2019-04-11 | 2019-06-28 | 合肥工业大学 | Energy consumption method for detecting abnormality, the apparatus and system of coating drying room |
CN110119340A (en) * | 2019-05-17 | 2019-08-13 | 北京字节跳动网络技术有限公司 | Method for monitoring abnormality, device, electronic equipment and storage medium |
CN110149258A (en) * | 2019-04-12 | 2019-08-20 | 北京航空航天大学 | A kind of automobile CAN-bus network data method for detecting abnormality based on isolated forest |
CN110243599A (en) * | 2019-07-02 | 2019-09-17 | 西南交通大学 | Multidimensional peels off train EMU axle box bearing temperature anomaly state monitoring method |
CN110888850A (en) * | 2019-12-04 | 2020-03-17 | 国网山东省电力公司威海供电公司 | Data quality detection method based on power Internet of things platform |
CN111428886A (en) * | 2020-04-10 | 2020-07-17 | 青岛聚好联科技有限公司 | Fault diagnosis deep learning model self-adaptive updating method and device |
CN111740856A (en) * | 2020-05-07 | 2020-10-02 | 北京直真科技股份有限公司 | Network communication equipment alarm acquisition abnormity early warning method based on abnormity detection algorithm |
CN112990246A (en) * | 2019-12-17 | 2021-06-18 | 杭州海康威视数字技术股份有限公司 | Method and device for establishing isolated tree model |
CN113112188A (en) * | 2021-05-14 | 2021-07-13 | 北京邮电大学 | Power dispatching monitoring data anomaly detection method based on pre-screening dynamic integration |
CN113721000A (en) * | 2021-07-16 | 2021-11-30 | 国家电网有限公司大数据中心 | Method and system for detecting abnormity of dissolved gas in transformer oil |
CN113762507A (en) * | 2021-08-24 | 2021-12-07 | 浙江中辰城市应急服务管理有限公司 | Semi-supervised deep learning arc voltage anomaly detection method based on phase space reconstruction |
CN116842322A (en) * | 2023-07-19 | 2023-10-03 | 深圳市精微康投资发展有限公司 | Electric motor operation optimization method and system based on data processing |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106960358A (en) * | 2017-01-13 | 2017-07-18 | 重庆小富农康农业科技服务有限公司 | A kind of financial fraud behavior based on rural area electronic commerce big data deep learning quantifies detecting system |
CN107657288A (en) * | 2017-10-26 | 2018-02-02 | 国网冀北电力有限公司 | A kind of power scheduling flow data method for detecting abnormality based on isolated forest algorithm |
-
2018
- 2018-11-29 CN CN201811439128.2A patent/CN109543765A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106960358A (en) * | 2017-01-13 | 2017-07-18 | 重庆小富农康农业科技服务有限公司 | A kind of financial fraud behavior based on rural area electronic commerce big data deep learning quantifies detecting system |
CN107657288A (en) * | 2017-10-26 | 2018-02-02 | 国网冀北电力有限公司 | A kind of power scheduling flow data method for detecting abnormality based on isolated forest algorithm |
Non-Patent Citations (3)
Title |
---|
丁智国,莫毓昌,杨凡: ""一种新的在线流数据异常检测方法"", 《计算机科学》 * |
单林森: ""配电变压器故障分析及随机森林诊断实现"", 《科技风》 * |
李星南,施展,亢中苗等: ""基于孤立森林算法和BP神经网络算法的电力运维数据清洗方法"", 《电气应用》 * |
Cited By (20)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109948738A (en) * | 2019-04-11 | 2019-06-28 | 合肥工业大学 | Energy consumption method for detecting abnormality, the apparatus and system of coating drying room |
CN110149258A (en) * | 2019-04-12 | 2019-08-20 | 北京航空航天大学 | A kind of automobile CAN-bus network data method for detecting abnormality based on isolated forest |
CN110119340A (en) * | 2019-05-17 | 2019-08-13 | 北京字节跳动网络技术有限公司 | Method for monitoring abnormality, device, electronic equipment and storage medium |
CN110243599A (en) * | 2019-07-02 | 2019-09-17 | 西南交通大学 | Multidimensional peels off train EMU axle box bearing temperature anomaly state monitoring method |
CN110243599B (en) * | 2019-07-02 | 2020-05-05 | 西南交通大学 | Method for monitoring temperature abnormal state of multi-dimensional outlier train motor train unit axle box bearing |
CN110888850A (en) * | 2019-12-04 | 2020-03-17 | 国网山东省电力公司威海供电公司 | Data quality detection method based on power Internet of things platform |
CN112990246B (en) * | 2019-12-17 | 2022-09-09 | 杭州海康威视数字技术股份有限公司 | Method and device for establishing isolated tree model |
CN112990246A (en) * | 2019-12-17 | 2021-06-18 | 杭州海康威视数字技术股份有限公司 | Method and device for establishing isolated tree model |
CN111428886A (en) * | 2020-04-10 | 2020-07-17 | 青岛聚好联科技有限公司 | Fault diagnosis deep learning model self-adaptive updating method and device |
CN111428886B (en) * | 2020-04-10 | 2023-08-04 | 青岛聚好联科技有限公司 | Method and device for adaptively updating deep learning model of fault diagnosis |
CN111740856B (en) * | 2020-05-07 | 2023-04-28 | 北京直真科技股份有限公司 | Network communication equipment alarm acquisition abnormity early warning method based on abnormity detection algorithm |
CN111740856A (en) * | 2020-05-07 | 2020-10-02 | 北京直真科技股份有限公司 | Network communication equipment alarm acquisition abnormity early warning method based on abnormity detection algorithm |
CN113112188B (en) * | 2021-05-14 | 2022-05-17 | 北京邮电大学 | Power dispatching monitoring data anomaly detection method based on pre-screening dynamic integration |
CN113112188A (en) * | 2021-05-14 | 2021-07-13 | 北京邮电大学 | Power dispatching monitoring data anomaly detection method based on pre-screening dynamic integration |
CN113721000A (en) * | 2021-07-16 | 2021-11-30 | 国家电网有限公司大数据中心 | Method and system for detecting abnormity of dissolved gas in transformer oil |
CN113721000B (en) * | 2021-07-16 | 2023-02-03 | 国家电网有限公司大数据中心 | Method and system for detecting abnormity of dissolved gas in transformer oil |
CN113762507A (en) * | 2021-08-24 | 2021-12-07 | 浙江中辰城市应急服务管理有限公司 | Semi-supervised deep learning arc voltage anomaly detection method based on phase space reconstruction |
CN113762507B (en) * | 2021-08-24 | 2023-12-29 | 浙江中辰城市应急服务管理有限公司 | Semi-supervised deep learning arc voltage anomaly detection method based on phase space reconstruction |
CN116842322A (en) * | 2023-07-19 | 2023-10-03 | 深圳市精微康投资发展有限公司 | Electric motor operation optimization method and system based on data processing |
CN116842322B (en) * | 2023-07-19 | 2024-02-23 | 深圳市精微康投资发展有限公司 | Electric motor operation optimization method and system based on data processing |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109543765A (en) | A kind of industrial data denoising method based on improvement IForest | |
CN104281674B (en) | It is a kind of based on the adaptive clustering scheme and system that gather coefficient | |
CN105471637B (en) | A kind of complex network node importance appraisal procedure and system | |
CN112640380A (en) | Apparatus and method for anomaly detection of an input stream of events | |
US11841839B1 (en) | Preprocessing and imputing method for structural data | |
CN110110322A (en) | Network new word discovery method, apparatus, electronic equipment and storage medium | |
CN103761249B (en) | Data lead-in method and system based on Data Matching | |
CN106294815B (en) | A kind of clustering method and device of URL | |
CN106909669A (en) | The detection method and device of a kind of promotion message | |
CN103761286B (en) | A kind of Service Source search method based on user interest | |
CN112149737A (en) | Selection model training method, model selection method, selection model training device and selection model selection device, and electronic equipment | |
CN102609501A (en) | Data cleaning method based on real-time historical database | |
CN103150470A (en) | Visualization method for concept drift of data stream in dynamic data environment | |
US20240177077A1 (en) | Attribution analysis method, electronic device, and storage medium | |
CN109670153A (en) | A kind of determination method, apparatus, storage medium and the terminal of similar model | |
CN104778202B (en) | The analysis method and system of event evolutionary process based on keyword | |
CN105930358B (en) | Case retrieval method and system based on relevance | |
CN107832450A (en) | Method for cleaning Data duplication record | |
CN105512270A (en) | Method and device for determining related objects | |
CN108874974A (en) | Parallelization Topic Tracking method based on frequent term set | |
da F. Vieira et al. | Modularity based hierarchical community detection in networks | |
CN109739840A (en) | Data processing empty value method, apparatus and terminal device | |
Li et al. | Dynamic multiobjective optimization algorithm based on average distance linear prediction model | |
CN112507181B (en) | Search request classification method, device, electronic equipment and storage medium | |
CN103955526A (en) | Data storage method and device |
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: 20190329 |
|
RJ01 | Rejection of invention patent application after publication |