CN102693452A - Multiple-model soft-measuring method based on semi-supervised regression learning - Google Patents
Multiple-model soft-measuring method based on semi-supervised regression learning Download PDFInfo
- Publication number
- CN102693452A CN102693452A CN2012101485886A CN201210148588A CN102693452A CN 102693452 A CN102693452 A CN 102693452A CN 2012101485886 A CN2012101485886 A CN 2012101485886A CN 201210148588 A CN201210148588 A CN 201210148588A CN 102693452 A CN102693452 A CN 102693452A
- Authority
- CN
- China
- Prior art keywords
- data
- model
- classification
- soft
- submodel
- 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
Images
Abstract
The invention provides a multiple-model soft-measuring method based on semi-supervised regression learning, comprising steps: step 1, clustering data; step 2, adopting a plurality of submodel prediction projects to predict a submodel together according to data classification results and obtaining submodel prediction results; step 3, merging a plurality of submodel prediction results and obtaining prediction results. The invention can reduce effectively model inaccurate problems caused by process complication. A weighting multiple-model soft-measuring modeling method does not ask for classifying beforehand input data. Data is classified in real time to obtain prediction results online as long as submodel adopts the online modeling method. Because the submodel of a switching multiple-model soft-measuring modeling method only predicts part of similar input data, model prediction accuracy of the method is higher relatively and data is predicted distinguishingly commendably based on operating conditions.
Description
Technical field
What the present invention relates to is a kind of method of soft measuring instrument technical field, specifically is a kind of multi-model flexible measurement method based on semi-supervised recurrence learning.
Background technology
In present commercial production, generally adopt two kinds of means to realize the measurement of control variable: the method that the method for in-line meter and off-line measurement are analyzed.But these two kinds of methods all exist certain shortcoming and inferior position: the method that in-line meter is measured often input cost is high; Maintenance difficulties is big; The method that off-line measurement is analyzed is then because the adding of artifact; Measure human error easily and become big, and it is serious to measure hysteresis quality, be difficult to satisfy the system that real-time is had relatively high expectations.The development of Along with computer technology; Soft-measuring technique becomes the effective ways that address this problem; Soft-measuring technique is that a kind of modeling technique that uses a computer is difficult to measure to some or the variable that can't measure carries out forecast method; This variable that is difficult to measure is commonly referred to as leading variable, and is relative with it, then is commonly referred to as auxiliary variable as the relatively easy variable of measuring of the other of model input.The input of soft-measuring technique is an auxiliary variable, and output is leading variable, and this method generally has the advantage that input cost is low, easy to maintenance, can measure in real time.
In traditional soft-measuring technique; Often only use a spot of flag data to carry out modeling; Wasted the value of the unmarked data that can obtain easily in a large number, and the study machine of semi-supervised learning how be a kind of research to utilize unmarked sample the obtains better performance and the ability of popularization.In the last few years, along with the continuous development of machine learning and the increase day by day of unmarked data, semi-supervised learning became a research focus just gradually.In general, obtaining of flag data is comparison " costliness ", in a system; Generally have no idea to obtain in large quantities flag data and set up model, and unmarked data are not because need the artificially that data are carried out mark, so can it be obtained in a large number; And discover; In unmarked data, there is the information that to explain the data structure characteristics,, in soft-measuring technique, can improves precision of prediction effectively if ability is used to these information.
Along with the fast development of modern society's economy with science and technology, industrial production environment becomes and becomes increasingly complex, and wants in complicated like this industrial environment, to accomplish system's precise control, must more accurate and diversified understanding be arranged to system.In soft field of measurement, because the leading variable of measuring generally is to dominate the principal element that control strategy changes, so, will bring very serious control problem if soft-sensing model is inaccurate.The method of soft sensor modeling has a variety of at present; Every kind of method all has its field own corresponding or that be good at; Characteristics such as multivariate, non-linear, strong coupling, multi-state, control performance composite request height also often appear in the control system of now; Be the combination or the coupling of polytype system, adopt single model to carry out soft sensor modeling to this system, effect is also bad.So; For the system of complicated multi-state, be necessary to set up a plurality of soft-sensing models, utilize a plurality of models that the leading variable of system is measured simultaneously; The many groups of basis measurement results are optimized again at last, thereby help to improve robustness, accuracy and the reliability of soft-sensing model.
Summary of the invention
The present invention is directed to and do not utilize unmarked data in the conventional softer measuring technique and, a kind of multi-model flexible measurement method based on semi-supervised recurrence learning is provided for multi-state, the bad problem of the single model prediction effect of non-linear strong complex industrial process.
According to an aspect of the present invention, a kind of multi-model flexible measurement method based on semi-supervised recurrence learning is provided, comprises the steps:
Step 1: data are carried out cluster;
Step 2: adopt a plurality of submodel prediction scheme to carry out the submodel prediction jointly according to the data qualification result, obtain submodel and predict the outcome;
Step 3: a plurality of submodels are predicted the outcome to be merged, and is predicted the outcome.
Preferably, said step 1 is specially: adopt G-K fuzzy clustering algorithm that data are carried out cluster with Weighting type multi-model soft-measuring modeling method.
Preferably, said step 3 comprises step: submodel is predicted the outcome carry out the optimization based on weighting.
Preferably, said step 1 is specially: adopt the minimum distance classification algorithm that data are carried out cluster with suitching type multi-model soft-measuring modeling method.
Preferably, said step 3 comprises step: submodel is predicted the outcome carry out the optimization based on switching.
Preferably; Before the soft-sensing model modeling; At first will classify as the data based similarity between them of the auxiliary variable of model input through data clusters; Then adopt different modeling strategy that a system is split as a plurality of soft measurement submodels according to classification results again and carry out the variable prediction jointly, merge predicting the outcome of a plurality of submodels more at last, the soft-sensing model that forms a kind of multi-model predicts the outcome.
Preferably, saidly data are carried out cluster comprise fuzzy clustering and hard cluster, wherein; Fuzzy clustering is meant a kind of undemanding cluster, and promptly some data points also not exclusively belong to a classification, and data point shows through fuzzy membership the said relation of each classification; Fuzzy membership is the decimal of a scope in 0 to 1, and this decimal is big more, and representative data point is similar more with the data in a certain classification; In addition, same data point adds the fuzzy membership of all categories and is 1; Hard cluster is meant a kind of cluster of strictness, and promptly some data points are strictly to belong to some classifications, and there is not the notion of degree of membership in this method, concerning the affiliated relation of a certain data point, is either-or.
Preferably; Weighting type multi-model soft-measuring modeling method refers to: at first utilize G-K fuzzy clustering algorithm that the input data point is carried out cluster analysis; Obtain the fuzzy membership of cluster centre and each data point; Then set up the submodel that equates with the classification number and carry out the prediction of soft-sensing model leading variable according to the characteristics of classification; Last to be weighting coefficient to all submodels of this data point predict the outcome carries out weighted calculation with the fuzzy membership of each data point again, finally obtains a multi-model soft-sensing model based on weighting and predict the outcome.
Preferably, described suitching type multi-model soft-measuring modeling method refers to: at first to carrying out cluster analysis as the auxiliary variable data of input, obtain the cluster numbers and the cluster centre of data; Then adopt the minimum distance classification method that the input data are classified; The minimum Cluster Classification here refers to through calculating a certain data point to judge to each distances of clustering centers which classification is this data point belong to, and gets the affiliated classification of the minimum classification of distance for this data point; And then data are wherein predicted according to the soft measurement submodel that classification results is set up each classification; Get the union that predicts the outcome of data in all categories at last, the direct collection that predicts the outcome as this multi-model flexible measurement method.
More particularly, two kinds of multi-model flexible measurement methods based on semi-supervised recurrence learning are provided: Weighting type multi-model soft-measuring modeling method and suitching type multi-model soft-measuring modeling method.At first; These two kinds of multi-model soft-measuring modeling methods all are the basis with the data clusters algorithm; Different is: Weighting type multi-model soft-measuring modeling method adopts G-K fuzzy clustering algorithm that data are carried out cluster, and suitching type multi-model soft-measuring modeling method adopts minimum distance classification that data are carried out cluster; Then adopt different submodel prediction scheme to carry out the submodel prediction respectively according to the data qualification result again; Last respectively submodel being predicted the outcome according to the difference of two kinds of modeling methods again carried out based on weighting or based on the optimization of switching, thereby obtains predicting the outcome under the different multi-model soft-measuring modeling methods.
Described learning method based on semi-supervised recurrence refers to: through semi-supervised kernel being introduced a kind of method of supervising in the recurrence learning algorithm target equation; It is become a kind of learning algorithm based on semi-supervised recurrence; Then carrying out minimum value through the objective function to this semi-supervised learning method again finds the solution; Thereby obtain the value of some known variables when objective function is obtained minimum value, come a kind of learning method of descriptive system funtcional relationship at last again through the value of these known variables.
The described multi-model soft-sensing model that is the basis with the data clusters algorithm refers to: before the soft-sensing model modeling; At first will classify as the data based similarity between them of the auxiliary variable of model input through data clusters; Then carry out the variable prediction jointly adopting different modeling strategy that a system is split as a plurality of soft measurement submodels according to classification results; Merge predicting the outcome of a plurality of submodels more at last, the soft-sensing model that forms a kind of multi-model predicts the outcome.The data clusters algorithm here can be divided into fuzzy clustering and hard cluster, and so-called fuzzy clustering is meant a kind of undemanding cluster, and promptly some data points also not exclusively belong to a classification; Data point shows through fuzzy membership the said relation of each classification; Fuzzy membership is the decimal of a scope in 0 to 1, and this decimal is big more, and representative data point is similar more with the data in a certain classification; In addition, same data point adds the fuzzy membership of all categories and is 1; So-called hard cluster is meant a kind of cluster of strictness, and promptly some data points are strictly to belong to some classifications, and there is not the notion of degree of membership in this method, concerning the affiliated relation of a certain data point, is either-or.
Described Weighting type multi-model soft-measuring modeling method refers to: at first utilize G-K fuzzy clustering algorithm that the input data point is carried out cluster analysis; Obtain the fuzzy membership of cluster centre and each data point; Then set up the submodel that equates with the classification number and carry out the prediction of soft-sensing model leading variable according to the characteristics of classification; Last to be weighting coefficient to all submodels of this data point predict the outcome carries out weighted calculation with the fuzzy membership of each data point again, finally obtains a multi-model soft-sensing model based on weighting and predict the outcome.As shown in Figure 1, its step is following:
1. analyze a part of data that obtain in advance, find out sub-operating mode (classification) number c, wherein c can be rule of thumb given or analyzes through other clustering algorithms.
2. be respectively each classification chooser model modelling approach, and parameters optimization, make this model optimum to the data prediction effect of affiliated classification.
3. read input data x
j, use all submodels to x
jPredict, obtain submodel and predict the outcome
4. prediction finishes, and changes 5.; Prediction does not finish j=j+1, changes 3..
5. input data set X is carried out the G-K cluster analysis, X is divided into c classification, obtain the degree of membership matrix U, export according to formula computes multi-model
U wherein
iBe the degree of membership vector of i classification to all data among the X,
Be all data predicting the outcome under i submodel among the X.
6. the multi-model of exporting all data predicts the outcome
Described suitching type multi-model soft-measuring modeling method refers to: at first to carrying out cluster analysis as the auxiliary variable data of input, obtain the cluster numbers and the cluster centre of data; Then adopt the minimum distance classification method that the input data are classified; The minimum Cluster Classification here refers to through calculating a certain data point to judge to each distances of clustering centers which classification is this data point belong to, and gets the affiliated classification of the minimum classification of distance for this data point; And then data are wherein predicted according to the soft measurement submodel that classification results is set up each classification; Get the union that predicts the outcome of data in all categories at last, the direct collection that predicts the outcome as this multi-model flexible measurement method.
As shown in Figure 2, the step of this method is following:
1. obtain whole input data and carry out cluster analysis, all data are divided into c classification, and obtain the cluster centre V={v of c classification
i, i=1,2 ..., c.
2. obtain an input data point x
j, utilize its Euclidean distance of computes to each cluster centre.
d
i=||x
j-v
i||
3. obtain the minimum classification p of d, and x
jClassify as this classification, it is joined X
p
4. all data read end, change 5.; All data do not read end, change 2..
Whether the marker samples number of 5. checking each classification is less than n, if change 6., otherwise change 7..
6. the classification that the marker samples number is less than n, half the marker samples and the marker samples of himself of each extraction gathered also from an other c-1 classification, jointly as such other marker samples collection.
7. respectively the data set of c classification is chosen submodel modeling method and parameters optimization, predict
8. merge all submodels and predict the outcome, obtain multi-model and predict the outcome
Compared with prior art, the present invention has following beneficial effect:
1. in the conventional softer measuring method, generally only input is carried out the foundation of single model, two kinds of multi-model soft-measuring modeling methods based on semi-supervised recurrence provided by the invention can effectively reduce because the inaccurate problem of the complicated model that brings of process.
2. Weighting type multi-model soft-measuring modeling method is a kind of method based on fuzzy clustering; This method does not require in advance just classifies to the input data; So, just can carry out real-time grading and predicted the outcome online to data as long as submodel takes the line modeling method.
3. suitching type multi-model soft-measuring modeling method submodel is because only only predict the similar input data of a part, so its model prediction accuracy than higher, can be distinguished prediction according to the operating mode situation to data well.
The present invention has a wide range of applications in industrial soft-measuring modeling method field.
Description of drawings
Fig. 1 is a Weighting type multi-model soft-measuring modeling method process flow diagram.
Fig. 2 is a suitching type multi-model soft-measuring modeling method process flow diagram.
Fig. 3 is the prediction effect figure of multi-model soft-measuring modeling method submodel A to flue gas oxygen content.
Fig. 4 is the prediction effect figure of multi-model soft-measuring modeling method multi-model to flue gas oxygen content.
Embodiment
The learning method based on semi-supervised recurrence described in the present invention refers to: through semi-supervised kernel being introduced a kind of method of supervising in the recurrence learning algorithm target equation; It is become a kind of learning algorithm based on semi-supervised recurrence; Then carrying out minimum value through the objective function to this semi-supervised learning method again finds the solution; Thereby obtain the value of some known variables when objective function is obtained minimum value, come a kind of learning method of descriptive system funtcional relationship at last again through the value of these known variables.
Elaborate in the face of embodiments of the invention down, present embodiment provided detailed embodiment and concrete operating process, but protection scope of the present invention is not limited to following embodiment being to implement under the prerequisite with technical scheme of the present invention.
Embodiment
Present embodiment utilizes certain 1000MW of power plant measured data to carry out the flue gas oxygen content prediction based on Weighting type multi-model soft-measuring modeling method, may further comprise the steps: at first, understand research object, choose auxiliary variable and leading variable; Secondly, choose submodel and choose the optimized parameter that is adapted to these submodel data according to data characteristics; Once more, the input data are carried out the prediction of a plurality of submodels, obtain predicting the outcome of many groups; At last, utilize G-K fuzzy clustering algorithm that the input data are carried out cluster and obtain the fuzzy membership matrix, and utilize fuzzy membership to carry out weighted calculation for weighting coefficient predicts the outcome to many groups submodel.
When estimating the performance of modeling method, used following index: relative root-mean-square error (RMSE), its computing formula is following:
RMSE mainly is used for estimating the precision of soft-sensing model, and RMSE is more little, and the expression precision is high more.
Present embodiment submodel number is three; Three submodels have been chosen based on the soft-measuring modeling method of semi-supervised local linear regression and the soft-measuring modeling method that returns based on semi-supervised Gaussian process, and have chosen the optimization model parameter to data characteristic in three kinds respectively.
Be comparative example result, except multi-model predicts the outcome, also three sub-predicted results carried out RMSE calculating, the result is as shown in table 1.
| Model | Submodel A | Submodel B | Submodel C | Multi-model |
| RMSE | 1.4572 | 4.1107 | 3.0437 | 0.2849 |
Table 1
Can find out that from table 1 after predicting the outcome of three groups of submodels carried out weighted calculation, the precision that predicts the outcome of multi-model improved very obvious; This is because three kinds of submodels all can not all have a good prediction to whole data, only predicting preferably the data in its classification, gives an example; From the predicting the outcome of Fig. 3 submodel A, can find out; Submodel A is at the non-constant of a certain data segment prediction effect, this is because at this moment the parameter of submodel A can not adapt to this one piece of data, but is optimized separately for other segment datas; But on the multi-model of Fig. 4 predicts the outcome; Then pass through weighted calculation, can be effectively submodel A prediction effect preferably data segment inherit, and rejected predicting the outcome of submodel A on the bad data segment of effect.
Claims (9)
1. the multi-model flexible measurement method based on semi-supervised recurrence learning is characterized in that, comprises the steps:
Step 1: data are carried out cluster;
Step 2: adopt a plurality of submodel prediction scheme to carry out the submodel prediction jointly according to the data qualification result, obtain submodel and predict the outcome;
Step 3: a plurality of submodels are predicted the outcome to be merged, and is predicted the outcome.
2. the multi-model flexible measurement method based on semi-supervised recurrence learning according to claim 1 is characterized in that said step 1 is specially: adopt G-K fuzzy clustering algorithm that data are carried out cluster with Weighting type multi-model soft-measuring modeling method.
3. the multi-model flexible measurement method based on semi-supervised recurrence learning according to claim 2 is characterized in that said step 3 comprises step: submodel is predicted the outcome carry out the optimization based on weighting.
4. according to each described multi-model flexible measurement method in the claim 1 to 3 based on semi-supervised recurrence learning; It is characterized in that; Said step 1 is specially: adopt the minimum distance classification algorithm that data are carried out cluster with suitching type multi-model soft-measuring modeling method.
5. the multi-model flexible measurement method based on semi-supervised recurrence learning according to claim 4 is characterized in that said step 3 comprises step: submodel is predicted the outcome carry out the optimization based on switching.
6. according to each described multi-model flexible measurement method in the claim 1 to 5 based on semi-supervised recurrence learning; It is characterized in that; Be specially: before the soft-sensing model modeling; At first will classify as the data based similarity between them of the auxiliary variable of model input through data clusters; Then adopt different modeling strategy that a system is split as a plurality of soft measurement submodels according to classification results again and carry out the variable prediction jointly, merge predicting the outcome of a plurality of submodels more at last, the soft-sensing model that forms a kind of multi-model predicts the outcome.
7. according to each described multi-model flexible measurement method in the claim 1 to 6, it is characterized in that, saidly data are carried out cluster comprise fuzzy clustering and hard cluster based on semi-supervised recurrence learning; Wherein, fuzzy clustering is meant a kind of undemanding cluster, and promptly some data points also not exclusively belong to a classification; Data point shows through fuzzy membership the said relation of each classification; Fuzzy membership is the decimal of a scope in 0 to 1, and this decimal is big more, and representative data point is similar more with the data in a certain classification; In addition, same data point adds the fuzzy membership of all categories and is 1; Hard cluster is meant a kind of cluster of strictness, and promptly some data points are strictly to belong to some classifications, and there is not the notion of degree of membership in this method, concerning the affiliated relation of a certain data point, is either-or.
8. according to each described multi-model flexible measurement method in the claim 2 to 7 based on semi-supervised recurrence learning; It is characterized in that; Weighting type multi-model soft-measuring modeling method refers to: at first utilize G-K fuzzy clustering algorithm that the input data point is carried out cluster analysis; Obtain the fuzzy membership of cluster centre and each data point; Then set up the submodel that equates with the classification number and carry out the prediction of soft-sensing model leading variable according to the characteristics of classification; Last to be weighting coefficient to all submodels of this data point predict the outcome carries out weighted calculation with the fuzzy membership of each data point again, finally obtains a multi-model soft-sensing model based on weighting and predict the outcome.
9. according to each described multi-model flexible measurement method in the claim 4 to 8 based on semi-supervised recurrence learning; It is characterized in that; Described suitching type multi-model soft-measuring modeling method refers to: at first to carrying out cluster analysis as the auxiliary variable data of input, obtain the cluster numbers and the cluster centre of data; Then adopt the minimum distance classification method that the input data are classified; The minimum Cluster Classification here refers to through calculating a certain data point to judge to each distances of clustering centers which classification is this data point belong to, and gets the affiliated classification of the minimum classification of distance for this data point; And then data are wherein predicted according to the soft measurement submodel that classification results is set up each classification; Get the union that predicts the outcome of data in all categories at last, the direct collection that predicts the outcome as this multi-model flexible measurement method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012101485886A CN102693452A (en) | 2012-05-11 | 2012-05-11 | Multiple-model soft-measuring method based on semi-supervised regression learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2012101485886A CN102693452A (en) | 2012-05-11 | 2012-05-11 | Multiple-model soft-measuring method based on semi-supervised regression learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102693452A true CN102693452A (en) | 2012-09-26 |
Family
ID=46858865
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2012101485886A Pending CN102693452A (en) | 2012-05-11 | 2012-05-11 | Multiple-model soft-measuring method based on semi-supervised regression learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102693452A (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104778298A (en) * | 2015-01-26 | 2015-07-15 | 江南大学 | Gaussian process regression soft measurement modeling method based on EGMM (Error Gaussian Mixture Model) |
| CN106339357A (en) * | 2016-08-12 | 2017-01-18 | 北京同余科技有限公司 | Optimizing parameterized network information security detection system and method |
| TWI574169B (en) * | 2014-12-08 | 2017-03-11 | 英業達股份有限公司 | Method of operating a solution searching system and solution searching system |
| CN107451102A (en) * | 2017-07-28 | 2017-12-08 | 江南大学 | A kind of semi-supervised Gaussian process for improving self-training algorithm returns soft-measuring modeling method |
| CN108307057A (en) * | 2018-01-22 | 2018-07-20 | 广东欧珀移动通信有限公司 | Fall the analysis method and Related product of data |
| CN108596415A (en) * | 2017-12-15 | 2018-09-28 | 阿里巴巴集团控股有限公司 | A kind of model integration method and device |
| CN108734207A (en) * | 2018-05-14 | 2018-11-02 | 江南大学 | A kind of model prediction method based on double preferred Semi-Supervised Regression algorithms |
| CN108764295A (en) * | 2018-04-28 | 2018-11-06 | 江南大学 | A kind of soft-measuring modeling method based on semi-supervised integrated study |
| CN109299279A (en) * | 2018-11-29 | 2019-02-01 | 北京奇安信科技有限公司 | A kind of data processing method, equipment, system and medium |
| CN109344561A (en) * | 2018-12-11 | 2019-02-15 | 上海长庚信息技术股份有限公司 | One kind realizing flexible measurement method based on similarity principle |
| WO2019213860A1 (en) * | 2018-05-09 | 2019-11-14 | Jiangnan University | Advanced ensemble learning strategy based semi-supervised soft sensing method |
| WO2020155300A1 (en) * | 2019-01-31 | 2020-08-06 | 网宿科技股份有限公司 | Model prediction method and device |
| CN112650063A (en) * | 2020-12-30 | 2021-04-13 | 浙江大学 | Self-adaptive soft measurement method based on semi-supervised incremental Gaussian mixture regression |
| CN112696667A (en) * | 2020-12-31 | 2021-04-23 | 华电国际电力股份有限公司天津开发区分公司 | Bed temperature early warning system of circulating fluidized bed boiler unit |
| CN112797402A (en) * | 2020-12-31 | 2021-05-14 | 华电国际电力股份有限公司天津开发区分公司 | Underpants leg type circulating fluidized bed boiler unit and bed pressure prediction system |
| CN113177358A (en) * | 2021-04-30 | 2021-07-27 | 燕山大学 | Soft measurement method for cement quality based on fuzzy fine-grained feature extraction |
-
2012
- 2012-05-11 CN CN2012101485886A patent/CN102693452A/en active Pending
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI574169B (en) * | 2014-12-08 | 2017-03-11 | 英業達股份有限公司 | Method of operating a solution searching system and solution searching system |
| CN104778298A (en) * | 2015-01-26 | 2015-07-15 | 江南大学 | Gaussian process regression soft measurement modeling method based on EGMM (Error Gaussian Mixture Model) |
| CN104778298B (en) * | 2015-01-26 | 2017-09-19 | 江南大学 | Gaussian process based on EGMM returns soft-measuring modeling method |
| CN106339357A (en) * | 2016-08-12 | 2017-01-18 | 北京同余科技有限公司 | Optimizing parameterized network information security detection system and method |
| CN107451102A (en) * | 2017-07-28 | 2017-12-08 | 江南大学 | A kind of semi-supervised Gaussian process for improving self-training algorithm returns soft-measuring modeling method |
| CN107451102B (en) * | 2017-07-28 | 2020-06-30 | 江南大学 | Method for predicting concentration of butane at bottom of debutanizer tower based on improved self-training algorithm semi-supervised Gaussian process regression soft measurement modeling |
| CN108596415A (en) * | 2017-12-15 | 2018-09-28 | 阿里巴巴集团控股有限公司 | A kind of model integration method and device |
| CN108596415B (en) * | 2017-12-15 | 2023-11-24 | 创新先进技术有限公司 | Model integration method and device |
| US11132624B2 (en) | 2017-12-15 | 2021-09-28 | Advanced New Technologies Co., Ltd. | Model integration method and device |
| CN108307057A (en) * | 2018-01-22 | 2018-07-20 | 广东欧珀移动通信有限公司 | Fall the analysis method and Related product of data |
| CN108764295B (en) * | 2018-04-28 | 2021-05-28 | 江南大学 | Method for predicting concentration of butane at bottom of debutanizer tower based on soft measurement modeling of semi-supervised ensemble learning |
| CN108764295A (en) * | 2018-04-28 | 2018-11-06 | 江南大学 | A kind of soft-measuring modeling method based on semi-supervised integrated study |
| US11488071B2 (en) | 2018-05-09 | 2022-11-01 | Jiangnan University | Advanced ensemble learning strategy based semi-supervised soft sensing method |
| WO2019213860A1 (en) * | 2018-05-09 | 2019-11-14 | Jiangnan University | Advanced ensemble learning strategy based semi-supervised soft sensing method |
| CN108734207B (en) * | 2018-05-14 | 2021-05-28 | 江南大学 | Method for predicting concentration of butane at bottom of debutanizer tower based on model of double-optimization semi-supervised regression algorithm |
| CN108734207A (en) * | 2018-05-14 | 2018-11-02 | 江南大学 | A kind of model prediction method based on double preferred Semi-Supervised Regression algorithms |
| CN109299279B (en) * | 2018-11-29 | 2020-08-21 | 奇安信科技集团股份有限公司 | Data processing method, device, system and medium |
| CN109299279A (en) * | 2018-11-29 | 2019-02-01 | 北京奇安信科技有限公司 | A kind of data processing method, equipment, system and medium |
| CN109344561A (en) * | 2018-12-11 | 2019-02-15 | 上海长庚信息技术股份有限公司 | One kind realizing flexible measurement method based on similarity principle |
| WO2020155300A1 (en) * | 2019-01-31 | 2020-08-06 | 网宿科技股份有限公司 | Model prediction method and device |
| CN112650063A (en) * | 2020-12-30 | 2021-04-13 | 浙江大学 | Self-adaptive soft measurement method based on semi-supervised incremental Gaussian mixture regression |
| CN112696667A (en) * | 2020-12-31 | 2021-04-23 | 华电国际电力股份有限公司天津开发区分公司 | Bed temperature early warning system of circulating fluidized bed boiler unit |
| CN112797402A (en) * | 2020-12-31 | 2021-05-14 | 华电国际电力股份有限公司天津开发区分公司 | Underpants leg type circulating fluidized bed boiler unit and bed pressure prediction system |
| CN113177358A (en) * | 2021-04-30 | 2021-07-27 | 燕山大学 | Soft measurement method for cement quality based on fuzzy fine-grained feature extraction |
| CN113177358B (en) * | 2021-04-30 | 2022-06-03 | 燕山大学 | Soft measurement method for cement quality based on fuzzy fine-grained feature extraction |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102693452A (en) | Multiple-model soft-measuring method based on semi-supervised regression learning | |
| CN101630376B (en) | Soft-sensing modeling method and soft meter of multi-model neural network in biological fermentation process | |
| CN111199016A (en) | DTW-based improved K-means daily load curve clustering method | |
| CN110728411B (en) | High-low altitude area combined rainfall prediction method based on convolutional neural network | |
| CN111369070B (en) | Multimode fusion photovoltaic power prediction method based on envelope clustering | |
| CN108985380B (en) | Point switch fault identification method based on cluster integration | |
| CN108241901B (en) | Transformer early warning evaluation method and device based on prediction data | |
| CN110400231B (en) | Failure rate estimation method for electric energy metering equipment based on weighted nonlinear Bayes | |
| CN104914723A (en) | Industrial process soft measurement modeling method based on cooperative training partial least squares model | |
| CN110648014A (en) | Regional wind power prediction method and system based on space-time quantile regression | |
| CN102004444A (en) | Multi-model predictive control method for component content in process of extracting rare earth | |
| CN113344019A (en) | K-means algorithm for improving decision value selection initial clustering center | |
| CN105868164A (en) | Soft measurement modeling method based on monitored linear dynamic system model | |
| CN108334988A (en) | A kind of short-term Load Forecasting based on SVM | |
| Su et al. | New particle formation event detection with Mask R-CNN | |
| CN102621953A (en) | Automatic online quality monitoring and prediction model updating method for rubber hardness | |
| CN104778480A (en) | Hierarchical spectral clustering method based on local density and geodesic distance | |
| CN108830405B (en) | Real-time power load prediction system and method based on multi-index dynamic matching | |
| Decheng et al. | Improved quantum clustering analysis based on the weighted distance and its application | |
| CN114117852A (en) | Regional heat load rolling prediction method based on finite difference working domain division | |
| CN101702172A (en) | Data discretization method based on category-attribute relation dependency | |
| CN110517730A (en) | A method of thermophilic protein is identified based on machine learning | |
| CN102122349B (en) | Method for building multi-classification support vector machine classifier based on Bhattacharyya distance and directed acyclic graph | |
| Zhang et al. | Applicability evaluation of different algorithms for daily reference evapotranspiration model in KBE system | |
| CN114297582A (en) | Modeling method of discrete counting data based on multi-probe locality sensitive Hash negative binomial regression model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20120926 |