CN110413494A - A kind of LightGBM method for diagnosing faults improving Bayes's optimization - Google Patents
A kind of LightGBM method for diagnosing faults improving Bayes's optimization Download PDFInfo
- Publication number
- CN110413494A CN110413494A CN201910532139.3A CN201910532139A CN110413494A CN 110413494 A CN110413494 A CN 110413494A CN 201910532139 A CN201910532139 A CN 201910532139A CN 110413494 A CN110413494 A CN 110413494A
- Authority
- CN
- China
- Prior art keywords
- hyper parameter
- lightgbm
- model
- bayes
- function
- 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.)
- Granted
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
-
- 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
- G06F11/3457—Performance evaluation by simulation
-
- 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)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Biology (AREA)
- Computer Hardware Design (AREA)
- Quality & Reliability (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of LightGBM method for diagnosing faults of improvement Bayes optimization, comprising the following steps: 1) determines the hyper parameter and hyper parameter value range that LightGBM model needs to optimize；2) Bayesian Optimization Algorithm is improved, obtains improving Bayesian Optimization Algorithm GP-ProbHedge；3) it combines five folding cross validation modes to choose the optimal hyper parameter of fault diagnosis model using the method for step 2) to combine；4) building improves Bayes's optimization LightGBM fault diagnosis model, and provides model iterative process and optimum results.By using above-mentioned technology, compared with prior art, the present invention proposes that a kind of improved Bayesian Optimization Algorithm optimizes selection to the parameter of fault model, it is improved by the covariance function of acquisition function and its Gaussian process to traditional Bayesian Optimization Algorithm, it proposes that improving Bayes optimizes LightGBM method for diagnosing faults simultaneously, diagnosis prediction is carried out to the failure of equipment.
Description
Technical field
The present invention relates to fault diagnosis technology fields more particularly to a kind of LightGBM failure for improving Bayes's optimization to examine
Disconnected method.
Background technique
The research hotspot of fault diagnosis is mainly the data-driven method for diagnosing faults based on machine learning algorithm at present, but
It is the fault diagnosis model based on this method building there are still the uncertain situation of quantity of parameters, leads to Symbolic fault diagnosis precision
Fluctuate big problem.The variation of different hyper parameters the predictablity rate of fault diagnosis model is influenced it is very big, choose one compared with
Good parameter combination can play the superior function of fault diagnosis model, and largely promote fault diagnosis discrimination, from
And great economic benefit is brought to enterprise.So how to choose the hyper parameter combination that model can be made to reach highest recognition accuracy
Will be most important, also increasingly by the attention of researcher.
Summary of the invention
For the above-mentioned problems in the prior art, fault diagnosis accuracy band is given in order to solve hyper parameter uncertainty
Come the problem of influence, the object of the invention with a kind of LightGBM method for diagnosing faults of improvement Bayes optimization is provided, be used for
Solve the problems, such as that hyper parameter uncertainty is affected to fault diagnosis model accuracy, so as to improve fault diagnosis precision and mould
Type robustness.
Technical scheme is as follows:
A kind of LightGBM method for diagnosing faults improving Bayes's optimization, it is characterised in that the following steps are included:
1) hyper parameter and hyper parameter value range that LightGBM model needs to optimize are determined；
2) Bayesian Optimization Algorithm is improved, obtains improving Bayesian Optimization Algorithm GP-ProbHedge；
3) it combines five folding cross validation modes to choose the optimal hyper parameter of fault diagnosis model using the method for step 2) to combine；
4) building improves Bayes's optimization LightGBM fault diagnosis model, and provides model iterative process and optimization knot
Fruit.
A kind of LightGBM method for diagnosing faults of improvement Bayes optimization, it is characterised in that the determination of step 1)
The hyper parameter and hyper parameter value range that LightGBM model needs to optimize are as follows:
Hyper parameter max_depth sets value range as section [1,11]；
Hyper parameter learning_rate sets value range as section [0.1,0.9]；
Hyper parameter colsample_bytree sets value range as section [0.1,0.9]；
Hyper parameter subsample sets value range as section [0.1,0.9]；
Hyper parameter max_bin sets value range as section [25,150].
A kind of LightGBM method for diagnosing faults of described improvement Bayes optimization, it is characterised in that step 2) to shellfish
The step of this optimization algorithm of leaf improves is as follows:
2.1) Gaussian process is improved: being used to using the covariance function of more strategy combination forms while being captured objective function
Flatness and amplitude, shown in covariance function such as formula (1):
Wherein, r=| λ-λ ' |, λ indicates that evaluation point, the next evaluation point of λ ' expression, l are the hyper parameter of covariance function,Belong to Mat é rn covariance function cluster, for capturing the fluctuation situation of objective function；With
To capture the Secular Variation Tendency of objective function, ρ1And ρ2Two coefficients respectively in formula, for indicating the corresponding combination form
Covariance function be partial to capture target overall trend or periodicity；
It is optimal that hyper parameter uses the method for maximizing marginal likelihood to obtain, specific as follows: ρ in wushu1And ρ2As super ginseng
Number, then the hyper parameter θ={ ρ being related in the covariance function of this combining form1,ρ2, l } and by maximizing marginal likelihood come excellent
Change, shown in the marginal likelihood such as formula (2) of log form:
Wherein: y indicates history observation, σnIndicate variance, I indicates that unit matrix, k indicate that covariance function, p indicate general
Rate；Above-mentioned marginal likelihood is maximized by gradient descent method, is obtained an optimal estimation to θ, is come more using optimal θ
New Gauss model, convenient for the iterative search best-evaluated point of next step；
2.2) acquisition function improves: improving on the basis of Bayesian Optimization Algorithm GP-Hedge, after obtaining improvement
Acquisition function GP-ProbHedge, the acquisition function t walk iteration when, calculate every kind acquisition function maximum value obtain
Corresponding candidate point, and optimal point is chosen from all candidate points, every kind is acquired the candidate point and its Effective Probability of function
It combines,As best-evaluated point, N indicates acquisition function number in formula.
Detailed process is as follows: setting parameter η and initial accumulated earnings firstThen according to the calculating pair of each acquisition function
The candidate point answeredSimultaneously according to initial accumulated earningsCalculate corresponding Effective Probability pt(j), according topt(j) and most
Best-evaluated point is calculated in the calculation formula of whole evaluation point, and final updating raw data set is DtAnd the association side of Gauss model
Poor hyper parameter θ, and updating accumulated earnings at this timeEnter next round iteration later；
2.3) it improves Bayesian Optimization Algorithm to realize, including calculates step as follows:
Step 2.3.1) according to the acquisition function GP-ProbHedge calculation of proposition, and 5 hyper parameters of initialization
Sample point obtains next best-evaluated point λt；
Step 2.3.2) according to λtAnd Gaussian process calculating target function value yt；
Step 2.3.3) usage history sample D1:t-1And new samples (λt,yt) more new data set be D1:t；
Step 2.3.4) update the more strategy combination form covariance functions of GP model hyper parameter θ={ ρ1,ρ2,l}；
Step 2.3.5) calculate current coefficient of correspondenceAnd total revenue
A kind of LightGBM method for diagnosing faults of improvement Bayes optimization, it is characterised in that the use of step 3)
Step 2) method combines five folding cross validation modes to choose the optimal hyper parameter combination of fault diagnosis model, the five foldings cross validation
Mode is as follows: first the initial data of acquisition being upset, is then randomly divided into five parts, is successively trained using four parts of data therein
Model uses remaining a data to collect as verifying, to detect the identification classification accuracy of fault diagnosis model, finally takes and is testing
Card collects the average value of five parts of obtained accuracy to assess diagnosis performance of the model in failure problems.
A kind of LightGBM method for diagnosing faults of improvement Bayes optimization, it is characterised in that the building of step 4)
It improves Bayes and optimizes LightGBM fault diagnosis model, and provide model iterative process and optimum results, detailed process is as follows:
The hyper parameter for needing to optimize is determined according to LightGBM fault diagnosis model first, then sets each hyper parameter
Value range, using Bayes's optimization method is improved, simultaneously five folding cross validation modes choose optimal hyper parameter combination, are corresponded to
Improvement Bayes optimize LightGBM fault diagnosis model, and record cast iterative process and optimum results.
By using above-mentioned technology, compared with prior art:
1) present invention proposes that a kind of improved Bayesian Optimization Algorithm optimizes selection to the parameter of fault model, passes through
The covariance function of acquisition function and its Gaussian process to traditional Bayesian Optimization Algorithm improves, while proposing to improve shellfish
Ye Si optimizes LightGBM method for diagnosing faults, carries out diagnosis prediction to the failure of equipment.
2) this method is gathered around than common grid search Optimizing fault diagnosis method with Stochastic search optimization method for diagnosing faults
There is better performance, model computation complexity is low, and failure predication efficiency and accuracy are high.
3) the invention proposes a kind of LightGBM method for diagnosing faults of improvement Bayes optimization, for solving hyper parameter
The problem of uncertainty is affected to fault diagnosis model accuracy, so as to improve fault diagnosis precision and model robustness.
Detailed description of the invention
Fig. 1 is to improve Bayes's Optimization Framework；
Fig. 2 is five folding cross validation modes；
Fig. 3 is to improve Bayes to optimize LightGBM fault diagnosis model building process.
Specific embodiment
Below in conjunction with Figure of description, the invention will be further described, but protection scope of the present invention is not limited to
This:
Improve the LightGBM method for diagnosing faults of Bayes's optimization, comprising the following steps:
1) hyper parameter and hyper parameter value range that LightGBM model needs to optimize are determined；The determination of step 1)
The hyper parameter and hyper parameter value range that LightGBM model needs to optimize, in the following ways:
Hyper parameter max_depth sets value range as section [1,11]；
Hyper parameter learning_rate sets value range as section [0.1,0.9]；
Hyper parameter colsample_bytree sets value range as section [0.1,0.9]；
Hyper parameter subsample sets value range as section [0.1,0.9]；
Hyper parameter max_bin sets value range as section [25,150].
2) Bayesian Optimization Algorithm is improved；
2.1) Gaussian process is improved: in Gauss model initialization, covariance function therein is to calculate two data points
Between similitude function, it specifies the flatness and amplitude of unknown object function.Therefore, straight to the selection of covariance function
The matching degree affected between Gaussian process and data character is connect, while single covariance function can not be in all problems
On show best performance, and inappropriate covariance hyper parameter also results in the estimation inaccuracy of Posterior distrbutionp.Based on this,
The covariance function using more strategy combination forms is proposed, for capturing the flatness and amplitude of objective function, form simultaneously
Are as follows:
Wherein, r=| λ-λ ' |, λ indicates that evaluation point, the next evaluation point of λ ' expression, l are the super ginseng of covariance function
Number, formula first item belong to Mat é rn covariance function cluster, have high flexibility, for capturing the fluctuation situation of target, formula
Section 2 is used to capture the Secular Variation Tendency of target, ρ1And ρ2Two coefficients respectively in formula, they are used to indicate the group
The covariance function of conjunction form is partial to capture target overall trend or periodicity.The covariance function of this combining form is opposite
In single covariance function more robust, it is adapted to the solution of the Function Fitting in a variety of situations, makes Gaussian process and mesh
The matching degree marked between property is higher, guarantees to obtain more preferably prediction result with this.
ρ in wushu1And ρ2It is equally used as hyper parameter, complete Bayes, which is handled in the covariance function of this combining form, to be related to
And the hyper parameter θ={ ρ arrived1,ρ2, l } and optimized by maximizing marginal likelihood.The marginal likelihood of its log form are as follows:
Above-mentioned likelihood, an available optimal estimation to θ are maximized by gradient descent method.
When maximum likelihood method optimizes and calculates hyper parameter, its complexity is this more strategy combination form covariance function
Twice of O (n3), although computation complexity can be improved, it ensure that the flatness and amplitude of unknown object function, adaptively
Learn the coefficient of correspondence of covariance function to be fitted various situations, improves model quality.
2.2) acquisition function improves: in the latest Progress of acquisition function, using the Bayes of single acquisition function
Optimization algorithm cannot all show best performance on all problems.Therefore in order to obtain one with strong robustness
Method, Shahriari etc. propose a kind of combined strategy ESP (Entropy Search Portfolio, ESP) based on information
Method.This method forms candidate point set the candidate point that function obtains is acquired from every kind in each iteration, is then based on
Entropy searching method selects that the candidate point of most information can be provided as evaluation point for globally optimal solution.Brochu etc. proposes one kind
Use the Bayesian Optimization Algorithm of a variety of acquisition functions of the strategy combination that liquidates, referred to as GP-Hedge.The difference is that it should with ESP
Method chooses evaluation point according to the strategy that liquidates from candidate point.The so-called strategy that liquidates is exactly according to the effective of each acquisition function i
Probability p (i) chooses evaluation point, and Probability p (i) is calculated according to the accumulated earnings of acquisition function i.
Although the above method obtains best-evaluated point by comparing the candidate point that each acquisition function provides, substantially
What the evaluation point was still only obtained from single acquisition function, therefore the information that the point includes also only is determined by the single acquisition function
It is fixed, have ignored other acquisition possible incomes of function.Based on this, one kind is proposed on the basis of the GP-Hedge of the studies above
Improved acquisition function, is named as GP-ProbHedge.The acquisition function can equally calculate every kind when t walks iteration
The maximum value of acquisition function obtains corresponding candidate point, but is different from GP-Hedge and chooses optimal point from all candidate points,
It combines the candidate point that every kind acquires function with its Effective Probability,As best-evaluated point,
N indicates acquisition function number in formula.
Table 1 is the specific calculation process of GP-ProbHedge.Parameter η and initial accumulated earnings are set firstThen root
Corresponding candidate point is calculated according to each acquisition functionSimultaneously according to accumulated earningsCalculate corresponding Effective Probability pt(j), root
According topt(j) and final evaluation point λ is calculated in the formula of propositiont, final updating raw data set is DtAnd Gauss model
Covariance hyper parameter θ, and update accumulated earnings at this time calculatingEnter next round iteration later.Improved acquisition letter
Number not only possesses whole advantages of GP-Hedge method, but also has merged the information in acquisition function not of the same race, and this method can
There is the acquisition function of performance difference in tolerance combination, the tactful initial stage that solves the problems, such as to liquidate may make wrong choice, with more
Strong robustness and higher global optimization efficiency.
1 GP-ProbHedge calculation process of table
2.3) it improves Bayesian Optimization Algorithm to realize: in conclusion improved Bayesian Optimization Algorithm frame such as Fig. 1 institute
Show.Steps are as follows for its specific calculating:
Step 2.3.1) according to the acquisition function GP-ProbHedge calculation of proposition, and 5 samples initialized,
Obtain next best-evaluated point λt。
Step 2.3.2) according to λtAnd GP model calculating target function value yt。
Step 2.3.3) usage history sample D1:t-1And new samples (λt,yt) more new data set be D1:t.(in t-1 iteration
All samples, the new samples of current t step are (λt,yt), altogether it is exactly all samples of t iteration).
Step 2.3.4) update the more strategy combination form covariance functions of GP model hyper parameter θ={ ρ1,ρ2,l}。
Step 2.3.5) calculate current coefficient of correspondenceAnd total revenueDetailed process is as shown in table 1.
3) it combines five folding cross validation modes to choose the optimal hyper parameter of fault diagnosis model using step 2) method to combine；
3.1) in order to ensure the stability and accuracy of model prediction result, herein by the way of five folding cross validations,
First the initial data of acquisition is upset, is then randomly divided into five parts, successively carrys out training pattern using four parts of data therein, is used
Remaining portion data are as verifying collection, to detect the identification classification accuracy of fault diagnosis model, division mode such as Fig. 2 institute
Show.The average value of the five parts of accuracy obtained on verifying collection is finally taken to assess diagnosis performance of the model in failure problems.
4) building improves Bayes's optimization LightGBM fault diagnosis model, and provides model iterative process and optimum results.
It is illustrated in figure 3 and improves Bayes's optimization LightGBM fault diagnosis model building process.It is examined first according to failure
Disconnected model determines the hyper parameter set for needing to optimize, and then sets the value interval of each hyper parameter, and using improving, Bayes is excellent
Change method combines five folding cross validation modes to obtain corresponding fault diagnosis model, and records iterative process and optimum results.
In order to guarantee the reliability of fault data and the repeatability of result, fault data derives from UCI machine learning
Database.It is the fault data of stainless steel plate, it shares seven kinds of possible defect failure (spot corrosion, Z_ scratch, K_ scratches, stain
Trace, dirty, salient point, other failures), 30 dimensional parameters, totally 1941 data instances.The research purpose of the data is correct
The type of Forecasting recognition stainless steel surface defect failure.
Table 2 is based on above-mentioned data set, using the iterative calculation for improving Bayes's optimization LightGBM fault diagnosis model
Process.Serial number 1-5 is five sample points of initialization, the result that serial number 6-20 is Optimized Iterative 15 times after initialization.From table
In it can be found that improve Bayes optimize LightGBM model utilize existing five sample point information, capture model quickly
The fluctuation situation that objective function is really distributed, and quickly reach higher accuracy 97.5%, then by subsequent iteration several times,
Posterior distrbutionp perfection is gradually set to be fitted true distribution situation.It improves Bayes and optimizes the standard that LightGBM fault diagnosis model obtains
Exactness extreme value is 98.2%, and extraordinary model hyper parameter is only just achieved under fewer iterations number.As it can be seen that improving Bayes
Optimize LightGBM method for diagnosing faults, required calculating time complexity is lower, and failure predication accuracy is higher, has centainly
Practicability and validity.
Table 2 improves BOA-LightGBM diagnostic model calculated result
。
Claims (5)
1. a kind of LightGBM method for diagnosing faults for improving Bayes's optimization, it is characterised in that the following steps are included:
1) hyper parameter and hyper parameter value range that LightGBM model needs to optimize are determined；
2) Bayesian Optimization Algorithm is improved, obtains improving Bayesian Optimization Algorithm GP-ProbHedge；
3) it combines five folding cross validation modes to choose the optimal hyper parameter of fault diagnosis model using the method for step 2) to combine；
4) building improves Bayes's optimization LightGBM fault diagnosis model, and provides model iterative process and optimum results.
2. a kind of LightGBM method for diagnosing faults for improving Bayes's optimization according to claim 1, it is characterised in that
The hyper parameter and hyper parameter value range that the determination LightGBM model of step 1) needs to optimize are as follows:
Hyper parameter max_depth sets value range as section [1,11]；
Hyper parameter learning_rate sets value range as section [0.1,0.9]；
Hyper parameter colsample_bytree sets value range as section [0.1,0.9]；
Hyper parameter subsample sets value range as section [0.1,0.9]；
Hyper parameter max_bin sets value range as section [25,150].
3. a kind of LightGBM method for diagnosing faults for improving Bayes's optimization according to claim 1, it is characterised in that
Step 2) it is as follows the step of being improved to Bayesian Optimization Algorithm:
2.1) Gaussian process is improved: being used to using the covariance function of more strategy combination forms while being captured the smooth of objective function
Property and amplitude, shown in covariance function such as formula (1):
Wherein, r=| λ-λ ' |, λ indicates that evaluation point, the next evaluation point of λ ' expression, l are the hyper parameter of covariance function,Belong to Mat é rn covariance function cluster, for capturing the fluctuation situation of objective function；
For capturing the Secular Variation Tendency of objective function, ρ1And ρ2Two coefficients respectively in formula, for indicating the corresponding combination shape
The covariance function of formula is partial to capture target overall trend or periodicity；
It is optimal that hyper parameter uses the method for maximizing marginal likelihood to obtain, specific as follows: ρ in wushu1And ρ2As hyper parameter,
The hyper parameter θ={ ρ being then related in the covariance function of this combining form1,ρ2, l } optimized by maximizing marginal likelihood,
Shown in the marginal likelihood such as formula (2) of its log form:
Wherein: y indicates history observation, σnIndicate variance, I indicates that unit matrix, k indicate that covariance function, p indicate probability；It is logical
Gradient descent method is crossed to maximize above-mentioned marginal likelihood, an optimal estimation to θ is obtained, carrys out more new peak using optimal θ
This model, convenient for the iterative search best-evaluated point of next step；
2.2) acquisition function improves: improving on the basis of Bayesian Optimization Algorithm GP-Hedge, obtains improved adopt
Set function GP-ProbHedge, when t walks iteration, the maximum value for calculating every kind of acquisition function is corresponded to the acquisition function
Candidate point, and choose optimal point from all candidate points, the candidate point that every kind acquires function mutually tied with its Effective Probability
It closes,As best-evaluated point, N indicates acquisition function number in formula；
Detailed process is as follows: setting parameter η and initial accumulated earnings firstThen it is calculated according to each acquisition function corresponding
Candidate pointSimultaneously according to initial accumulated earningsCalculate corresponding Effective Probability pt(j), according topt(j) and finally it assesses
Best-evaluated point is calculated in the calculation formula of point, and final updating raw data set is DtAnd the super ginseng of covariance of Gauss model
Number θ, and updating accumulated earnings at this timeEnter next round iteration later；
2.3) it improves Bayesian Optimization Algorithm to realize, including calculates step as follows:
Step 2.3.1) according to the acquisition function GP-ProbHedge calculation of proposition, and 5 hyper parameter samples of initialization
Point obtains next best-evaluated point λt；
Step 2.3.2) according to λtAnd Gaussian process calculating target function value yt；
Step 2.3.3) usage history sample D1:t-1And new samples (λt,yt) more new data set be D1:t；
Step 2.3.4) update the more strategy combination form covariance functions of GP model hyper parameter θ={ ρ1,ρ2,l}；
Step 2.3.5) calculate current coefficient of correspondenceAnd total revenue
4. a kind of LightGBM method for diagnosing faults for improving Bayes's optimization according to claim 1, it is characterised in that
The use step 2) method of step 3) combines five folding cross validation modes to choose the optimal hyper parameter combination of fault diagnosis model, described
Five folding cross validation modes are as follows: first the initial data of acquisition being upset, is then randomly divided into five parts, successively uses therein four
Part data carry out training pattern, and remaining a data is used to collect as verifying, to detect the identification classification of fault diagnosis model accurately
Degree finally takes the average value of the five parts of accuracy obtained on verifying collection diagnostic in failure problems to assess the model
Energy.
5. a kind of LightGBM method for diagnosing faults for improving Bayes's optimization according to claim 1, it is characterised in that
The building of step 4) improves Bayes and optimizes LightGBM fault diagnosis model, and provides model iterative process and optimum results,
Detailed process is as follows:
The hyper parameter for needing to optimize is determined according to LightGBM fault diagnosis model first, then sets the value of each hyper parameter
Range, using Bayes's optimization method is improved, simultaneously five folding cross validation modes choose optimal hyper parameter combination, obtain corresponding change
Optimize LightGBM fault diagnosis model, and record cast iterative process and optimum results into Bayes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910532139.3A CN110413494B (en) | 2019-06-19 | 2019-06-19 | LightGBM fault diagnosis method for improving Bayesian optimization |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910532139.3A CN110413494B (en) | 2019-06-19 | 2019-06-19 | LightGBM fault diagnosis method for improving Bayesian optimization |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110413494A true CN110413494A (en) | 2019-11-05 |
| CN110413494B CN110413494B (en) | 2022-05-13 |
Family
ID=68359301
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910532139.3A Active CN110413494B (en) | 2019-06-19 | 2019-06-19 | LightGBM fault diagnosis method for improving Bayesian optimization |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110413494B (en) |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110889450A (en) * | 2019-11-27 | 2020-03-17 | 腾讯科技（深圳）有限公司 | Method and device for super-parameter tuning and model building |
| CN111027709A (en) * | 2019-11-29 | 2020-04-17 | 腾讯科技（深圳）有限公司 | Information recommendation method and device, server and storage medium |
| CN111176832A (en) * | 2019-12-06 | 2020-05-19 | 重庆邮电大学 | Performance optimization and parameter configuration method based on memory computing framework Spark |
| CN111340110A (en) * | 2020-02-25 | 2020-06-26 | 湖南师范大学 | Fault early warning method based on industrial process running state trend analysis |
| CN111475921A (en) * | 2020-03-13 | 2020-07-31 | 重庆邮电大学 | Tool residual life prediction method based on edge calculation and L STM network |
| CN111711608A (en) * | 2020-05-20 | 2020-09-25 | 南方电网调峰调频发电有限公司信息通信分公司 | Method and system for detecting abnormal flow of power data network and electronic equipment |
| CN111859264A (en) * | 2020-07-09 | 2020-10-30 | 北京工商大学 | Time sequence prediction method and device based on Bayes optimization and wavelet decomposition |
| CN111898212A (en) * | 2020-06-17 | 2020-11-06 | 西安交通大学 | Impeller mechanical profile design optimization method based on BezierGAN and Bayesian optimization |
| CN111901330A (en) * | 2020-07-24 | 2020-11-06 | 中移（杭州）信息技术有限公司 | Ensemble learning model construction method, ensemble learning model identification device, server and medium |
| CN111984872A (en) * | 2020-09-09 | 2020-11-24 | 北京中科研究院 | Multi-modal information social media popularity prediction method based on iterative optimization strategy |
| CN112597686A (en) * | 2020-12-30 | 2021-04-02 | 复旦大学 | Finite element analysis parameter Bayes optimization method and device |
| CN112598148A (en) * | 2020-09-30 | 2021-04-02 | 新天绿色能源股份有限公司 | Fan variable pitch motor temperature fault early warning method based on collaborative expression and LightGBM algorithm |
| CN113313049A (en) * | 2021-06-11 | 2021-08-27 | 北京百度网讯科技有限公司 | Method, device, equipment, storage medium and computer program product for determining hyper-parameters |
| CN113655391A (en) * | 2021-08-26 | 2021-11-16 | 江苏慧智能源工程技术创新研究院有限公司 | Energy storage power station battery fault diagnosis method based on LightGBM model |
| CN113707228A (en) * | 2021-07-29 | 2021-11-26 | 北京工业大学 | LightGBM algorithm-based wet flue gas desulfurization optimization method |
| CN113792762A (en) * | 2021-08-24 | 2021-12-14 | 华南理工大学 | Water chilling unit fault diagnosis method, system and medium based on Bayesian optimization LightGBM |
| CN114860298A (en) * | 2022-03-25 | 2022-08-05 | 上海师范大学 | Bayes-LightGBM software defect prediction method based on RFECV feature selection |
| CN115544115A (en) * | 2022-09-24 | 2022-12-30 | 湖南万脉医疗科技有限公司 | Respirator fault prediction method and respirator |
| CN117763356A (en) * | 2023-12-26 | 2024-03-26 | 中国地质科学院地质力学研究所 | Rapid earthquake phase identification method based on LightGBM algorithm |
| CN118509592A (en) * | 2024-07-18 | 2024-08-16 | 宁波康达凯能医疗科技有限公司 | Inter-frame coding method, device and medium based on semantic segmentation under Bayesian framework |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109829236A (en) * | 2019-01-31 | 2019-05-31 | 浙江工业大学 | A kind of Compressor Fault Diagnosis method based on XGBoost feature extraction |
-
2019
- 2019-06-19 CN CN201910532139.3A patent/CN110413494B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109829236A (en) * | 2019-01-31 | 2019-05-31 | 浙江工业大学 | A kind of Compressor Fault Diagnosis method based on XGBoost feature extraction |
Non-Patent Citations (2)
| Title |
|---|
| 谭江红等: "一种机器学习方法在湖北定时气温预报中的应用试验", 《气象科技进展》 * |
| 高欣等: "基于模型自适应选择融合的智能电表故障多分类方法", 《电网技术》 * |
Cited By (29)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110889450B (en) * | 2019-11-27 | 2023-08-11 | 腾讯科技（深圳）有限公司 | Super-parameter tuning and model construction method and device |
| CN110889450A (en) * | 2019-11-27 | 2020-03-17 | 腾讯科技（深圳）有限公司 | Method and device for super-parameter tuning and model building |
| CN111027709A (en) * | 2019-11-29 | 2020-04-17 | 腾讯科技（深圳）有限公司 | Information recommendation method and device, server and storage medium |
| CN111176832A (en) * | 2019-12-06 | 2020-05-19 | 重庆邮电大学 | Performance optimization and parameter configuration method based on memory computing framework Spark |
| CN111176832B (en) * | 2019-12-06 | 2022-07-01 | 重庆邮电大学 | Performance optimization and parameter configuration method based on memory computing framework Spark |
| CN111340110A (en) * | 2020-02-25 | 2020-06-26 | 湖南师范大学 | Fault early warning method based on industrial process running state trend analysis |
| CN111340110B (en) * | 2020-02-25 | 2023-04-18 | 湖南师范大学 | Fault early warning method based on industrial process running state trend analysis |
| CN111475921A (en) * | 2020-03-13 | 2020-07-31 | 重庆邮电大学 | Tool residual life prediction method based on edge calculation and L STM network |
| CN111711608B (en) * | 2020-05-20 | 2022-06-21 | 南方电网调峰调频发电有限公司信息通信分公司 | Method and system for detecting abnormal flow of power data network and electronic equipment |
| CN111711608A (en) * | 2020-05-20 | 2020-09-25 | 南方电网调峰调频发电有限公司信息通信分公司 | Method and system for detecting abnormal flow of power data network and electronic equipment |
| CN111898212A (en) * | 2020-06-17 | 2020-11-06 | 西安交通大学 | Impeller mechanical profile design optimization method based on BezierGAN and Bayesian optimization |
| CN111859264B (en) * | 2020-07-09 | 2024-02-02 | 北京工商大学 | Time sequence prediction method and device based on Bayesian optimization and wavelet decomposition |
| CN111859264A (en) * | 2020-07-09 | 2020-10-30 | 北京工商大学 | Time sequence prediction method and device based on Bayes optimization and wavelet decomposition |
| CN111901330A (en) * | 2020-07-24 | 2020-11-06 | 中移（杭州）信息技术有限公司 | Ensemble learning model construction method, ensemble learning model identification device, server and medium |
| CN111984872A (en) * | 2020-09-09 | 2020-11-24 | 北京中科研究院 | Multi-modal information social media popularity prediction method based on iterative optimization strategy |
| CN112598148A (en) * | 2020-09-30 | 2021-04-02 | 新天绿色能源股份有限公司 | Fan variable pitch motor temperature fault early warning method based on collaborative expression and LightGBM algorithm |
| CN112597686B (en) * | 2020-12-30 | 2022-02-22 | 复旦大学 | Finite element analysis parameter Bayes optimization method and device |
| CN112597686A (en) * | 2020-12-30 | 2021-04-02 | 复旦大学 | Finite element analysis parameter Bayes optimization method and device |
| CN113313049A (en) * | 2021-06-11 | 2021-08-27 | 北京百度网讯科技有限公司 | Method, device, equipment, storage medium and computer program product for determining hyper-parameters |
| CN113707228B (en) * | 2021-07-29 | 2024-04-16 | 北京工业大学 | Wet flue gas desulfurization optimization method based on LightGBM algorithm |
| CN113707228A (en) * | 2021-07-29 | 2021-11-26 | 北京工业大学 | LightGBM algorithm-based wet flue gas desulfurization optimization method |
| CN113792762A (en) * | 2021-08-24 | 2021-12-14 | 华南理工大学 | Water chilling unit fault diagnosis method, system and medium based on Bayesian optimization LightGBM |
| CN113792762B (en) * | 2021-08-24 | 2023-06-20 | 华南理工大学 | Water chilling unit fault diagnosis method, system and medium based on Bayesian optimization LightGBM |
| CN113655391A (en) * | 2021-08-26 | 2021-11-16 | 江苏慧智能源工程技术创新研究院有限公司 | Energy storage power station battery fault diagnosis method based on LightGBM model |
| CN114860298A (en) * | 2022-03-25 | 2022-08-05 | 上海师范大学 | Bayes-LightGBM software defect prediction method based on RFECV feature selection |
| CN114860298B (en) * | 2022-03-25 | 2024-09-13 | 上海师范大学 | RFECV feature selection-based Bayes-LightGBM software defect prediction method |
| CN115544115A (en) * | 2022-09-24 | 2022-12-30 | 湖南万脉医疗科技有限公司 | Respirator fault prediction method and respirator |
| CN117763356A (en) * | 2023-12-26 | 2024-03-26 | 中国地质科学院地质力学研究所 | Rapid earthquake phase identification method based on LightGBM algorithm |
| CN118509592A (en) * | 2024-07-18 | 2024-08-16 | 宁波康达凯能医疗科技有限公司 | Inter-frame coding method, device and medium based on semantic segmentation under Bayesian framework |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110413494B (en) | 2022-05-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110413494A (en) | A kind of LightGBM method for diagnosing faults improving Bayes's optimization | |
| CN109597968B (en) | SMT big data-based solder paste printing performance influence factor analysis method | |
| CN108921604B (en) | Advertisement click rate prediction method based on cost-sensitive classifier integration | |
| CN106599230A (en) | Method and system for evaluating distributed data mining model | |
| CN108416373A (en) | A kind of unbalanced data categorizing system based on regularization Fisher threshold value selection strategies | |
| CN109255029A (en) | A method of automatic Bug report distribution is enhanced using weighted optimization training set | |
| CN106127262A (en) | The clustering method of one attribute missing data collection | |
| CN112557034A (en) | Bearing fault diagnosis method based on PCA _ CNNS | |
| CN113947157B (en) | Dynamic brain effect connection network generation method based on hierarchical clustering and structural equation model | |
| CN110390358A (en) | A kind of deep learning method based on feature clustering | |
| CN118114083B (en) | Student academic performance prediction method based on big data | |
| CN109447153A (en) | Divergence-excitation self-encoding encoder and its classification method for lack of balance data classification | |
| CN115791174A (en) | Rolling bearing abnormity diagnosis method and system, electronic equipment and storage medium | |
| Wu et al. | Evolutionary feature construction for ultrasound image processing and its application to automatic liver disease diagnosis | |
| CN118114031A (en) | Radio waveform prediction method and system based on machine learning | |
| CN113128518B (en) | Sift mismatch detection method based on twin convolution network and feature mixing | |
| CN105824785A (en) | Rapid abnormal point detection method based on penalized regression | |
| CN110298399B (en) | Freeman chain code and moment feature fusion-based pumping well fault diagnosis method | |
| CN109389313B (en) | Fault classification diagnosis method based on weighted neighbor decision | |
| CN116350234A (en) | ECG arrhythmia classification method and system based on GCNN-LSTM model | |
| CN111199209A (en) | Bearing time-frequency spectrogram identification method based on IWO-KFCM algorithm | |
| CN116313111A (en) | Breast cancer risk prediction method, system, medium and equipment based on combined model | |
| CN113296947B (en) | Resource demand prediction method based on improved XGBoost model | |
| CN114463268A (en) | Image analysis method based on Bayesian deep learning | |
| CN112884748A (en) | Non-woven fabric surface small defect detection method based on multi-core support vector machine |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |