CN110490234A - The construction method and classification method of classifier based on Cluster Classification associative mechanism - Google Patents
The construction method and classification method of classifier based on Cluster Classification associative mechanism Download PDFInfo
- Publication number
- CN110490234A CN110490234A CN201910652753.3A CN201910652753A CN110490234A CN 110490234 A CN110490234 A CN 110490234A CN 201910652753 A CN201910652753 A CN 201910652753A CN 110490234 A CN110490234 A CN 110490234A
- Authority
- CN
- China
- Prior art keywords
- population
- cluster
- classification
- classifier
- parameter
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biophysics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Medical Informatics (AREA)
- Physiology (AREA)
- Genetics & Genomics (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses the construction method and classification method of a kind of classifier based on Cluster Classification associative mechanism, classifier construction method therein includes: to obtain initialization population firstly, encoding the distribution situation of cluster centre into population at individual;Then optimal population is searched for using multi-objective optimization algorithm, as the Pareto disaggregation for meeting cluster two targets of objective function and class object function;Optimal solution is selected from Pareto solution concentration further according to mutual information index, and according to the optimal solution selected, the parameter of the classifier of joint Cluster Classification mechanism is configured, the classifier of the parameter building Cluster Classification associative mechanism based on setting.The present invention with method can be effectively by Cluster-Fusion into classification task, to improve accuracy and the search efficiency of classifier.
Description
Technical field
The present invention relates to data minings and big data technical field, and in particular to a kind of based on Cluster Classification associative mechanism
The construction method and classification method of classifier.
Background technique
In data mining and big data field, cluster and classification are two mutually independent processes.Cluster is based on unsupervised
Study can forecast sample label by the potential structure of mining data, be classified based on supervised learning, relationship square can be passed through
Battle array predicts the label of unknown data.When classification information gives, the two can receive benefits mutually.Based on this, fuzzy relation
Classifier proposes hereafter to improve the demand of aspect, parallel Cluster Classification calculation based on efficiency as serial Cluster Classification algorithm
Method proposes to carry out classification task.
At least there is following technology in implementing the present invention, it may, the method for finding the prior art in present inventor
Problem:
In the prior art, none multiple-objection optimization search strategy efficiently customized is applied to parallel clustering classification and calculates
Inside method.Also, current cluster objective function can spend it is a large amount of calculate the time and meanwhile instruct effect be not it is especially outstanding,
In addition, the frame of current parallel clustering sorting algorithm is not classic frame.
It follows that the bad technical problem of classifier performance exists in the prior art.
Summary of the invention
In view of this, the present invention provides a kind of construction method of classifier based on Cluster Classification associative mechanism and dividing
Class method, to solve or at least partly solve the bad technical problem of performance of classifier existing in the prior art.
The present invention provides a kind of construction methods of classifier based on Cluster Classification associative mechanism, comprising:
Step S1: the distribution situation of cluster centre is encoded into population at individual, initialization population is obtained, wherein is initial
Change includes cluster objective function and class object function in population;
Step S2: searching for optimal population using multi-objective optimization algorithm, clusters objective function as meeting and divides
The Pareto disaggregation of two targets of class objective function;
Step S3: optimal solution is selected from Pareto solution concentration according to mutual information index, and optimal according to what is selected
Solution, the parameter of classifier of joint Cluster Classification mechanism is configured, then the building Cluster Classification connection of the parameter based on setting
The classifier of conjunction mechanism.
In one embodiment, step S1 is specifically included:
Step S1.1: multiple populations are directed to, corresponding cluster centre distribution situation is set;
Step S1.2: coding cluster centre distribution situation, and it will be used as decision variable, according to fitness calculation formula, meter
Corresponding cluster target and class object value are calculated, and using corresponding cluster target as cluster objective function, class object value
As class object function, wherein shown in the concrete form such as formula (1) and formula (2) of cluster objective function and class object function:
In formula (1), f (xi) label predicted of presentation class device, yiExpression is really label, δ (f (xi), yi) indicate pre-
Whether mark label are consistent with true tag, consistent results 1, inconsistent results 0, and N indicates test data total quantity;
In formula (2), nnijIndicate distance sample xiMinimum distance j-th of sample, p (ck|nnij) indicate j-th of sample
Belong to the conditional probability of k-th of cluster,It is a parameter, works as nnijAnd xiWhen belonging to cluster k, it is 0, is otherwise 1/j;
Step S1.3: decision variable, cluster objective function and class object function are encapsulated into population at individual simultaneously,
Multiple individual composition initialization populations.
In one embodiment, step S2 is specifically included:
Step S2.1: it in former generation, is evolved using the crossover operator and mutation operator of difference algorithm to current population
Processing, obtains progeny population;
Step S2.2: in former generation, using backward learning algorithm, inverse change is carried out to progeny population, and will be reversed
The population of front and back carries out non-dominated ranking after merging, and obtains reversed sub- population;
Step S2.3: in former generation, using fireworks algorithm, calculating the burst radius and spark number of reversed sub- population,
And it is based on burst radius and spark number, to generate spark population in peripheral extent using reversed sub- population as fireworks population,
Fireworks population and spark population carry out non-dominated ranking and generate next-generation population;
Step S2.4: using next-generation population as former generation population is worked as, step S2.1~step S2.3 is repeated, is changed
For evolutionary process;
Step S2.5: when iterative evolution process meets termination condition, iterative evolution is terminated, and exports Pareto disaggregation.
In one embodiment, the termination condition in step S2.5 includes: that the number of iterations satisfaction reaches greatest iteration time
The whole variation of several and population is less than the minimum population variation of setting.
In one embodiment, step S3 is specifically included
Step S3.1: the association relationship of all Pareto disaggregation is calculated;
Step S3.2: according to the size of association relationship, the maximum Pareto disaggregation of association relationship is selected as optimal
Solution is configured the parameter of the classifier of joint Cluster Classification mechanism;
Step S3.3: the classifier of the parameter building Cluster Classification associative mechanism based on setting.
In one embodiment, after step S3.3, the method also includes step S3.4:
It is tested using classifier of the test data to building.
Based on same inventive concept, second aspect of the present invention provides a kind of based on classifier constructed by first aspect
Classification method, which includes:
After the cluster centre and clustering relationships that calculate data to be processed, classification results are obtained by classifier.This Shen
Please said one or multiple technical solutions in embodiment, at least there are following one or more technical effects:
The construction method of a kind of classifier based on Cluster Classification associative mechanism provided by the invention, firstly, initialization kind
Group, the distribution situation of cluster centre is encoded into population at individual, then searches for optimal population using multi-objective optimization algorithm,
Make it as the Pareto disaggregation for meeting two targets, selects optimal solution from Pareto solution concentration further according to mutual information index
The parameter setting of classifier as joint Cluster Classification mechanism.
Due to the construction method of the classifier of Cluster Classification associative mechanism provided by the invention, pass through multiple-objection optimization first
The optimal population of algorithm search, so as to find the relationship between optimal classifier, and using mutual information index as classification
The foundation of device parameter setting, so as to effectively by Cluster-Fusion into classification task, so the property of classifier can be improved
Can, improve the accuracy of classification.
Further, the cluster target letter based on silhouette coefficient design that the present invention passes through a kind of quick low complex degree simultaneously
Several and class object function carrys out subsidiary classification device, so as to further improve the performance of classifier.
Further, the present invention is searching for optimal population using multi-objective optimization algorithm, clusters mesh as meeting
When the Pareto disaggregation of two targets of scalar functions and class object function, automatic cluster can save initialization time, reversed to learn
The effective range for increasing search is practised, mutual information index is as last judgment basis, so as to improve the accuracy of classifier
And search efficiency.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is the present invention
Some embodiments for those of ordinary skill in the art without creative efforts, can also basis
These attached drawings obtain other attached drawings.
Fig. 1 is a kind of flow diagram of the construction method of the classifier based on Cluster Classification associative mechanism of the present invention;
Fig. 2 is the schematic diagram of the classifier based on Cluster Classification associative mechanism in the present invention;
Fig. 3 is the flow chart of the classifier of the Cluster Classification mechanism optimized based on fireworks algorithm and difference algorithm.
Specific embodiment
It is an object of the invention to provide a kind of based on poly- for the bad technical problem of classifier performance in the prior art
The construction method and classification method of the classifier of class classification associative mechanism, and achieve the purpose that improve classifier performance.
In order to achieve the above objectives, central scope of the invention is as follows:
A kind of construction method of classifier based on Cluster Classification associative mechanism is provided, firstly, initialization population, will cluster
The distribution situation at center is encoded into population at individual, is then searched for optimal population using multi-objective optimization algorithm, is made its conduct
The Pareto disaggregation for meeting two targets is concentrated from Pareto solution further according to mutual information index and selects optimal solution as joint
The parameter setting of the classifier of Cluster Classification mechanism.
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without creative efforts, shall fall within the protection scope of the present invention.
Embodiment one
A kind of construction method of classifier based on Cluster Classification associative mechanism is present embodiments provided, it referring to Figure 1, should
Method includes:
Step S1: the distribution situation of cluster centre is encoded into population at individual, initialization population is obtained, wherein is initial
Change includes cluster objective function and class object function in population.
Specifically, step S1 is that coding obtains initial population.
In one embodiment, step S1 is specifically included:
Step S1.1: multiple populations are directed to, corresponding cluster centre distribution situation is set;
Step S1.2: coding cluster centre distribution situation, and it will be used as decision variable, according to fitness calculation formula, meter
Corresponding cluster target and class object value are calculated, and using corresponding cluster target as cluster objective function, class object value
As class object function, wherein shown in the concrete form such as formula (1) and formula (2) of cluster objective function and class object function:
In formula (1), f (xi) label predicted of presentation class device, yiExpression is really label, δ (f (xi), yi) indicate pre-
Whether mark label are consistent with true tag, consistent results 1, inconsistent results 0, and N indicates test data total quantity;
In formula (2), nnijIndicate distance sample xiMinimum distance j-th of sample, p (ck|nnij) indicate j-th of sample
Belong to the conditional probability of k-th of cluster,It is a parameter, works as nnijAnd xiWhen belonging to cluster k, it is 0, is otherwise 1/j;
Step S1.3: decision variable, cluster objective function and class object function are encapsulated into population at individual simultaneously,
Multiple individual composition initialization populations.
Specifically, in step S1.1, corresponding cluster centre distribution situation can be set at random for multiple populations.Step
Rapid S1.2 is that cluster objective function is designed based on the method for silhouette coefficient using quick and low complex degree, after assisting
The building of continuous classifier.
Step S2: searching for optimal population using multi-objective optimization algorithm, clusters objective function as meeting and divides
The Pareto disaggregation of two targets of class objective function.
Specifically, step S2 is to find optimal classifier using a kind of knockdown search Multi-object searching algorithm
Interior relational matrix specifically obtains cluster centre after can clustering to training data, by cluster centre and known mark
Label value finds out relational matrix.
In one embodiment, step S2 is specifically included:
Step S2.1: it in former generation, is evolved using the crossover operator and mutation operator of difference algorithm to current population
Processing, obtains progeny population;
Step S2.2: in former generation, using backward learning algorithm, inverse change is carried out to progeny population, and will be reversed
The population of front and back carries out non-dominated ranking after merging, and obtains reversed sub- population;
Step S2.3: in former generation, using fireworks algorithm, calculating the burst radius and spark number of reversed sub- population,
And it is based on burst radius and spark number, to generate spark population in peripheral extent using reversed sub- population as fireworks population,
Fireworks population and spark population carry out non-dominated ranking and generate next-generation population;
Step S2.4: using next-generation population as former generation population is worked as, step S2.1~step S2.3 is repeated, is changed
For evolutionary process;
Step S2.5: when iterative evolution process meets termination condition, iterative evolution is terminated, and exports Pareto disaggregation.
Specifically, differential evolution algorithm (Differential Evolution, DE) is by Storn and Price in 1995
Year is put forward for the first time.It is mainly used for solving real number optimization problem.The algorithm is a kind of adaptive global optimization algorithm based on group,
One kind for belonging to evolution algorithmic, since it has the characteristics that structure is simple, easy to accomplish, convergence is quick, strong robustness.
In evolution algorithm, when generating the population of a new generation, in order to guarantee the newly diversity for population, calculated by backward learning
Method, so that the solution that population has certain small probability to select not degree of being accommodated preference.Thus, it is more preferably reversed obtained in step S2.2
Sub- population.
For the chromosome not dominated mutually, claiming these chromosomes to be in same layer, then all chromosome can be divided
It is exactly the sort algorithm for being layered chromosome to several layers non-dominated ranking, the layer got is known as the non-dominant layer of the first order, and second
Wherein the non-dominant layer of the first order is in the forward position Pareto (Pareto Front) to the non-dominant layer ... of grade.
Fireworks algorithm (Fireworks Algorithm), is abbreviated as FWA, be by the night sky fireworks explode inspiration and
A kind of Swarm Intelligence Algorithm proposed.Fireworks algorithm starts iteration, successively using explosion operator, mutation operator, mapping ruler and
Selection strategy, until reaching termination condition, that is, meet the required precision of problem or reaching maximal function assessment number.
The realization of fireworks algorithm includes following several steps:
1) some fireworks are randomly generated in specific solution space, each fireworks represents a solution of solution space.
2) fitness value of each fireworks is calculated according to fitness function, and spark is generated according to fitness value.Spark
Number calculated based on the thought of the immune concentration in immunology, i.e., the better fireworks of fitness value produce pyrophoric number more
It is more.
3) it according to the fireworks attribute in reality and the actual conditions of combination search problem, is generated in the radiation space of fireworks
Spark.(size of the explosion amplitude of some fireworks determines that fitness value is bigger by fitness value of the fireworks on function, quick-fried
Fried amplitude is bigger, and vice versa).Each spark represents a solution in solution space.In order to guarantee the diversity of population, need
It is suitably made a variation to fireworks, such as Gaussian mutation.
4) optimal solution for calculating population, determines whether to meet the requirements, and stops search if meeting, does not meet, continue
Iteration.The initial value of iteration is the other solutions for the best solution and selection that this circulation obtains.
Wherein, the termination condition in step S2.5 includes: that the number of iterations satisfaction reaches maximum number of iterations and population is whole
The variation of body is less than the minimum population variation of setting.
Step S3: optimal solution is selected from Pareto solution concentration according to mutual information index, and optimal according to what is selected
Solution, the parameter of classifier of joint Cluster Classification mechanism is configured, then the building Cluster Classification connection of the parameter based on setting
The classifier of conjunction mechanism.
Wherein, step S3 is specifically included:
Step S3.1: the association relationship of all Pareto disaggregation is calculated;
Step S3.2: according to the size of association relationship, the maximum Pareto disaggregation of association relationship is selected as optimal
Solution is configured the parameter of the classifier of joint Cluster Classification mechanism;
Step S3.3: the classifier of the parameter building Cluster Classification associative mechanism based on setting.
Specifically, Pareto optimal solution, also referred to as Pareto efficiency (Pareto efficiency) refer to resource point
A kind of perfect condition matched, it is assumed that intrinsic group and assignable resource, from a kind of distribution state to another state
In variation, under the premise of not making anyone circumstances degenerate, so that at least one people becomes more preferable.Pareto-optimality is just
It is the leeway that impossible have more Pareto improvements again.
Mutual information is a kind of useful measure information in information theory, it can regard the pass for including in a stochastic variable as
In the information content of another stochastic variable, or perhaps a stochastic variable is reduced not due to another known stochastic variable
Certainty.Wherein, calculate the formula of mutual information be it is known, the mutual information index in the embodiment of the present invention refers to prediction bid
Association relationship between the test data and truthful data of label, for this value closer to 1, classification accuracy is higher.
After the present invention obtains one group of Pareto disaggregation by evolution algorithmic, optimal solution is therefrom chosen by step S3, is pressed
It is ranked up according to association relationship size, parameter of the solution of maximum mutual information as classifier is selected, to obtain optimal classification
Device setting.
In one embodiment, after step S3.3, the method also includes step S3.4:
It is tested using classifier of the test data to building.
Specifically, in order to be further improved classifier, the present invention also passes through test data and tests corresponding data set feelings
Condition.It if test result does not meet preset condition, can must sort according to association relationship, successively choose the parameter met the requirements and arrive
It is tested in classifier.
Fig. 2 is referred to, for the schematic diagram of the classifier based on Cluster Classification associative mechanism in the present invention, when learning classifier
It, can be with the quality of testing classification device by test data by the available optimal parameter of training data after general framework.Its
Middle training process constructs the process of classifier.
Fig. 3 is the flow chart of the classifier of the Cluster Classification mechanism optimized based on fireworks algorithm and difference algorithm, is first
Data collection is carried out, the number of cluster centre is then assessed, two meta templates are set, is initialized after random value is set to individual
Population, upper right portion indicate to search for optimal population using multi-objective optimization algorithm (fireworks difference combinational algorithm).
Embodiment two
Based on same inventive concept, a kind of classification based on the classifier constructed in embodiment one is present embodiments provided
Method, the classification method include:
After the cluster centre and clustering relationships that calculate data to be processed, classification results are obtained by classifier.
The beneficial effect of the method provided in order to further illustrate the present invention, below by several specific examples to this hair
The classifier of bright building is compared for classifying with existing method, wherein table 1 is the present invention on 19 UCI data sets
Classification accuracy and the comparison with other algorithms, table 2 be the present invention 4 image segmentation standards test libraries mutual information index
And the comparison with other algorithms, MSCC-DE/FWA is the method for use of the invention, from Tables 1 and 2, it will be seen that, it is of the invention
The available preferable classifying quality of method.
Classification accuracy on 1 19 UCI data sets of table and the comparison with other algorithms
Table 24 image segmentation standards test libraries mutual information index and with the comparison of other algorithms
NMI | MASCC-DE/FWA | MOASCC | MSCC | SVM | RBFNN | semi-MOCK |
two categories image | 0.963 | 0.959 | 0.905 | 0.932 | 0.897 | 0.957 |
Three categories image | 0.971 | 0.973 | 0.914 | 0.892 | 0.953 | 0.878 |
four categories image | 0.977 | 0.982 | 0.924 | 0.947 | 0.862 | 0.94 |
two categories image | 0.981 | 0.974 | 0.932 | 0.921 | 0.903 | 0.917 |
In general, it is proposed that be a kind of automatic categorizer based on Cluster Classification associative mechanism.Firstly, proposing
A kind of knockdown search Multi-object searching algorithm finds the relational matrix in optimal classifier.Secondly, proposing a kind of fast
The cluster objective function based on silhouette coefficient design of speed while low complex degree is come subsidiary classification device.Finally, in Cluster Classification machine
Done 3 points of improvement again in system the inside: automatic cluster saves initialization time, and backward learning increases the effective range of search, mutual information
Index is as last judgment basis.The algorithm can be effectively by Cluster-Fusion into classification task the inside, thus the standard of classifier
Exactness and search efficiency.
Although preferred embodiments of the present invention have been described, it is created once a person skilled in the art knows basic
Property concept, then additional changes and modifications can be made to these embodiments.So it includes excellent that the following claims are intended to be interpreted as
It selects embodiment and falls into all change and modification of the scope of the invention.
Obviously, those skilled in the art can carry out various modification and variations without departing from this hair to the embodiment of the present invention
The spirit and scope of bright embodiment.In this way, if these modifications and variations of the embodiment of the present invention belong to the claims in the present invention
And its within the scope of equivalent technologies, then the present invention is also intended to include these modifications and variations.
Claims (7)
1. a kind of construction method of the classifier based on Cluster Classification associative mechanism characterized by comprising
Step S1: the distribution situation of cluster centre is encoded into population at individual, initialization population is obtained, wherein initialization kind
It include cluster objective function and class object function in group;
Step S2: searching for optimal population using multi-objective optimization algorithm, clusters objective function and classification mesh as meeting
The Pareto disaggregation of two targets of scalar functions;
Step S3: selecting optimal solution from Pareto solution concentration according to mutual information index, and according to the optimal solution selected,
The parameter of the classifier of joint Cluster Classification mechanism is configured, then the parameter based on setting constructs Cluster Classification associative mechanism
Classifier.
2. the method as described in claim 1, which is characterized in that step S1 is specifically included:
Step S1.1: multiple populations are directed to, corresponding cluster centre distribution situation is set;
Step S1.2: coding cluster centre distribution situation, and will be calculated as decision variable according to fitness calculation formula
Corresponding cluster target and class object value, and using corresponding cluster target as cluster objective function, class object value conduct
Class object function, wherein shown in the concrete form such as formula (1) and formula (2) of cluster objective function and class object function:
In formula (1), f (xi) label predicted of presentation class device, yiExpression is really label, δ (f (xi), yi) indicate pre- mark
, consistent results 1 whether consistent with true tag, inconsistent results 0 are signed, N indicates test data total quantity;
In formula (2), nnijIndicate distance sample xiMinimum distance j-th of sample, p (ck|nnij) indicate that j-th of sample belongs to
The conditional probability of k-th of cluster,It is a parameter, works as nnijAnd xiWhen belonging to cluster k, it is 0, is otherwise 1/i;
Step S1.3: decision variable, cluster objective function and class object function are encapsulated into population at individual simultaneously, multiple
Individual composition initialization population.
3. the method as described in claim 1, which is characterized in that step S2 is specifically included:
Step S2.1: in former generation, doing evolution processing to current population using the crossover operator and mutation operator of difference algorithm,
Obtain progeny population;
Step S2.2: in former generation, using backward learning algorithm, inverse change is carried out to progeny population, and will reversed front and back
Population merge after carry out non-dominated ranking, obtain reversed sub- population;
Step S2.3: in former generation, using fireworks algorithm, the burst radius and spark number of reversed sub- population, and base are calculated
In burst radius and spark number, to generate spark population, fireworks in peripheral extent using reversed sub- population as fireworks population
Population and spark population carry out non-dominated ranking and generate next-generation population;
Step S2.4: using next-generation population as work as former generation population, repeat step S2.1~step S2.3, be iterated into
Change process;
Step S2.5: when iterative evolution process meets termination condition, iterative evolution is terminated, and exports Pareto disaggregation.
4. method as claimed in claim 3, which is characterized in that the termination condition in step S2.5 includes: that the number of iterations meets
Reach maximum number of iterations and whole the changing of population is less than the minimum population variation of setting.
5. the method as described in claim 1, which is characterized in that step S3 is specifically included
Step S3.1: the association relationship of all Pareto disaggregation is calculated;
Step S3.2: according to the size of association relationship, selecting the maximum Pareto disaggregation of association relationship as optimal solution, right
The parameter of the classifier of joint Cluster Classification mechanism is configured;
Step S3.3: the classifier of the parameter building Cluster Classification associative mechanism based on setting.
6. method as claimed in claim 5, which is characterized in that after step S3.3, the method also includes step S3.4:
It is tested using classifier of the test data to building.
7. a kind of classification method based on classifier constructed by any one of claim 1 to 5 claim, which is characterized in that
The classification method includes:
After the cluster centre and clustering relationships that calculate data to be processed, classification results are obtained by classifier.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910652753.3A CN110490234A (en) | 2019-07-19 | 2019-07-19 | The construction method and classification method of classifier based on Cluster Classification associative mechanism |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910652753.3A CN110490234A (en) | 2019-07-19 | 2019-07-19 | The construction method and classification method of classifier based on Cluster Classification associative mechanism |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110490234A true CN110490234A (en) | 2019-11-22 |
Family
ID=68547451
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910652753.3A Pending CN110490234A (en) | 2019-07-19 | 2019-07-19 | The construction method and classification method of classifier based on Cluster Classification associative mechanism |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110490234A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111325284A (en) * | 2020-03-09 | 2020-06-23 | 武汉大学 | Self-adaptive learning method and device based on multi-target dynamic distribution |
CN112654081A (en) * | 2020-12-14 | 2021-04-13 | 西安邮电大学 | User clustering and resource allocation optimization method, system, medium, device and application |
CN114239742A (en) * | 2021-12-21 | 2022-03-25 | 中国人民解放军国防科技大学 | Medical data classification method based on rule classifier and related equipment |
-
2019
- 2019-07-19 CN CN201910652753.3A patent/CN110490234A/en active Pending
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111325284A (en) * | 2020-03-09 | 2020-06-23 | 武汉大学 | Self-adaptive learning method and device based on multi-target dynamic distribution |
CN112654081A (en) * | 2020-12-14 | 2021-04-13 | 西安邮电大学 | User clustering and resource allocation optimization method, system, medium, device and application |
CN112654081B (en) * | 2020-12-14 | 2023-02-07 | 西安邮电大学 | User clustering and resource allocation optimization method, system, medium, device and application |
CN114239742A (en) * | 2021-12-21 | 2022-03-25 | 中国人民解放军国防科技大学 | Medical data classification method based on rule classifier and related equipment |
CN114239742B (en) * | 2021-12-21 | 2024-04-02 | 中国人民解放军国防科技大学 | Medical data classification method based on rule classifier and related equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111191732B (en) | Target detection method based on full-automatic learning | |
CN105589806B (en) | A kind of software defect tendency Forecasting Methodology based on SMOTE+Boosting algorithms | |
Kang et al. | A weight-incorporated similarity-based clustering ensemble method based on swarm intelligence | |
Kuo et al. | Integration of particle swarm optimization and genetic algorithm for dynamic clustering | |
CN103164713B (en) | Image classification method and device | |
CN110363344A (en) | Probability integral parameter prediction method based on MIV-GP algorithm optimization BP neural network | |
CN110163258A (en) | A kind of zero sample learning method and system reassigning mechanism based on semantic attribute attention | |
CN108363810A (en) | Text classification method and device | |
CN107683469A (en) | A kind of product classification method and device based on deep learning | |
CN108897791B (en) | Image retrieval method based on depth convolution characteristics and semantic similarity measurement | |
CN107292097B (en) | Chinese medicine principal symptom selection method based on feature group | |
CN110490234A (en) | The construction method and classification method of classifier based on Cluster Classification associative mechanism | |
Gu et al. | An improved SMOTE algorithm based on genetic algorithm for imbalanced data classification | |
CN110991653A (en) | Method for classifying unbalanced data sets | |
CN111931505A (en) | Cross-language entity alignment method based on subgraph embedding | |
CN105046323B (en) | Regularization-based RBF network multi-label classification method | |
CN110287985B (en) | Depth neural network image identification method based on variable topology structure with variation particle swarm optimization | |
CN110147808A (en) | A kind of novel battery screening technique in groups | |
Zhou et al. | ECMdd: Evidential c-medoids clustering with multiple prototypes | |
CN113269647A (en) | Graph-based transaction abnormity associated user detection method | |
CN107451617B (en) | Graph transduction semi-supervised classification method | |
CN113255873A (en) | Clustering longicorn herd optimization method, system, computer equipment and storage medium | |
CN114556364A (en) | Neural architecture search based on similarity operator ordering | |
CN108737429B (en) | Network intrusion detection method | |
Liu et al. | A weight-incorporated similarity-based clustering ensemble method |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191122 |