CN109829473A - The degree of membership for equipping services platform user classification for intelligence determines method and system - Google Patents
The degree of membership for equipping services platform user classification for intelligence determines method and system Download PDFInfo
- Publication number
- CN109829473A CN109829473A CN201811603637.4A CN201811603637A CN109829473A CN 109829473 A CN109829473 A CN 109829473A CN 201811603637 A CN201811603637 A CN 201811603637A CN 109829473 A CN109829473 A CN 109829473A
- Authority
- CN
- China
- Prior art keywords
- particle
- population
- indicate
- data
- lateral cross
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 65
- 239000002245 particle Substances 0.000 claims description 180
- 239000011159 matrix material Substances 0.000 claims description 108
- 230000003044 adaptive effect Effects 0.000 claims description 71
- 238000010606 normalization Methods 0.000 claims description 9
- 238000013213 extrapolation Methods 0.000 claims description 5
- 238000012790 confirmation Methods 0.000 claims 1
- 238000012545 processing Methods 0.000 abstract description 3
- 238000007796 conventional method Methods 0.000 abstract 1
- 239000008186 active pharmaceutical agent Substances 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003032 molecular docking Methods 0.000 description 1
- 230000003472 neutralizing effect Effects 0.000 description 1
- 238000011017 operating method Methods 0.000 description 1
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a kind of degrees of membership for equipping services platform user classification for intelligence to determine method and system, propose the Fuzzy C-Means Cluster Algorithm (CSO-FCM) based on crossover algorithm in length and breadth, this method is suitable for complicated multiple features, various dimensions, nonlinear problem, it not only overcomes the shortcomings that conventional method, also remain FCM processing uncertain information ability, CSO is simple and easy and has the characteristics that stronger global convergence type.Known FCM is by determining that each sample to the degree of membership of cluster centre, to realize the classification to sample, calculates the user attribute data of each user and the degree of membership of cluster centre, improves the accuracy rate and stability of determining degree of membership.
Description
Technical field
The present invention relates to technical field of data processing, equip services platform user for intelligence more particularly, to one kind
The degree of membership of classification determines method and system.
Background technique
Intelligence equipment service platform be serve local government, section consults wall telephone set structure and the big data industry development of enterprise is flat
Platform can provide Information Navigation Service and direction of industry service for each mechanism, user, and meet its supply and demand docking, cooperation and exchange, money
News push, industry know clearly, Techno-sharing, talent introduction, expert consulting, the demands such as invite outside investment.
When carrying out user's classification to platform, generally require to handle user attribute data.User attribute data is
Reflect a kind of achievement data of user characteristics, which includes mainly enterprise nature, is expert in the registration information of user
Industry, location/city, product applications, user's position/job specification etc..In view of the deficiency of current many clustering algorithms,
Such as Fuzzy C-Means Cluster Algorithm (FCM) although simple, quickly and be widely used, its cluster result is unstable and easy falls into
Enter local optimum;Fuzzy C-Means Cluster Algorithm (PSO-FCM) based on particle group optimizing is compared with FCM, although its global search
Ability increases, but optimal result is still unstable;Although simplifying coding PSO cluster eliminates sample vector dimension to PSO
Influence, but operating procedure complexity is cumbersome, needs the parameter that controls excessive.
Summary of the invention
The present invention is directed to solve above-mentioned technical problem at least to a certain extent.
Primary and foremost purpose of the invention is to provide a kind of degree of membership determination side that services platform user classification is equipped for intelligence
Method, proposes the Fuzzy C-Means Cluster Algorithm (CSO-FCM) based on crossover algorithm in length and breadth, and this method is suitable for complicated mostly spy
The shortcomings that sign, various dimensions, nonlinear problem, it not only overcomes the above method, also remains the energy of FCM processing uncertain information
Power, CSO are simple and easy and have the characteristics that stronger global convergence type.Known FCM is by determining each sample to the person in servitude of cluster centre
Category degree, to realize the classification to sample.
It is determining that the further object of the present invention is to provide a kind of degree of membership for equipping services platform user classification for intelligence
System.
In order to solve the above technical problems, technical scheme is as follows:
It is a kind of to determine method for the intelligent degree of membership for equipping services platform user classification, comprising the following steps:
S1: the user attribute data including number of users, user's registration information, and root are obtained from intelligence equipment service platform
Raw data matrix is established according to the user attribute data;
S2: normalizing the raw data matrix, obtains normalized matrix;
S3: determining initial cluster center, carries out particle coding to initial cluster center based on the coding mode of cluster centre,
Generate initial population;
Initial population: being set as the parent population of lateral cross by S4, calculates the laterally friendship according to the normalized matrix
The adaptive value of each particle in the parent population of fork;
S5: according to the adaptive value of the parent population of lateral cross and its particle, calculating the first population using lateral cross method,
First population is set as crossed longitudinally parent population, the crossed longitudinally parent population is calculated according to the normalized matrix
In each particle adaptive value, the adaptive value of its particle is neutralized according to crossed longitudinally parent population, using crossed longitudinally method meter
Calculate the second population;
S6: judging whether the number of iterations is greater than setting the number of iterations, if it is not, the second population to be then set as to the father of lateral cross
For population, the adaptive value of each particle in the parent population of the lateral cross is calculated according to the normalized matrix, using transverse direction
Interior extrapolation method and crossed longitudinally method, which are iterated, calculates the second population;If so, the second population is split as c cluster centre, and count
Calculate the attribute data of each user and the degree of membership of cluster centre, wherein c indicates preset clusters number.
Preferably, the raw data matrix is normalized in step S2, obtains normalized matrix, comprising the following steps:
S2.1: normalization data is calculated:
In formula, a ∈ (1, n), n indicate that number of users, b ∈ (1, d), d indicate the corresponding user attribute data of each user
Quantity, minx*bIndicate in raw data matrix minimum user attribute data, maxx in b column*bIndicate raw data matrix
In maximum user attribute data in b column, x'abIndicate the corresponding normalization number of user attribute data of a-th of user, b column
According to;
S2.2: normalized matrix is obtained according to normalization data.
Preferably, initial population is set as to the parent population of lateral cross, according to the normalized matrix meter in step S4
Calculate the adaptive value of each particle in the parent population of the lateral cross, comprising the following steps:
S4.1: initial population is set as to the parent population of lateral cross;
S4.2: each particle in the parent population of lateral cross is split as c cluster centre respectively, obtains each particle
Corresponding cluster centre;
S4.3: the person in servitude between the corresponding user attribute data of each user in the normalized matrix and cluster centre is calculated
Category degree:
In formula, 1≤x≤c, 1≤y≤n, n indicate that number of users, m indicate fuzzy coefficient, dxyIndicate the normalized matrix
In Euclidean distance between y-th of user attribute data and x-th of cluster centre, dkyIt indicates in the normalized matrix y-th
Euclidean distance between the corresponding user attribute data of user and k-th of cluster centre, uxyIndicate every in the normalized matrix
Degree of membership between the corresponding user attribute data of a user y and cluster centre x;
S4.4: the adaptive value of each particle in the parent population of lateral cross is calculated:
In formula, fit indicates the adaptive value of the corresponding particle of cluster centre.
Preferably, according to the adaptive value of the parent population of lateral cross and its particle in step S5, using lateral cross method
The first population is calculated, the first population is set as crossed longitudinally parent population, the longitudinal direction is calculated according to the normalized matrix
The adaptive value of each particle in the parent population of intersection, comprising the following steps:
S5.1: the particle in the parent population of lateral cross is subjected to combination of two, obtains particle pair, and be numbered;
S5.2: successively taking out particle pair in numerical order, and following formula pair is used under the lateral cross probability of setting
The d of particle pair0Dimension executes horizontal line and intersects:
MShc(i,d0)=r1·X(i,d0)+(1-r1)·X(j,d0)+c1·(X(i,d0)-X(j,d0))
MShc(j,d0)=r1·X(j,d0)+(1-r2)·X(i,d0)+c1·(X(j,d0)-X(i,d0))
Wherein, d0∈ (1, D), D=c × d, d indicate the quantity of the corresponding user attribute data of each user;i∈(1,
N), (1, N) j ∈, N indicate initial population particle number;r1、r2For the uniform random number on [0,1], c1、c2For [- 1,1]
Between uniform random number;X(i,d0) indicate i-th of particle d0Dimension data;X(j,d0) indicate the of j-th particle
d0Dimension data;MShc(i,d0) and MShc(j,d0) indicate the golden mean of the Confucian school solution after lateral cross;
S5.3: each particle is stored in the first matrix corresponding golden mean of the Confucian school solution, calculates each in first matrix
The adaptive value of golden mean of the Confucian school solution, and adaptive value corresponding with particle each in the parent population of lateral cross is compared, by fitness
Big particle is stored in the first population, and the first population is set as crossed longitudinally parent population.
Preferably, the adaptive value of its particle is neutralized in step S5 according to crossed longitudinally parent population, use is crossed longitudinally
Method calculates the second population, comprising the following steps:
S5.4: by being normalized per one-dimensional for crossed longitudinally parent population, carrying out combination of two for each dimension, and
It is numbered;
S5.5: every a pair is successively taken out in numerical order;
S5.6: crossed longitudinally to each pair of data execution using following formula according to crossed longitudinally rate is set:
MSvc(i,d1)=rX (i, d1)+(1-r)·X(i,d2)
In formula, X (i, d1) indicate i-th of particle d1Dimension data, X (i, d2) indicate i-th of particle d2Dimension
According to d1,d2∈ (1, D), d1≠d2, D=c × d, d indicate the quantity of the corresponding user attribute data of each user;i∈(1,N);
R is the uniform random number on [0,1];MSvc(i,d1) indicate i-th of particle d1Tie up offspring data;
S5.7: each particle is respectively tieed up into offspring data and carries out renormalization, obtains corresponding golden mean of the Confucian school solution, and save it in
In second matrix, calculate the adaptive value of each golden mean of the Confucian school solution in second matrix, and with it is each in crossed longitudinally parent population
The adaptive value of particle is compared, and the big particle of fitness is stored in the second population, wherein crossed longitudinally parent population
The first population obtained for lateral cross method.
Preferably, being normalized per one-dimensional by crossed longitudinally parent population in step S5.4, use is following
Formula is normalized crossed longitudinally parent population per one-dimensional:
In formula, d1∈ (1, D), Pd1maxFor d1The upper limit of dimension control variable, Pd1minFor d1The lower limit of dimension control variable,
K is current iteration number, Xk(i,d1) indicate the number of iterations be k when i-th of particle d1Dimension data, Xk-1(i,d1) indicate to change
Generation number is the d of i-th of particle of k-11Dimension data.
Preferably, each particle is respectively tieed up into offspring data in the step S5.7 and carries out renormalization, using following formula
Each particle is respectively tieed up into offspring data and carries out renormalization:
MSvc'(i,d1)=MSvc(i,d1)·(Pd1max-Pd1min)+Pd1min
Wherein, MSvc(i,d1) indicate i-th of particle d1Tie up offspring data, MSvc'(i,d1) indicate i-th of particle
D1Data after tieing up offspring data renormalization.
It is a kind of to determine system for the intelligent degree of membership for equipping services platform user classification, comprising:
Raw data matrix establishes module, includes number of users, user's registration for obtaining from intelligence equipment service platform
The user attribute data of information, and raw data matrix is established according to the user attribute data;
Normalized module normalizes the raw data matrix, obtains normalized matrix;
Initial population generation module, for initial population to be set as to the parent population of lateral cross, according to the standardization
Matrix calculates the adaptive value of each particle in the parent population of the lateral cross;
Degree of membership determining module, for the adaptive value according to the parent population of lateral cross and its particle, using lateral friendship
Fork method calculates the first population, and the first population is set as crossed longitudinally parent population, according to normalized matrix calculating
The adaptive value of each particle in crossed longitudinally parent population, the adaptation of its particle is neutralized according to crossed longitudinally parent population
Value calculates the second population using crossed longitudinally method;Judge whether the number of iterations is greater than setting the number of iterations, if it is not, then by second
Population is set as the parent population of lateral cross, calculates each grain in the parent population of the lateral cross according to the normalized matrix
The adaptive value of son is iterated using lateral cross method and crossed longitudinally method and calculates the second population;If so, the second population is split
For c cluster centre, and calculate the attribute data of each user and the degree of membership of cluster centre, wherein c indicates preset cluster
Number.
Preferably, the degree of membership determining module, is used for:
Particle in the parent population of lateral cross is subjected to combination of two, obtains particle pair, and be numbered;
Particle pair is successively taken out in numerical order, and uses following formula to particle pair under the lateral cross probability of setting
D0Dimension executes horizontal line and intersects:
MShc(i,d0)=r1·X(i,d0)+(1-r1)·X(j,d0)+c1·(X(i,d0)-X(j,d0))
MShc(j,d0)=r1·X(j,d0)+(1-r2)·X(i,d0)+c1·(X(j,d0)-X(i,d0))
Wherein, d0∈ (1, D), D=c × d, d indicate the quantity of the corresponding user attribute data of each user;i∈(1,
N), (1, N) j ∈, N indicate initial population particle number;r1、r2For the uniform random number on [0,1], c1、c2For [- 1,1]
Between uniform random number;X(i,d0) indicate i-th of particle d0Dimension data;X(j,d0) indicate the of j-th particle
d0Dimension data;MShc(i,d0) and MShc(j,d0) indicate the golden mean of the Confucian school solution after lateral cross;
Each particle is stored in the first matrix corresponding golden mean of the Confucian school solution, calculates each golden mean of the Confucian school solution in first matrix
Adaptive value, and adaptive value corresponding with particle each in the parent population of lateral cross is compared, by the big grain of fitness
Son is stored in the first population, and the first population is set as crossed longitudinally parent population.
Preferably, the degree of membership determining module, is used for:
By being normalized per one-dimensional for crossed longitudinally parent population, each dimension is subjected to combination of two, and carry out
Number;
Every a pair is successively taken out in numerical order;
It is crossed longitudinally to each pair of data execution using following formula according to crossed longitudinally rate is set:
MSvc(i,d1)=rX (i, d1)+(1-r)·X(i,d2)
In formula, X (i, d1) indicate i-th of particle d1Dimension data, X (i, d2) indicate i-th of particle d2Dimension
According to d1,d2∈ (1, D), d1≠d2, D=c × d, d indicate the quantity of the corresponding user attribute data of each user;i∈(1,N);
R is the uniform random number on [0,1];MSvc(i,d1) indicate i-th of particle d1Tie up offspring data;
Each particle is respectively tieed up into offspring data and carries out renormalization, obtains corresponding golden mean of the Confucian school solution, and save it in second
In matrix, calculate the adaptive value of each golden mean of the Confucian school solution in second matrix, and with each particle in crossed longitudinally parent population
Adaptive value be compared, the big particle of fitness is stored in the second population, wherein crossed longitudinally parent population be cross
The first population obtained to interior extrapolation method.
Compared with prior art, the beneficial effect of technical solution of the present invention is:
It by establishing raw data matrix, and is normalized, subsequent to handle normalized matrix, Ke Yiti
High treatment efficiency.Particle coding is carried out to initial cluster center based on the coding mode of cluster centre, generates initial population.?
When an iteration, initial population is set as to the parent population of lateral cross, the laterally friendship is calculated according to the normalized matrix
The adaptive value of each particle in the parent population of fork, and the second population is calculated using lateral cross method and crossed longitudinally method, do not having
Have reach setting the number of iterations when, the second population is set as to the parent population of lateral cross, according to the normalized matrix calculate
The adaptive value of each particle in the parent population of the lateral cross, is iterated calculating using lateral cross method and crossed longitudinally method
Second population, until reaching setting the number of iterations, the second population is split as c by the second optimal population available at this time
Cluster centre to obtain optimal cluster centre, and calculates the user attribute data of each user and being subordinate to for cluster centre
Degree, improves the accuracy rate and stability of determining degree of membership.
Detailed description of the invention
Fig. 1 determines method flow diagram for a kind of degree of membership for equipping services platform user classification for intelligence.
Fig. 2 determines system schematic for a kind of degree of membership for equipping services platform user classification for intelligence.
Specific embodiment
The attached figures are only used for illustrative purposes and cannot be understood as limitating the patent;
In order to better illustrate this embodiment, the certain components of attached drawing have omission, zoom in or out, and do not represent actual product
Size;
To those skilled in the art, it is to be understood that certain known features and its explanation, which may be omitted, in attached drawing
's.
The following further describes the technical solution of the present invention with reference to the accompanying drawings and examples.
Embodiment 1
The present embodiment provides a kind of degrees of membership for equipping services platform user classification for intelligence to determine method, such as Fig. 1, wraps
Include following steps:
S1: the user attribute data including number of users, user's registration information, and root are obtained from intelligence equipment service platform
Raw data matrix is established according to the user attribute data;
User attribute data is a kind of achievement data for reflecting user characteristics.It can establish original matrix X=(xab)n×d。a
∈ (1, n), n indicate intelligence equipment services platform user quantity in total, and b ∈ (1, d), d indicate the corresponding user of each user
The quantity of attribute data, xabIndicate the user attribute data of a-th of user, b column.
S2: normalizing the raw data matrix, obtains normalized matrix;
S3: determining initial cluster center, carries out particle coding to initial cluster center based on the coding mode of cluster centre,
Initial population is generated, each particle is made of c cluster centre, and c is the classification number of data set.Normalized matrix vector dimension
For d, then the dimension of particle is c × d, the coding structure of particle v are as follows:
Initial population: being set as the parent population of lateral cross by S4, calculates the laterally friendship according to the normalized matrix
The adaptive value of each particle in the parent population of fork;
S5: according to the adaptive value of the parent population of lateral cross and its particle, calculating the first population using lateral cross method,
First population is set as crossed longitudinally parent population, the crossed longitudinally parent population is calculated according to the normalized matrix
In each particle adaptive value, the adaptive value of its particle is neutralized according to crossed longitudinally parent population, using crossed longitudinally method meter
Calculate the second population;
S6: judging whether the number of iterations is greater than setting the number of iterations, if it is not, the second population to be then set as to the father of lateral cross
For population, the adaptive value of each particle in the parent population of the lateral cross is calculated according to the normalized matrix, using transverse direction
Interior extrapolation method and crossed longitudinally method, which are iterated, calculates the second population;If so, the second population is split as c cluster centre, and count
Calculate the attribute data of each user and the degree of membership of cluster centre, wherein c indicates preset clusters number.
It is Optimal cluster centers by the cluster centre that the second population splits.Pass through each use in normalized matrix
Degree of membership between the corresponding user attribute data in family and cluster centre, each user is corresponding in available raw data matrix
User attribute data and cluster centre between degree of membership, degree of membership can be greater than to the corresponding user property of setting degree of membership
Data are divided into one kind, and then realize the classification to user.
The raw data matrix is normalized in step S2, obtains normalized matrix, comprising the following steps:
S2.1: normalization data is calculated:
In formula, a ∈ (1, n), n indicate that number of users, b ∈ (1, d), d indicate the corresponding user attribute data of each user
Quantity, minx*bIndicate in raw data matrix minimum user attribute data, maxx in b column*bIndicate raw data matrix
In maximum user attribute data in b column, x'abIndicate the corresponding normalization number of user attribute data of a-th of user, b column
According to;
S2.2: normalized matrix is obtained according to normalization data, with X'=(x'ij)n×dIndicate normalized matrix.
Initial population is set as to the parent population of lateral cross in step S4, the cross is calculated according to the normalized matrix
The adaptive value of each particle into the parent population of intersection, comprising the following steps:
S4.1: initial population is set as to the parent population of lateral cross;
S4.2: each particle in the parent population of lateral cross is split as c cluster centre respectively, obtains each particle
Corresponding cluster centre;
S4.3: the person in servitude between the corresponding user attribute data of each user in the normalized matrix and cluster centre is calculated
Category degree:
In formula, 1≤x≤c, 1≤y≤n, n indicate that number of users, m indicate fuzzy coefficient, dxyIndicate the normalized matrix
In Euclidean distance between y-th of user attribute data and x-th of cluster centre, dkyIt indicates in the normalized matrix y-th
Euclidean distance between the corresponding user attribute data of user and k-th of cluster centre, uxyIndicate every in the normalized matrix
Degree of membership between the corresponding user attribute data of a user y and cluster centre x;
S4.4: the adaptive value of each particle in the parent population of lateral cross is calculated:
In formula, fit indicates the adaptive value of the corresponding particle of cluster centre.
According to the adaptive value of the parent population of lateral cross and its particle in step S5, first is calculated using lateral cross method
Population, is set as crossed longitudinally parent population for the first population, calculates the crossed longitudinally father according to the normalized matrix
For the adaptive value of particle each in population, comprising the following steps:
S5.1: the particle in the parent population of lateral cross is subjected to combination of two, obtains particle pair, and be numbered;
The step is that particle individuals all in population are carried out to two neither to repeat to match, and shares N/2 pairs, and be numbered,
Purpose is subsequent can successively to take out.
In first time iteration, the parent population of lateral cross is initial population;When pth time iteration, the father of lateral cross
The second population for population to be obtained when -1 iteration of pth.Second population is properly termed as the crossed longitudinally solution that is dominant, Ke Yiyong again
DSvcIt indicates.
S5.2: successively taking out particle pair in numerical order, and following formula pair is used under the lateral cross probability of setting
The d of particle pair0Dimension executes horizontal line and intersects:
MShc(i,d0)=r1·X(i,d0)+(1-r1)·X(j,d0)+c1·(X(i,d0)-X(j,d0))
MShc(j,d0)=r1·X(j,d0)+(1-r2)·X(i,d0)+c1·(X(j,d0)-X(i,d0))
Wherein, d0∈ (1, D), D=c × d, d indicate the quantity of the corresponding user attribute data of each user;i∈(1,
N), (1, N) j ∈, N indicate initial population particle number;r1、r2For the uniform random number on [0,1], c1、c2For [- 1,1]
Between uniform random number;X(i,d0) indicate i-th of particle d0Dimension data;X(j,d0) indicate the of j-th particle
d0Dimension data;MShc(i,d0) and MShc(j,d0) indicate the golden mean of the Confucian school solution after lateral cross;It repeats the above steps, the available grain
Son is to the corresponding golden mean of the Confucian school solution of each dimension.Same treatment is carried out to other particles again, obtains each particle to corresponding golden mean of the Confucian school solution.
S5.3: each particle is stored in the first matrix corresponding golden mean of the Confucian school solution, calculates each in first matrix
The adaptive value of golden mean of the Confucian school solution, and adaptive value corresponding with particle each in the parent population of lateral cross is compared, by fitness
Big particle is stored in the first population, and the first population is set as crossed longitudinally parent population.
The adaptive value for neutralizing its particle in step S5 according to crossed longitudinally parent population calculates the using crossed longitudinally method
Two populations, comprising the following steps:
S5.4: by being normalized per one-dimensional for crossed longitudinally parent population, carrying out combination of two for each dimension, and
It is numbered;Being dominant for lateral cross solves DS during crossed longitudinally parent population, that is, current iterationhc, this step is in population
All dimensions carry out two and neither repeat pairing (shared N/2 to).
S5.5: every a pair is successively taken out in numerical order;
S5.6: crossed longitudinally to each pair of data execution using following formula according to crossed longitudinally rate is set:
MSvc(i,d1)=rX (i, d1)+(1-r)·X(i,d2)
In formula, X (i, d1) indicate i-th of particle d1Dimension data, X (i, d2) indicate i-th of particle d2Dimension
According to d1,d2∈ (1, D), d1≠d2, D=c × d, d indicate the quantity of the corresponding user attribute data of each user;i∈(1,N);
R is the uniform random number on [0,1];MSvc(i,d1) indicate i-th of particle d1Tie up offspring data;
S5.7: each particle is respectively tieed up into offspring data and carries out renormalization, obtains corresponding golden mean of the Confucian school solution, and save it in
In second matrix, calculate the adaptive value of each golden mean of the Confucian school solution in second matrix, and with it is each in crossed longitudinally parent population
The adaptive value of particle is compared, and the big particle of fitness is stored in the second population, wherein crossed longitudinally parent population
The first population obtained for lateral cross method;
Each iterative process can all update the first population and the second population, keep finally obtained second population optimal.Whole
Before a scheme, clusters number c, fuzzy coefficient m, Population Size N can also be set, is controlled variable number D (i.e. dimension c × d), it is maximum
The number of iterations, crossed longitudinally rate pvc, lateral cross rate phc;
Crossed longitudinally parent population being normalized per one-dimensional in S5.4, using following formula by longitudinal friendship
The parent population of fork is normalized per one-dimensional:
In formula, d1∈ (1, D), Pd1maxFor d1The upper limit of dimension control variable, Pd1minFor d1The lower limit of dimension control variable,
K is current iteration number, Xk(i,d1) indicate the number of iterations be k when i-th of particle d1Dimension data, Xk-1(i,d1) indicate to change
Generation number is the d of i-th of particle of k-11Dimension data.
Each particle is respectively tieed up into offspring data in step S5.7 and carries out renormalization, it is using following formula that each particle is each
It ties up offspring data and carries out renormalization:
MSvc'(i,d1)=MSvc(i,d1)·(Pd1max-Pd1min)+Pd1min
Wherein, MSvc(i,d1) indicate i-th of particle d1Tie up offspring data, MSvc'(i,d1) indicate i-th of particle
D1Data after tieing up offspring data renormalization.
The present embodiment is normalized by establishing raw data matrix, it is subsequent will be at normalized matrix
Reason, can be improved treatment effeciency.Particle coding is carried out to initial cluster center based on the coding mode of cluster centre, is generated initial
Population.In first time iteration, initial population is set as to the parent population of lateral cross, institute is calculated according to the normalized matrix
The adaptive value of each particle in the parent population of lateral cross is stated, and is calculated second using lateral cross method and crossed longitudinally method
Second population is set as the parent population of lateral cross, according to the standardization square when not reaching setting the number of iterations by group
Battle array calculates the adaptive value of each particle in the parent population of the lateral cross, is changed using lateral cross method and crossed longitudinally method
In generation, calculates the second population, and until reaching setting the number of iterations, the second optimal population available at this time splits the second population
For a cluster centre, to obtain optimal cluster centre, and the user attribute data and cluster centre of each user is calculated
Degree of membership improves the accuracy rate and stability of determining degree of membership.
Embodiment 2
Method, this implementation are determined based on the degree of membership for equipping services platform user classification for intelligence that embodiment 1 provides
Example provide it is a kind of for intelligence equip services platform user classification degree of membership determine system, such as Fig. 2, including raw data matrix
Module is established, for obtaining the user attribute data including number of users, user's registration information from intelligence equipment service platform, and
Raw data matrix is established according to the user attribute data;
Normalized module normalizes the raw data matrix, obtains normalized matrix;
Initial population generation module, for initial population to be set as to the parent population of lateral cross, according to the standardization
Matrix calculates the adaptive value of each particle in the parent population of the lateral cross;
Degree of membership determining module, for the adaptive value according to the parent population of lateral cross and its particle, using lateral friendship
Fork method calculates the first population, and the first population is set as crossed longitudinally parent population, according to normalized matrix calculating
The adaptive value of each particle in crossed longitudinally parent population, the adaptation of its particle is neutralized according to crossed longitudinally parent population
Value calculates the second population using crossed longitudinally method;Judge whether the number of iterations is greater than setting the number of iterations, if it is not, then by second
Population is set as the parent population of lateral cross, calculates each grain in the parent population of the lateral cross according to the normalized matrix
The adaptive value of son is iterated using lateral cross method and crossed longitudinally method and calculates the second population;If so, the second population is split
For c cluster centre, and calculate the attribute data of each user and the degree of membership of cluster centre, wherein c indicates preset cluster
Number.
Degree of membership determining module, is used for:
Particle in the parent population of lateral cross is subjected to combination of two, obtains particle pair, and be numbered;
Particle pair is successively taken out in numerical order, and uses following formula to particle pair under the lateral cross probability of setting
D0Dimension executes horizontal line and intersects:
MShc(i,d0)=r1·X(i,d0)+(1-r1)·X(j,d0)+c1·(X(i,d0)-X(j,d0))
MShc(j,d0)=r1·X(j,d0)+(1-r2)·X(i,d0)+c1·(X(j,d0)-X(i,d0))
Wherein, d0∈ (1, D), D=c × d, d indicate the quantity of the corresponding user attribute data of each user;i∈(1,
N), (1, N) j ∈, N indicate initial population particle number;r1、r2For the uniform random number on [0,1], c1、c2For [- 1,1]
Between uniform random number;X(i,d0) indicate i-th of particle d0Dimension data;X(j,d0) indicate the of j-th particle
d0Dimension data;MShc(i,d0) and MShc(j,d0) indicate the golden mean of the Confucian school solution after lateral cross;
Each particle is stored in the first matrix corresponding golden mean of the Confucian school solution, calculates each golden mean of the Confucian school solution in first matrix
Adaptive value, and adaptive value corresponding with particle each in the parent population of lateral cross is compared, by the big grain of fitness
Son is stored in the first population, and the first population is set as crossed longitudinally parent population.
Degree of membership determining module, is used for:
By being normalized per one-dimensional for crossed longitudinally parent population, each dimension is subjected to combination of two, and carry out
Number;
Every a pair is successively taken out in numerical order;
It is crossed longitudinally to each pair of data execution using following formula according to crossed longitudinally rate is set:
MSvc(i,d1)=rX (i, d1)+(1-r)·X(i,d2)
In formula, X (i, d1) indicate i-th of particle d1Dimension data, X (i, d2) indicate i-th of particle d2Dimension
According to d1,d2∈ (1, D), d1≠d2, D=c × d, d indicate the quantity of the corresponding user attribute data of each user;i∈(1,N);
R is the uniform random number on [0,1];MSvc(i,d1) indicate i-th of particle d1Tie up offspring data;
Each particle is respectively tieed up into offspring data and carries out renormalization, obtains corresponding golden mean of the Confucian school solution, and save it in second
In matrix, calculate the adaptive value of each golden mean of the Confucian school solution in second matrix, and with each particle in crossed longitudinally parent population
Adaptive value be compared, the big particle of fitness is stored in the second population, wherein crossed longitudinally parent population be cross
The first population obtained to interior extrapolation method.
The same or similar label correspond to the same or similar components;
The terms describing the positional relationship in the drawings are only for illustration, should not be understood as the limitation to this patent;
Obviously, the above embodiment of the present invention be only to clearly illustrate example of the present invention, and not be pair
The restriction of embodiments of the present invention.For those of ordinary skill in the art, may be used also on the basis of the above description
To make other variations or changes in different ways.There is no necessity and possibility to exhaust all the enbodiments.It is all this
Made any modifications, equivalent replacements, and improvements etc., should be included in the claims in the present invention within the spirit and principle of invention
Protection scope within.
Claims (10)
1. a kind of degree of membership for equipping services platform user classification for intelligence determines method, which is characterized in that including following step
It is rapid:
S1: the user attribute data including number of users, user's registration information is obtained from intelligence equipment service platform, and according to institute
It states user attribute data and establishes raw data matrix;
S2: normalizing the raw data matrix, obtains normalized matrix;
S3: determining initial cluster center, carries out particle coding to initial cluster center based on the coding mode of cluster centre, generates
Initial population;
Initial population: being set as the parent population of lateral cross by S4, calculates the lateral cross according to the normalized matrix
The adaptive value of each particle in parent population;
S5: according to the adaptive value of the parent population of lateral cross and its particle, calculating the first population using lateral cross method, by the
One population is set as crossed longitudinally parent population, is calculated according to the normalized matrix every in the crossed longitudinally parent population
The adaptive value of a particle neutralizes the adaptive value of its particle according to crossed longitudinally parent population, calculates the using crossed longitudinally method
Two populations;
S6: judging whether the number of iterations is greater than setting the number of iterations, if it is not, the second population to be then set as to the parent kind of lateral cross
Group, the adaptive value of each particle in the parent population of the lateral cross is calculated according to the normalized matrix, using lateral cross
Method and crossed longitudinally method, which are iterated, calculates the second population;If so, the second population is split as c cluster centre, and calculate every
The attribute data of a user and the degree of membership of cluster centre, wherein c indicates preset clusters number.
2. the degree of membership according to claim 1 for equipping services platform user classification for intelligence determines method, feature
It is, the raw data matrix is normalized in step S2, obtains normalized matrix, comprising the following steps:
S2.1: normalization data is calculated:
In formula, a ∈ (1, n), n indicate number of users, and b ∈ (1, d), d indicate the number of the corresponding user attribute data of each user
Amount, minx*bIndicate in raw data matrix minimum user attribute data, maxx in b column*bIndicate b in raw data matrix
Maximum user attribute data, x' in columnabIndicate the corresponding normalization data of user attribute data of a-th of user, b column;
S2.2: normalized matrix is obtained according to normalization data.
3. the degree of membership according to claim 2 for equipping services platform user classification for intelligence determines method, feature
It is, initial population is set as to the parent population of lateral cross in step S4, the transverse direction is calculated according to the normalized matrix
The adaptive value of each particle in the parent population of intersection, comprising the following steps:
S4.1: initial population is set as to the parent population of lateral cross;
S4.2: being split as c cluster centre for each particle in the parent population of lateral cross respectively, and it is corresponding to obtain each particle
Cluster centre;
S4.3: being subordinate between the corresponding user attribute data of each user in the normalized matrix and cluster centre is calculated
Degree:
In formula, 1≤x≤c, 1≤y≤n, n indicate that number of users, m indicate fuzzy coefficient, dxyIt indicates in the normalized matrix
Euclidean distance between y user attribute data and x-th of cluster centre, dkyIndicate y-th of user in the normalized matrix
Euclidean distance between corresponding user attribute data and k-th of cluster centre, uxyIndicate each use in the normalized matrix
Degree of membership between the corresponding user attribute data of family y and cluster centre x;
S4.4: the adaptive value of each particle in the parent population of lateral cross is calculated:
In formula, fit indicates the adaptive value of the corresponding particle of cluster centre.
4. the degree of membership according to claim 3 for equipping services platform user classification for intelligence determines method, feature
It is, according to the adaptive value of the parent population of lateral cross and its particle in step S5, the first is calculated using lateral cross method
Group, is set as crossed longitudinally parent population for the first population, calculates the crossed longitudinally parent according to the normalized matrix
The adaptive value of each particle in population, comprising the following steps:
S5.1: the particle in the parent population of lateral cross is subjected to combination of two, obtains particle pair, and be numbered;
S5.2: successively taking out particle pair in numerical order, and uses following formula to particle under the lateral cross probability of setting
Pair d0Dimension executes horizontal line and intersects:
MShc(i,d0)=r1·X(i,d0)+(1-r1)·X(j,d0)+c1·(X(i,d0)-X(j,d0))
MShc(j,d0)=r1·X(j,d0)+(1-r2)·X(i,d0)+c1·(X(j,d0)-X(i,d0))
Wherein, d0∈ (1, D), D=c × d, d indicate the quantity of the corresponding user attribute data of each user;I ∈ (1, N), j ∈
(1, N), N indicate initial population particle number;r1、r2For the uniform random number on [0,1], c1、c2Between [- 1,1]
Uniform random number;X(i,d0) indicate i-th of particle d0Dimension data;X(j,d0) indicate j-th of particle d0Dimension
According to;MShc(i,d0) and MShc(j,d0) indicate the golden mean of the Confucian school solution after lateral cross;
S5.3: each particle is stored in the first matrix corresponding golden mean of the Confucian school solution, calculates each golden mean of the Confucian school in first matrix
The adaptive value of solution, and adaptive value corresponding with particle each in the parent population of lateral cross is compared, fitness is big
Particle is stored in the first population, and the first population is set as crossed longitudinally parent population.
5. the degree of membership confirmation method according to claim 4 for equipping services platform user classification for intelligence, feature
It is, neutralizes the adaptive value of its particle in step S5 according to crossed longitudinally parent population, calculates second using crossed longitudinally method
Population, comprising the following steps:
S5.4: by being normalized per one-dimensional for crossed longitudinally parent population, each dimension is subjected to combination of two, and carry out
Number;
S5.5: every a pair is successively taken out in numerical order;
S5.6: crossed longitudinally to each pair of data execution using following formula according to crossed longitudinally rate is set:
MSvc(i,d1)=rX (i, d1)+(1-r)·X(i,d2)
In formula, X (i, d1) indicate i-th of particle d1Dimension data, X (i, d2) indicate i-th of particle d2Dimension data, d1,
d2∈ (1, D), d1≠d2, D=c × d, d indicate the quantity of the corresponding user attribute data of each user;i∈(1,N);R be [0,
1] uniform random number on;MSvc(i,d1) indicate i-th of particle d1Tie up offspring data;
S5.7: each particle is respectively tieed up into offspring data and carries out renormalization, obtains corresponding golden mean of the Confucian school solution, and save it in second
In matrix, calculate the adaptive value of each golden mean of the Confucian school solution in second matrix, and with each particle in crossed longitudinally parent population
Adaptive value be compared, the big particle of fitness is stored in the second population, wherein crossed longitudinally parent population be cross
The first population obtained to interior extrapolation method.
6. the degree of membership according to claim 5 for equipping services platform user classification for intelligence determines method, feature
It is, being normalized per one-dimensional by crossed longitudinally parent population in step S5.4 will be longitudinal using following formula
The parent population of intersection is normalized per one-dimensional:
In formula, d1∈ (1, D), Pd1maxFor d1The upper limit of dimension control variable, Pd1minFor d1The lower limit of dimension control variable, k are to work as
Preceding the number of iterations, Xk(i,d1) indicate the number of iterations be k when i-th of particle d1Dimension data, Xk-1(i,d1) indicate iteration time
Number is the d of i-th of particle of k-11Dimension data.
7. the degree of membership according to claim 6 for equipping services platform user classification for intelligence determines method, feature
It is, each particle is respectively tieed up into offspring data in the step S5.7 and carries out renormalization, using following formula by each particle
Each dimension offspring data carries out renormalization:
MSvc'(i,d1)=MSvc(i,d1)·(Pd1max-Pd1min)+Pd1min
Wherein, MSvc(i,d1) indicate i-th of particle d1Tie up offspring data, MSvc'(i,d1) indicate i-th of particle d1
Data after tieing up offspring data renormalization.
8. a kind of degree of membership for equipping services platform user classification for intelligence confirms system characterized by comprising
Raw data matrix establishes module, includes number of users, user's registration information for obtaining from intelligence equipment service platform
User attribute data, and raw data matrix is established according to the user attribute data;
Normalized module normalizes the raw data matrix, obtains normalized matrix;
Initial population generation module, for initial population to be set as to the parent population of lateral cross, according to the normalized matrix
Calculate the adaptive value of each particle in the parent population of the lateral cross;
Degree of membership determining module, for the adaptive value according to the parent population of lateral cross and its particle, using lateral cross method
The first population is calculated, the first population is set as crossed longitudinally parent population, the longitudinal direction is calculated according to the normalized matrix
The adaptive value of each particle in the parent population of intersection, the adaptive value of its particle is neutralized according to crossed longitudinally parent population, is adopted
The second population is calculated with crossed longitudinally method;Judge whether the number of iterations is greater than setting the number of iterations, if it is not, then setting the second population
For the parent population of lateral cross, the suitable of each particle in the parent population of the lateral cross is calculated according to the normalized matrix
It should be worth, be iterated using lateral cross method and crossed longitudinally method and calculate the second population;If so, the second population is split as c
Cluster centre, and calculate the attribute data of each user and the degree of membership of cluster centre, wherein c indicates preset clusters number.
9. the degree of membership according to claim 8 for equipping services platform user classification for intelligence determines system, feature
It is, the degree of membership determining module is used for:
Particle in the parent population of lateral cross is subjected to combination of two, obtains particle pair, and be numbered;
Successively take out particle pair in numerical order, and under the lateral cross probability of setting using following formula to the of particle pair
d0Dimension executes horizontal line and intersects:
MShc(i,d0)=r1·X(i,d0)+(1-r1)·X(j,d0)+c1·(X(i,d0)-X(j,d0))
MShc(j,d0)=r1·X(j,d0)+(1-r2)·X(i,d0)+c1·(X(j,d0)-X(i,d0))
Wherein, d0∈ (1, D), D=c × d, d indicate the quantity of the corresponding user attribute data of each user;I ∈ (1, N), j ∈
(1, N), N indicate initial population particle number;r1、r2For the uniform random number on [0,1], c1、c2Between [- 1,1]
Uniform random number;X(i,d0) indicate i-th of particle d0Dimension data;X(j,d0) indicate j-th of particle d0Dimension
According to;MShc(i,d0) and MShc(j,d0) indicate the golden mean of the Confucian school solution after lateral cross;
Each particle is stored in the first matrix corresponding golden mean of the Confucian school solution, each golden mean of the Confucian school solution is suitable in calculating first matrix
It should be worth, and adaptive value corresponding with particle each in the parent population of lateral cross is compared, and the big particle of fitness is deposited
First population is set as crossed longitudinally parent population in the first population by storage.
10. the degree of membership according to claim 8 for equipping services platform user classification for intelligence determines method, feature
It is, the degree of membership determining module is used for:
By being normalized per one-dimensional for crossed longitudinally parent population, each dimension is subjected to combination of two, and be numbered;
Every a pair is successively taken out in numerical order;
It is crossed longitudinally to each pair of data execution using following formula according to crossed longitudinally rate is set:
MSvc(i,d1)=rX (i, d1)+(1-r)·X(i,d2)
In formula, X (i, d1) indicate i-th of particle d1Dimension data, X (i, d2) indicate i-th of particle d2Dimension data, d1,
d2∈ (1, D), d1≠d2, D=c × d, d indicate the quantity of the corresponding user attribute data of each user;i∈(1,N);R be [0,
1] uniform random number on;MSvc(i,d1) indicate i-th of particle d1Tie up offspring data;
Each particle is respectively tieed up into offspring data and carries out renormalization, obtains corresponding golden mean of the Confucian school solution, and save it in the second matrix
In, calculate the adaptive value of each golden mean of the Confucian school solution in second matrix, and in crossed longitudinally parent population each particle it is suitable
It should be worth and be compared, the big particle of fitness is stored in the second population, wherein crossed longitudinally parent population is laterally to hand over
The first population that fork method obtains.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811603637.4A CN109829473A (en) | 2018-12-26 | 2018-12-26 | The degree of membership for equipping services platform user classification for intelligence determines method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811603637.4A CN109829473A (en) | 2018-12-26 | 2018-12-26 | The degree of membership for equipping services platform user classification for intelligence determines method and system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109829473A true CN109829473A (en) | 2019-05-31 |
Family
ID=66861201
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811603637.4A Pending CN109829473A (en) | 2018-12-26 | 2018-12-26 | The degree of membership for equipping services platform user classification for intelligence determines method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109829473A (en) |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104239961A (en) * | 2014-07-16 | 2014-12-24 | 广东工业大学 | Electric system economic dispatching optimization method based on criss-cross algorithm |
CN104504127A (en) * | 2014-12-29 | 2015-04-08 | 广东电网有限责任公司茂名供电局 | Membership determining method and system for power consumer classification |
-
2018
- 2018-12-26 CN CN201811603637.4A patent/CN109829473A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104239961A (en) * | 2014-07-16 | 2014-12-24 | 广东工业大学 | Electric system economic dispatching optimization method based on criss-cross algorithm |
CN104504127A (en) * | 2014-12-29 | 2015-04-08 | 广东电网有限责任公司茂名供电局 | Membership determining method and system for power consumer classification |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Kazakovtsev et al. | Genetic algorithm with fast greedy heuristic for clustering and location problems | |
CN110084512B (en) | Multi-robot task allocation method for intelligent warehousing system | |
He et al. | A study of drift analysis for estimating computation time of evolutionary algorithms | |
Yang et al. | Applying family competition to evolution strategies for constrained optimization | |
WO2018166270A2 (en) | Index and direction vector combination-based multi-objective optimisation method and system | |
CN107977740A (en) | A kind of scene O&M intelligent dispatching method | |
CN103942419B (en) | Hybrid intelligent optimization method | |
CN109272058B (en) | Integrated power load curve clustering method | |
Dai et al. | An orthogonal evolutionary algorithm with learning automata for multiobjective optimization | |
CN110110395A (en) | A kind of multi-state System Reliability appraisal procedure based on markov and general generating function | |
CN104504127B (en) | Degree of membership defining method and system for classification of power customers | |
CN109902808A (en) | A method of convolutional neural networks are optimized based on floating-point numerical digit Mutation Genetic Algorithms Based | |
CN111369000A (en) | High-dimensional multi-target evolution method based on decomposition | |
CN105578472B (en) | A kind of wireless sensor network performance online Method for optimized planning based on immunity principle | |
CN115938505A (en) | Drug molecule screening method and system based on fusion of graph neural network block structure and multi-head attention mechanism | |
CN107451617A (en) | One kind figure transduction semisupervised classification method | |
Kim et al. | A modified genetic algorithm for fast training neural networks | |
CN109829473A (en) | The degree of membership for equipping services platform user classification for intelligence determines method and system | |
Sooda et al. | A comparative analysis for determining the optimal path using PSO and GA | |
CN115438585A (en) | Double-layer multi-objective evolutionary algorithm-based site selection optimization method | |
Li et al. | An efficient artificial immune network with elite-learning | |
CN105142198B (en) | A kind of method that wireless sensor network optimal path is solved using differential evolution | |
CN110781352B (en) | Method for optimizing topological structure to realize network structure controllability at lowest cost | |
CN107767035A (en) | A kind of electric energy meter detection mixed production line dispatching method based on genetic algorithm | |
Liu et al. | Finding the optimal number of clusters using genetic algorithms |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190531 |
|
RJ01 | Rejection of invention patent application after publication |