CN106844637A - Method is recommended based on the film for just giving cluster to prune improvement multi-objective genetic algorithm - Google Patents
Method is recommended based on the film for just giving cluster to prune improvement multi-objective genetic algorithm Download PDFInfo
- Publication number
- CN106844637A CN106844637A CN201710044461.2A CN201710044461A CN106844637A CN 106844637 A CN106844637 A CN 106844637A CN 201710044461 A CN201710044461 A CN 201710044461A CN 106844637 A CN106844637 A CN 106844637A
- Authority
- CN
- China
- Prior art keywords
- film
- individual
- population
- individuality
- algorithm
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/73—Querying
- G06F16/735—Filtering based on additional data, e.g. user or group profiles
-
- 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)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Physiology (AREA)
- Genetics & Genomics (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Multimedia (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Method is recommended based on the film for just giving cluster to prune improvement multi-objective genetic algorithm, this innovatory algorithm OTNSGA II are proposed for not enough of both distributivity present in NSGA II and convergence, can be used for solving various multi objective function optimization problems.Algorithm design tomography multi-Objective Orthogonal experiment initialization population, it is to avoid individual uneven and cause distributivity to lack;And Evolution of Population process is safeguarded with self-adaption cluster pruning strategy, and an appropriate number of worst individual is removed, maintain the convergence and distributivity of population.With reference to user behavior and the information excavating of film native, the algorithm is applied to film personalized recommendation this practical problem, versatility and validity that test comparison illustrates algorithm are carried out by with existing algorithm, more excellent recommendation results are obtained, improve accuracy rate, recall rate and the coverage rate of recommendation, and there is provided the suggested design combination of more horn of plenty, be conducive to the point of interest of abundant digging user to provide more structurally sound recommendation service.
Description
Technical field
The invention belongs to personalized recommendation technical field.The many mesh being improved with the deficiency for NSGA-II algorithms
Mark genetic algorithm OTNSGA-II algorithms (specifically related to repair by NSGA-II algorithms, the experiment of tomography multi-Objective Orthogonal and self-adaption cluster
Cut strategy) realize the personalized recommendation to film.
Background technology
With the application popularization of Internet technologies and the fast development of hyundai electronicses commercial affairs, the money in internet is full of
The situation that source quantity is exponentially increased.Substantial amounts of information is presented and often causes that user feels at loose ends simultaneously, it is difficult to therefrom seek
Oneself real resource interested is found, so as to occur in that so-called " information explosion " and " information overload " phenomenon.Search engine
Occur to alleviate problem of information overload with information retrieval technique.In information-based today, user is commonly using search
Engine finds the resource required for oneself.But traditional search engine technique does not account for the individual difference of user, by institute
There are user's equivalent processes, return to what the resource of user was just as, simultaneously because feedack amount is also very big so that use
Family is difficult to choose the resource oneself liked.Therefore, how according to the preference characteristics of each user, largely believe from internet
The information for meeting user's request is found in breath, and then recommends user, had become when previous urgently to be resolved hurrily studies a question.
Personalized recommendation system (personlized recommender systems) be exactly to meet the tendency of under this background and
Raw.It obtains the preference characteristics of user by the behavioural characteristic of user in collection system, and then according to these preference characteristics
Digging user is potential interested from the bulk information on network or resources of needs, and makes corresponding recommendation.Recommend
It is exactly to predict user to non-selected resource (such as music, film, book by analyzing resource that user selected on question essence
Nationality, webpage, restaurant, tourist attractions etc.) fancy grade, and the result of prediction is presented to use in certain effective form
Family, such as by predicted value resource recommendation higher to user.
Commending system is widely used in many fields at present, and common website is such as in daily life:Purchase by group net,
Jingdone district store, Taobao, only product meeting, Amazon, Dangdang.com etc. are all typical commending systems.And it is popular at present go where
Can the website of the characteristic theory such as net, ctrip.com, effectively find and draw over to one's side client, allow the recommendation system in client's dependence
Can system, keeps supply of long duration relation, so as to improve sales achievement, namely choose ripe effective commending system relation
To the final and decisive juncture of e-commerce venture.Also, it is recommended to system regions are always by academia as one of temperature research topic, and
It is progressively independent into a special disciplines.
As academia, engineering circles and business circles are to commending system in-depth study, other field is also widely applied to
Commending system and its technology.Nowadays, commending system related algorithm or will be applied to include the data in library and the network information
The information service that retrieval and DTV are watched etc, or even some fairly simple commending systems are mostly applied such as
In " bean cotyledon net ", the network forum of " Baidu's mhkc ".It can be seen that by development for many years, commending system application field also will increasingly
Wide in range, related personnel and scholar are increasing to the research interest in the field.
Commending system performance depends entirely on selected proposed algorithm.The quality of the quality of recommendation has many evaluation marks
Standard, such as the precision recommended, the personalization level of recommendation, accuracy rate, recall rate of recommendation etc..Generally certain or it is a few
When individual standard is optimal, recommend quality relatively good.Many optimization problems can all sum up in scientific research and engineering practice
It is multi-objective optimization question (MOP) that personalized recommendation is also the multi-objective optimization question for considering many aspects.
Multiple-objection optimization is a new branch of science of the applied mathematics developed rapidly over nearly more than 20 years.It studies vector
Optimization problem of object function when meeting certain constraints under certain meaning.Due to a large amount of optimization problems of real world,
The optimization problem containing multiple targets can be all attributed to, since the seventies, for the research of multiple-objection optimization, at home and state
People greatly concern and attention are all caused on border.Particularly during the nearly last ten years, theory study deepens continuously, range of application day
Beneficial extensive, studying team grows rapidly, shows vitality.Meanwhile, with medium-and-large-sized multiple to social economy and engineering design
Miscellaneous system research is goed deep into, and the theoretical and method of multiobjective optimization is also constantly subject to severe challenge and is rapidly developed.
Multiple-objection optimization has it to be widely applied field as an important research direction in optimization field, and research solves its effective calculation
Method has great scientific meaning and application value.
The research purpose of multi-objective Evolutionary Algorithm is mainly makes the disaggregation tried to achieve as much as possible close to the Pareto of problem
Preferable forward position, and widely distributed and uniform, this just determines that the performance indications of evaluation algorithms are distributivity and convergence two
Aspect.Distributivity and convergence have great importance for solving multi-objective optimization question, and good distributivity can be to certainly
Plan person provides more rationally effective selection schemes;Good convergence can more accurately solve the answer of practical problem.
In recent years, evolutionary computation field proposes some multi-objective Evolutionary Algorithms in succession.Wherein, it is most representational mainly to have:
The SPEA (Strength Pareto Evolutionary Algorithm) that Zitzler and Thiele is proposed, Kim et al. is at it
On the basis of propose SPEA2, Srinivas and Deb propose non-dominated sorted genetic algorithm NSGA (Non-dominated
Sorting Genetic Algorithm), and the proposition such as the NSGA-II that Deb etc. is proposed on its basis, Corne PESA
(Pareto Envelope-based Selection Algorithm) and PESA-II.NSGA-II is that a kind of application is the widest
The characteristics of general multi-objective Evolutionary Algorithm, algorithm is to determine individuality according to the Pareto dominance relations between individuality and density information
Adaptive value, but such fitness calculation there is distributivity and convergence safeguards improperly defect.The roots such as Wen Shihua
Retain some representative individualities according to the mode of distance metric to improve NSGA-II, this mode only considered
Crowding distance to keeping the influence of Species structure, from the angle of similar individual collections consider comprehensively feature it is close it is individual with
Worst individual there is a problem of the convergence for causing and distributivity missing.For the defect of NSGA-II these two aspects, carry herein
Gone out a kind of improved convergence and distributivity and kept strategy, algorithm before evolution, using tomography non-dominated ranking and crowded
The individual mode of Distance evaluation sets multi-Objective Orthogonal experiment initialization population, prevents population to be easily trapped into because of random initializtion
Local convergence restrained slow, and it is individual uneven and cause distributivity to lack to avoid initial population;Algorithm was being evolved
Cheng Zhong, the evolution result to every generation is clustered, the dynamics that the size dynamic regulation in class according to similarity is pruned, and is passed through
An appropriate number of feature of removal of self adaptation is close and non-dominated ranking is individual with the poor class of crowding distance and away from
A small amount of individuality of leading surface carries out population maintenance, accelerates to maintain distributivity again while convergence in population.Should by the innovatory algorithm
For the particular problem of film personalized recommendation, by with NSGA-II and Collaborative Filtering Recommendation Algorithm based on user and
The conventional recommendation such as content-based recommendation algorithm algorithm carries out contrast experiment under the same conditions, demonstrates the effect of algorithm.
The content of the invention
Film is used for based on the improvement multi-objective genetic algorithm for just giving cluster to prune the purpose of the present invention is to propose to a kind of
The solution of this practical problem of TOP-N in personalized recommendation, it is pre- with the film user's behavior prediction scoring of N portions and N portions film native
Test and appraisal are divided into two targets and optimize, and realize personalized recommendation.
It is of the invention based on just give cluster prune improvement multi-objective genetic algorithm (OTNSGA-II), it is characterised in that:
Algorithm set multi-Objective Orthogonal experiment just before evolution by the way of tomography non-dominated ranking and crowding distance evaluation individuality
Beginningization population, prevents population local convergence to be easily trapped into because of random initializtion or was restrained slowly, and avoid initial population
It is individual uneven and cause distributivity to lack;During evolution, the evolution result to every generation is clustered algorithm, in class
The dynamics that size dynamic regulation according to similarity is pruned, by the way that an appropriate number of feature of the removal of self adaptation is close and non-branch
With sequence, individual and away from leading surface a small amount of individuality carries out population maintenance with the poor class of crowding distance, accelerates population and receives
Distributivity is maintained while holding back again.
Method is recommended based on the film for just giving cluster to prune improvement multi-objective genetic algorithm, is comprised the following steps:
S1 carries out individual UVR exposure, initialization data, and setup parameter
The individual numbering for representing N number of film, wherein N represents the film number for needing to recommend;Use diRepresenting i-th needs
The film to be recommended is numbered, and in order to be applied to solve the problems, such as the TOP-N in personalized recommendation field, is numbered using N number of film and combined,
Individual UVR exposure form is:<d1, d2, di。。。dN>, using real coding mode, scope of the coding range where film numbering is simultaneously
And be integer form;Population Size is initialized as popszie by the initialization data, and each offspring produces popsize big
Small population;The setup parameter includes:It is 0.9 to set crossover probability Pc, and mutation probability Pm is 0.1, and individual lengths are N,
Film is numbered in order and not repeated in keeping coding, as a kind of combination of N portions difference film to be recommended.
S2 tomography multi-Objective Orthogonals experiment method initializes population
S2.1 is used and is generated M N individual, the interim initial population P of composition according to following solution space discretization mode;
S2.1.1 finds the s dimensions for meeting following formula;
If S2.2.2 solution spaces are [l, u], l, u are two specific numerals, represent the upper bound of feasible solution with respectively
Boundary, then be divided into S sub-spaces in S Wei Chu by solution space:
[l (1), u (1)], [l (2), u (2)] ... [l (S), u (S)]
Wherein, Is=[c1,j]1×N,IsRepresent a column vector, the tool for calculating l (i) and u (i)
Body numerical value, c1,jBe one according to the value of j determine 0 or 1 digital j=1,2 ... S, for constituting column vector IsOne
Element.
During above-mentioned steps due to be related to film number as encode gene position, to each individuality in N number of gene
Position uses floor operation, and keeps the film that repetition is occurred without in individuality to number;If there is the film numbering for repeating, to the volume
Number it is removed, and it is unduplicated other films numbering to supplement.
S2.2 calculates each individuality in population P the numerical value f of each target in problem to be optimizedit, it is expressed as i-th
The individual numerical value on t-th object function.
S2.3 is to each individuality in population P according to each target value fitBetween good and bad relation carry out non-dominated ranking
The calculating of layering, and the level Si where marking each individuality, the level where being expressed as i-th individuality.
S2.4 in layer obtains individuality and is put into another set successively according to level ranking since ground floor, until
The individual amount of this set is more than or equal to 4*popszie;If the individual amount of population P takes this kind less than 4*popszie
Group individuality complete or collected works, now the several numbers of different layers in set are m, and this is gathered to be and alternatively gathers, wherein, popszie is to plant
Group's individual amount.
S2.5 calculates crowding distance d in case selected works are combined into complete or collected works space to each individuality in alternative seti, it is expressed as
I-th numerical value of individual crowding distance.
S2.6 in alternative set, with each individual non-dominated ranking level SiWith crowding distance diAs two targets,
The non-dominant relation between individuality is evaluated to select the individual generation initial population P of optimal first popszie0。
S3 selection operations
Individual choice is carried out in the way of championship, i.e., randomly chooses k, k from this popszie individuality<Popszie
Individuality, takes k for n/2 (rounding) here, and an optimum individual is chosen from this k individuality.
Selection standard is with number of plies S where non-dominated rankingiWith crowding distance numerical value diAs two targets, according to non-branch
With winning relation, each individuality is compared, relatively more winning individuality is more excellent individuality.
S4 crossover operations
Crossover operation uses SBX crossover operators, it is assumed that when former generation (the t generations in evolutionary process) two individualities to be intersected
It is XA t、XB t, α is to intersect the parameter (span is 0~1) being related to, then XA t+1、XB t+1It is two individualities for producing of future generation.
Form is as follows:
S5 mutation operations
Gen during to entering is for population PgenIn any individual pi=(pi1, pi2…piN), i ∈ { 1,2 ...
N }, mutation operation is participated in probability P m:Produce decimal r ∈ [0,1], and random integers j ∈ [1, N];Make pi, j=
Lj+r* (uj-lj), to colony PgenEnter row variation and produce new population PgenNew。
S6 self-adaption clusters locally prune strategy
S6.1 sets population PgenNewIn have popszie it is individual, the population is clustered using K-means clustering algorithms,
Obtain k class;
S6.2 in the class in k class of cluster result in S6.1 two-by-two individuality between calculate similarity, then calculate class
Interior average similarity Pk;
S6.3 average similarity P in the class in each class calculated in S6.2kProtected in the calculating of self adaptation each class
The individual amount for staying, in reserved category the percentage computing formula of individual amount with reference to as follows, wherein, prune parameter δ for 0.12~
0.15;
1-δ*Pk
S6.4 according to the reservation individual amount calculated in S6.3, the non-dominant that prunes away layering ranking and crowding distance calculate compared with
Differ from and away from a small amount of individuality of preferable leading surface;
S6.5 by these prune after remaining individual reformulate new population PgenNew, continue executing with follow-on evolution
Process.
S7 end conditions judge
If reaching the algebraically of regulation or obtaining satisfied result, terminate and output result, otherwise turn S3 steps.
S8 enters next heredity circulation
By population PgenNewAs the initial population evolved of future generation, proceed S3 steps.
It is more than the main process flow steps based on the improvement multi-objective genetic algorithm for just giving cluster to prune, uses the algorithm
Flow is applied to the recommendation list that film personalized recommendation problem obtains film.
Compared with prior art, the present invention has the advantages that.
Based on just give cluster prune improvement multi-objective genetic algorithm, algorithm before evolution, using tomography non-dominant
Sequence and the individual mode of crowding distance evaluation set multi-Objective Orthogonal experiment initialization population, prevent population because of random initializtion
And be easily trapped into local convergence or restrained slow, and it is individual uneven and cause distributivity to lack to avoid initial population;Calculate
During evolution, the evolution result to every generation is clustered method, and the size dynamic regulation in class according to similarity is pruned
Dynamics, by an appropriate number of feature of removal of self adaptation is close and non-dominated ranking and the poor class of crowding distance in
Body and a small amount of individuality away from leading surface carry out population maintenance, accelerate to maintain distributivity again while convergence in population.Pass through
The innovatory algorithm is applied in the particular problem of film personalized recommendation, with NSGA-II and collaborative filtering based on user
Proposed algorithm and content-based recommendation algorithm both traditional proposed algorithms carry out contrast experiment under the same conditions, verify
The practical function of algorithm.
Brief description of the drawings
Fig. 1 is the flow chart based on the improvement multi-objective genetic algorithm for just giving cluster to prune.
Fig. 2 is that the experimental design of tomography multi-Objective Orthogonal initializes population broad flow diagram.
Fig. 3 is that self-adaption cluster prunes tactful broad flow diagram.
Fig. 4 is General Implementing method flow diagram of the present invention.
Fig. 5 is the Collaborative Filtering Recommendation Algorithm broad flow diagram based on user.
Fig. 6 is content-based recommendation algorithm broad flow diagram.
Specific embodiment
The present invention will be further described with specific embodiment below in conjunction with the accompanying drawings.
The data set that the present invention is recommended from Movielens as film, the data set is related to 100,000 film to comment
Point, 943 users and 1682 films.Collaborative filtering (UserCF) based on user, content-based recommendation are calculated
Method (CB), tetra- algorithms of NSGA-II and OTNSGA-II carry out contrast experiment.
In two kinds of multi-objective optimization algorithms of NSGA-II and OTNSGA-II, numbering is the film sequence number of N portions film
(N values are 5,10,15,20 here), numbering is limited to not repeat in order, and both operation algebraically gen are 100, population scale
Popsize is 50, and it is 0.9 to set crossover probability Pc, and mutation probability Pm is 0.1, with using most close in user's similarity matrix
K user calculate the corresponding prediction scoring sum of N portions film numbering and calculate similar using film native similarity matrix
The corresponding prediction scoring sum of N number of film numbering that film is obtained is used as two optimization object functions.It is above-mentioned for having used
Single portion's film prediction scoring that two ways is calculated is recorded, and is calculated same film numbering corresponding film next time and is commented
Dividing need not then calculate, and directly read score value summation.With commonly use recommendation evaluation criterion accuracy rate, recall rate and
The abundant degree of coverage rate and suggested design carrys out the effect of verification algorithm as index.Wherein, accuracy rate, recall rate and covering
The computational methods of rate are as follows:
Assuming that carrying out recommendation test to n user on test set I, the article collection for making user i like is combined into L (i), makes R
I () represents the set of the article recommended user i.
For accuracy rate index is then corresponding commending system, accuracy rate index can carry out table with following mathematical formulae
Show:
In above formula, | R (i) | expressions be article in recommendation list number.From above formula, accuracy rate describes to push away
The number for recommending the article that user in list likes accounts for the ratio of whole recommendation list.Usual businessman is intended to the article of user's recommendation
As far as possible all be article that user likes, so it has been generally acknowledged that accuracy rate the higher the better.
Recall rate can be then indicated with following mathematical formulae:
In above formula, | R (i) ∩ L (i) | represents that recommendation list and user i like the individual of the article of the part of overlap in list
Number, can be seen that recall rate is directed to for user from above-mentioned recall rate formula, and what recall rate was calculated is used in recommendation list
The article number that family is liked takes the ratio that list is entirely liked at family.
Coverage rate describes the ratio that recommendation list occupies total article, and coverage rate can carry out table with following mathematical formulae
Show:
Main flow such as Fig. 1 institutes based on the improvement multi-objective genetic algorithm for just giving cluster to prune proposed by the present invention
Show, whole flow process initializes population with the experimental design of tomography multi-Objective Orthogonal and self-adaption cluster prunes strategy and safeguards population
Distributivity and convergence.Be broadly divided into the experimental design of tomography multi-Objective Orthogonal initialization population, selection operation, crossover operation,
Mutation operation, self-adaption cluster prune tactful five parts, wherein, tomography multi-Objective Orthogonal experimental design initialization population is main
Flow is as shown in Fig. 2 the tactful main flow of self-adaption cluster pruning is as shown in Figure 3.Selection, intersection, three operating processes of variation
Operating process with NSGA-II algorithms is consistent.
Implementation process of the invention is described in detail with reference to Fig. 4.Embodiments of the invention are with the technology of the present invention
Implemented premised on scheme, given detailed implementation method and specific operating process, but protection scope of the present invention
It is not limited to following embodiments.
Embodiment is from film personalized recommendation problem to multiple target innovatory algorithm OTNSGA-II presented herein and base
Collaborative filtering (UserCF), content-based recommendation algorithm (CB) and NSGA-II in user are tested and compared.
By comparing it can be seen that algorithms of different is processing the performance state of same problem under same experiment condition.
Wherein, based on user the score value that collaborative filtering (UserCF) is seen a film according to user, calculates two
Similarity between two users, and select and sorted from big to small preceding K user with the similarity of user to be recommended, it is similar with these
Number of degrees value is weight, and the film seen and scored using these users is predicted to the film that user to be recommended has not seen
Scoring.Main flow is as shown in Figure 5.
Content-based recommendation algorithm (CB) calculates the similarity between each film according to the belonging relation of film types
Size, and the film seen to user to be recommended scores film higher, according to similar degree size to not seeing
The film crossed is predicted scoring.Main flow is as shown in Figure 6.
In this embodiment, for NSGA-II and OTNSGA-II, to use most phase in user's similarity matrix
K near user calculates the corresponding prediction scoring sum of N portions film numbering and calculates phase using film native similarity matrix
The corresponding prediction of N number of film numbering obtained like film scores sum as two optimization object functions.It is upper for having used
State single portion's film prediction scoring that two ways calculates to record, next time calculates same film and numbers corresponding film
Scoring need not then be calculated, and directly read score value summation.Purpose is the numbering group for solving N number of different films
Close, i.e., each individual gene position is a numbering for film, uses diRepresent the numbering of i portions film to be recommended, individual UVR exposure shape
Formula is:<d1, d2, di。。。dN>, using real coding mode, coding range is the scope where film numbering and is integer shape
Formula, film is numbered in order and not repeated in keeping coding, as a kind of combination of N portions difference film to be recommended.
The explanation of involved each detailed problem in the inventive technique scheme is given in detail below:
S1 carries out individual UVR exposure, initialization data, and setup parameter
The individual numbering for representing N number of film, wherein N represents the film number for needing to recommend;Use diRepresenting i-th needs
The film to be recommended is numbered, and in order to be applied to solve the problems, such as the TOP-N in personalized recommendation field, is numbered using N number of film and combined,
Individual UVR exposure form is:<d1, d2, di。。。dN>, using real coding mode, scope of the coding range where film numbering is simultaneously
And be integer form;Population Size is initialized as popszie by the initialization data, and each offspring produces popsize big
Small population;The setup parameter includes:It is 0.9 to set crossover probability Pc, and mutation probability Pm is 0.1, and individual lengths are N,
Film is numbered in order and not repeated in keeping coding, as a kind of combination of N portions difference film to be recommended.
S2 tomography multi-Objective Orthogonals experiment method initializes population
S2.1 is used and is generated M N individual, the interim initial population P of composition according to following solution space discretization mode;
S2.1.1 finds the s dimensions for meeting following formula;
If S2.2.2 solution spaces are that [l, u] (l, u are two specific numerals, represent the upper bound of feasible solution with respectively
Boundary), then solution space is divided into S sub-spaces in s Wei Chu:
[l (1), u (1)], [l (2), u (2)] ... [l (S), u (S)]
Wherein, Is=[c1,j]1×N,IsRepresent a column vector, the tool for calculating l (i) and u (i)
Body numerical value, c1,jBe one according to the value of j determine 0 or 1 numeral (j=1,2 ... S), for constituting column vector IsOne
Individual element.
During above-mentioned steps due to be related to film number as encode gene position, to each individuality in N number of gene
Position uses floor operation, and keeps occurring without the film numbering of repetition (if there is the film numbering for repeating, to the volume in individuality
Number it is removed, and it is unduplicated other films numbering to supplement).
S2.2 calculates each individuality in population P the numerical value f of each target in problem to be optimizedit, it is expressed as i-th
The individual numerical value on t-th object function.
S2.3 is to each individuality in population P according to each target value fitBetween good and bad relation carry out non-dominated ranking
The calculating of layering, and the level Si where marking each individuality, the level where being expressed as i-th individuality.
S2.4 in layer obtains individuality and is put into another set successively according to level ranking since ground floor, until
The individual amount of this set is more than or equal to 4*popszie (if the individual amount of population P takes this kind less than 4*popszie
Group's individuality complete or collected works), now the several numbers of different layers in set are m, and this is gathered to be and alternatively gathers, wherein, popszie is to plant
Group's individual amount.
S2.5 calculates crowding distance d in case selected works are combined into complete or collected works space to each individuality in alternative seti, it is expressed as
I-th numerical value of individual crowding distance.
S2.6 in alternative set, with each individual non-dominated ranking level SiWith crowding distance diAs two targets,
The non-dominant relation between individuality is evaluated to select the individual generation initial population P of optimal first popszie0。
S3 selection operations
Individual choice is carried out in the way of championship, i.e., k (k are randomly choosed from this popszie individuality<popszie)
Individuality, takes k for n/2 (rounding) here, and an optimum individual is chosen from this k individuality.
Selection standard is with number of plies S where non-dominated rankingiWith crowding distance numerical value diAs two targets, according to non-branch
With winning relation, each individuality is compared, relatively more winning individuality is more excellent individuality.
S4 crossover operations
Crossover operation uses SBX crossover operators, it is assumed that when former generation (the t generations in evolutionary process) two individualities to be intersected
It is XA t、XB t, α is to intersect the parameter (span is 0~1) being related to, then XA t+1、XB t+1It is two individualities for producing of future generation.
Form is as follows:
S5 mutation operations
Gen during to entering is for population PgenIn any individual pi=(pi1, pi2…piN), i ∈ { 1,2 ...
N }, mutation operation is participated in probability P m:Produce decimal r ∈ [0,1], and random integers j ∈ [1, N];Make pi, j=
Lj+r* (uj-lj), to colony PgenEnter row variation and produce new population PgenNew。
S6 self-adaption clusters locally prune strategy
S6.1 sets population PgenNewIn have popszie it is individual, the population is clustered using K-means clustering algorithms,
Obtain k class;
S6.2 in the class in k class of cluster result in S6.1 two-by-two individuality between calculate similarity, then calculate class
Interior average similarity Pk;
S6.3 average similarity P in the class in each class calculated in S6.2kProtected in the calculating of self adaptation each class
The individual amount for staying, in reserved category the percentage computing formula of individual amount with reference to as follows, wherein, prune parameter δ for 0.12~
0.15;
1-δ*Pk
S6.4 according to the reservation individual amount calculated in S6.3, the non-dominant that prunes away layering ranking and crowding distance calculate compared with
Differ from and away from a small amount of individuality of preferable leading surface;
S6.5 by these prune after remaining individual reformulate new population PgenNew, continue executing with follow-on evolution
Process.
S7 end conditions judge
If reaching the algebraically of regulation or obtaining satisfied result, terminate and output result, otherwise turn S3 steps.
S8 enters next heredity circulation
By population PgenNewAs the initial population evolved of future generation, proceed S3 steps.
It is more than the main process flow steps based on the improvement multi-objective genetic algorithm for just giving cluster to prune, uses the algorithm
Flow is applied to the recommendation list that film personalized recommendation problem obtains film.
Explanation experimental result of the invention is explained in detail below:
In order to prove validity of the method for the invention in film personalized recommendation problem, OTNSGA- is respectively adopted
II (method in the present invention) and UserCF, CB and NSGA-II carry out excellent to the TOP-N problems in film personalized recommendation
Change, wherein, N is the number (it is 5,10,15,20 to distinguish value here) of recommendation film in a combination.The experimental result such as institute of table 1
Show (wherein overstriking font is the preferable experimental data of performance).
The accuracy rate of the OTNSGA-II of table 1 and UserCF, CB and NSGA-II, recall rate and coverage rate contrast
As shown in Table 1, can under conditions of N=5,10,15,20 using OTNSGA-II (method in the present invention)
It is effective to improve accuracy rate, recall rate and the coverage rate recommended, and NSGA-II then shows poor in these two aspects, UserCF and
Both traditional recommendation methods of CB are then worse compared to OTNSGA-II.This is absolutely proved for the improved present invention sides of NSGA-II
The recommendation results that method OTNSGA-II is obtained increase than NSGAII method in terms of accuracy rate, recall rate and coverage rate is improved,
It is also significantly better than the traditional recommendation methods of two kinds of UserCF, CB.Therefore, compared with prior art, the present invention can be obtained more
Excellent recommendation results, improve accuracy rate, recall rate and the coverage rate of recommendation, are conducive to the point of interest of abundant digging user to carry
For more structurally sound recommendation service.
For the assembled scheme recommended, using OTNSGA-II (present invention in method) and UserCF, CB and
The combination experimental result that NSGA-II algorithms are obtained is as shown in table 2 (wherein overstriking font is the preferable experimental data of performance).
The OTNSGA-II of table 2 and the abundant degree contrast of the suggested design of UserCF, CB and NSGA-II
As shown in Table 2, more various solution can be obtained than NSGAII using OTNSGA-II (method in the present invention)
Scheme, and being recommended the characteristics of comprehensively examined user's history behavior with film native in itself so that recommend more comprehensively and
It is personalized;And UserCF and CB both traditional recommendation methods can only obtain a solution, this way of recommendation is not only
Personalized hobby that is dull and ignoring user, and this way of recommendation more depends on the calculating of the same way of recommendation
As a result, the fault tolerances of uniformity are easily caused.It is excellent from multiple target for these solutions that OTNSGA-II is obtained
All it is non-bad from the perspective of change, i.e., is each other equivalent good and bad relation, businessman can be according to different solution
More various film is provided and recommends combination, user can be more suitable for oneself current interest according to different solution selections
Film combination is watched so that recommendation results are more accurate and practical.
Claims (2)
1. method is recommended based on the film for just giving cluster to prune improvement multi-objective genetic algorithm, it is characterised in that:The method bag
Include following steps,
S1 carries out individual UVR exposure, initialization data, and setup parameter
The individual numbering for representing N number of film, wherein N represents the film number for needing to recommend;Use diRepresent that i-th needs is pushed away
The film numbering recommended, in order to be applied to solve the problems, such as the TOP-N in personalized recommendation field, is numbered using N number of film and combined, individual
Coding form is:<d1, d2, di;;;dN>, using real coding mode, coding range is the scope where film numbering and is
Integer form;Population Size is initialized as popszie by the initialization data, and each offspring produces popsize sizes
Population;The setup parameter includes:It is 0.9 to set crossover probability Pc, and mutation probability Pm is 0.1, and individual lengths are N, are kept
Film numbering is orderly in coding and does not repeat, as a kind of combination of N portions difference film to be recommended;
S2 tomography multi-Objective Orthogonals experiment method initializes population
S2.1 is used and is generated M N individual, the interim initial population P of composition according to following solution space discretization mode;
S2.1.1 finds the s dimensions for meeting following formula;
If S2.2.2 solution spaces are [l, u], l, u are two specific numerals, the upper bound and the lower bound of feasible solution are represented respectively, then
Solution space is divided into S sub-spaces in s Wei Chu:
[l (1), u (1)], [l (2), u (2)] ... [l (S), u (S)]
Wherein, Is=[c1,j]1×N,IsRepresent a column vector, the specific number for calculating l (i) and u (i)
Value, c1,jBe one according to the value of j determine 0 or 1 digital j=1,2 ... S, for constituting column vector IsA unit
Element;
During above-mentioned steps due to be related to film number as encode gene position, to each individuality in N number of gene position adopt
With floor operation, and the film that repetition is occurred without in individuality is kept to number;If there is repeat film numbering, to this number into
Row removal, and supplement as unduplicated other films are numbered;
S2.2 calculates each individuality in population P the numerical value f of each target in problem to be optimizedit, it is expressed as i-th individuality
Numerical value on t-th object function;
S2.3 is to each individuality in population P according to each target value fitBetween good and bad relation carry out non-dominated ranking layering
Calculating, and the level Si where marking each individuality, the level where being expressed as i-th individuality;
S2.4 in layer obtains individuality and is put into another set successively according to level ranking since ground floor, until this
The individual amount of set is more than or equal to 4*popszie, if the individual amount of population P takes the population less than 4*popszie
Body complete or collected works, the several numbers of different layers in now gathering are m, and this set is alternative set, wherein, popszie is population
Body number;
S2.5 calculates crowding distance d in case selected works are combined into complete or collected works space to each individuality in alternative seti, it is expressed as i-th
The numerical value of individual crowding distance;
S2.6 in alternative set, with each individual non-dominated ranking level SiWith crowding distance diAs two targets, evaluate
To select, optimal first popszie is individual to generate initial population P to non-dominant relation between individuality0;
S3 selection operations
Individual choice is carried out in the way of championship, i.e., randomly chooses k, k from this popszie individuality<Popszie
Body, takes k for n/2 is rounded here, and an optimum individual is chosen from this k individuality;
Selection standard is with number of plies S where non-dominated rankingiWith crowding distance numerical value diIt is excellent according to non-dominant as two targets
Victory relation, is compared to each individuality, and relatively more winning individuality is more excellent individuality;
S4 crossover operations
Crossover operation uses SBX crossover operators, it is assumed that when t generation two individualities to be intersected in former generation evolutionary process are XA t、
XB t, α is that the parameter value scope that intersection is related to is 0~1, then XA t+1、XB t+1It is two individualities for producing of future generation;Form is such as
Under:
S5 mutation operations
Gen during to entering is for population PgenIn any individual pi=(pi1, pi2…piN), i ∈ { 1,2 ... n }, with general
Rate Pm participates in mutation operation:Produce decimal r ∈ [0,1], and random integers j ∈ [1, N];Make pi, j=lj+r*
(uj-lj), to colony PgenEnter row variation and produce new population PgenNew;
S6 self-adaption clusters locally prune strategy
S6.1 sets population PgenNewIn have popszie it is individual, the population is clustered using K-means clustering algorithms, obtain
K class;
S6.2 in the class in k class of cluster result in S6.1 two-by-two individuality between calculate similarity, then calculate class in
Average similarity Pk;
S6.3 average similarity P in the class in each class calculated in S6.2kSelf adaptation calculate each class in retain
Body number, the percentage computing formula reference of individual amount is as follows in reserved category, wherein, it is 0.12~0.15 to prune parameter δ;
1-δ*Pk
S6.4 according to the reservation individual amount calculated in S6.3, the non-dominant that prunes away layering ranking and crowding distance calculate it is poor with
And away from a small amount of individuality of preferable leading surface;
S6.5 by these prune after remaining individual reformulate new population PgenNew, continue executing with follow-on evolutionary process;
S7 end conditions judge
If reaching the algebraically of regulation or obtaining satisfied result, terminate and output result, otherwise turn S3 steps;
S8 enters next heredity circulation
By population PgenNewAs the initial population evolved of future generation, proceed S3 steps;
It is more than the main process flow steps based on the improvement multi-objective genetic algorithm for just giving cluster to prune, uses the algorithm flow
It is applied to the recommendation list that film personalized recommendation problem obtains film.
It is 2. according to claim 1 that method is recommended based on the film for just giving cluster to prune improvement multi-objective genetic algorithm,
It is characterized in that:The method for calculating ideal adaptation angle value in population is as follows,
In this application scene, calculating for the individual relevance grade of multi-objective optimization question, it is necessary to calculate multiple object functions,
Then the number of plies where calculating each individuality with the method for non-dominated ranking, using the number of plies as individual virtually applicable angle value;
Two object functions are referred here to, refers respectively to calculate N using K user most close in user's similarity matrix
The corresponding prediction scoring sum of portion's film numbering and the N number of electricity obtained using film native similarity matrix calculating similar movies
The corresponding prediction scoring sum of shadow numbering;For the single portion's film prediction scoring note calculated using above two mode
Record is got off, and calculating the corresponding film scoring of same film numbering next time need not then calculate, and directly read the score value and ask
With;
Wherein, non-dominated ranking refers to being asked for multiple-objection optimization for famous multi-objective Optimization scholar Pareto propositions
Winning mode between two solutions of topic:If a solution PiCorresponding all desired values are better than another solution PjIt is corresponding all
Desired value, it is believed that PiBetter than Pj;If two solution PiAnd PjDo not possess such quality relation between corresponding all targets, then
Think that both are non-dominance relations;Individual, to be selected first in whole population non-dominant is selected in such a manner
Body set C1As the 1st layer, the non-dominant individual collections C for selecting to obtain then is proceeded in remaining all individualities2It is
2 layers, go on always until finding last individual collections Cm, wherein m is total number of plies;
Wherein, before first generation evolution, the score value that the algorithm is seen a film according to user, between two two users of calculating
Similarity, and select and sorted from big to small preceding K user with the similarity of user to be recommended;During evolution, the algorithm with
Above-mentioned K user similarity numerical value is weight, and the N portions film to be recommended of active user is calculated using the score data of these users
Prediction scoring, and sue for peace, as first aim numerical value;
Meanwhile, before first generation evolution, the algorithm calculates similar between each film according to the attribute value between film
Degree size;During evolution, the algorithm is scored as weight with the film that user to be recommended has seen, and calculates active user's
The prediction scoring of N portions film degree of correlation to be recommended, and sue for peace, as second target numerical value, so as to form multiple target electricity
Shadow recommends the functional value calculation of problem;
Single portion's film prediction scoring for having been calculated using above two mode is recorded, and next time calculates same
Film is numbered corresponding film scoring and need not then be calculated, and directly reads score value summation.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710044461.2A CN106844637B (en) | 2017-01-19 | 2017-01-19 | Movie recommendation method for improving multi-target genetic algorithm based on orthogonal and clustering pruning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710044461.2A CN106844637B (en) | 2017-01-19 | 2017-01-19 | Movie recommendation method for improving multi-target genetic algorithm based on orthogonal and clustering pruning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN106844637A true CN106844637A (en) | 2017-06-13 |
CN106844637B CN106844637B (en) | 2020-01-24 |
Family
ID=59120722
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710044461.2A Active CN106844637B (en) | 2017-01-19 | 2017-01-19 | Movie recommendation method for improving multi-target genetic algorithm based on orthogonal and clustering pruning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106844637B (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108153918A (en) * | 2018-02-05 | 2018-06-12 | 北京工业大学 | Method is recommended based on the personalized film of multi-objective optimization algorithm that adaptive quadrature intersects |
CN108681570A (en) * | 2018-05-04 | 2018-10-19 | 安徽大学 | A kind of individualized webpage recommending method based on multi-objective Evolutionary Algorithm |
CN108804586A (en) * | 2018-05-27 | 2018-11-13 | 北京工业大学 | The personalized film that fusion grid deposits the multiple-objection optimization of dominant strategy recommends method |
CN109948016A (en) * | 2017-10-31 | 2019-06-28 | 北京嘀嘀无限科技发展有限公司 | Application message method for pushing, device, server and computer readable storage medium |
CN110222089A (en) * | 2019-05-31 | 2019-09-10 | 华南理工大学 | A kind of double cluster results of the constraint based on order-preserving submatrix and missing values prediction technique |
CN110928873A (en) * | 2019-10-09 | 2020-03-27 | 中国人民解放军国防科技大学 | Spatial object indexing and querying method based on improved non-dominated sorting genetic algorithm |
CN110955702A (en) * | 2019-11-28 | 2020-04-03 | 江南大学 | Pattern data mining method based on improved genetic algorithm |
CN111143685A (en) * | 2019-12-30 | 2020-05-12 | 第四范式(北京)技术有限公司 | Recommendation system construction method and device |
CN111191076A (en) * | 2019-12-30 | 2020-05-22 | 北京工业大学 | Movie recommendation method based on prospect theory and multi-target evolution |
CN112765398A (en) * | 2021-01-04 | 2021-05-07 | 珠海格力电器股份有限公司 | Information recommendation method and device and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2014020214A1 (en) * | 2012-08-01 | 2014-02-06 | Universitat Politècnica De Catalunya | Improved method for spectral clustering by computer and uses thereof |
CN103985055A (en) * | 2014-05-30 | 2014-08-13 | 西安交通大学 | Stock market investment decision-making method based on network analysis and multi-model fusion |
CN104008204A (en) * | 2014-06-17 | 2014-08-27 | 南京大学 | Dynamic multi-dimensional context awareness film recommending system and achieving method thereof |
CN105868281A (en) * | 2016-03-23 | 2016-08-17 | 西安电子科技大学 | Location-aware recommendation system based on non-dominated sorting multi-target method |
-
2017
- 2017-01-19 CN CN201710044461.2A patent/CN106844637B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2014020214A1 (en) * | 2012-08-01 | 2014-02-06 | Universitat Politècnica De Catalunya | Improved method for spectral clustering by computer and uses thereof |
CN103985055A (en) * | 2014-05-30 | 2014-08-13 | 西安交通大学 | Stock market investment decision-making method based on network analysis and multi-model fusion |
CN104008204A (en) * | 2014-06-17 | 2014-08-27 | 南京大学 | Dynamic multi-dimensional context awareness film recommending system and achieving method thereof |
CN105868281A (en) * | 2016-03-23 | 2016-08-17 | 西安电子科技大学 | Location-aware recommendation system based on non-dominated sorting multi-target method |
Non-Patent Citations (1)
Title |
---|
刘士琛: "面向推荐系统的关键问题研究及应用", 《中国博士学位论文全文数据库·信息科技辑》 * |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109948016A (en) * | 2017-10-31 | 2019-06-28 | 北京嘀嘀无限科技发展有限公司 | Application message method for pushing, device, server and computer readable storage medium |
CN108153918B (en) * | 2018-02-05 | 2021-07-30 | 北京工业大学 | Personalized movie recommendation method based on adaptive orthogonal intersection multi-objective optimization algorithm |
CN108153918A (en) * | 2018-02-05 | 2018-06-12 | 北京工业大学 | Method is recommended based on the personalized film of multi-objective optimization algorithm that adaptive quadrature intersects |
CN108681570A (en) * | 2018-05-04 | 2018-10-19 | 安徽大学 | A kind of individualized webpage recommending method based on multi-objective Evolutionary Algorithm |
CN108681570B (en) * | 2018-05-04 | 2021-09-21 | 安徽大学 | Personalized webpage recommendation method based on multi-objective evolutionary algorithm |
CN108804586A (en) * | 2018-05-27 | 2018-11-13 | 北京工业大学 | The personalized film that fusion grid deposits the multiple-objection optimization of dominant strategy recommends method |
CN108804586B (en) * | 2018-05-27 | 2021-08-06 | 北京工业大学 | Multi-objective optimization personalized movie recommendation method fusing grid optimization strategy |
CN110222089A (en) * | 2019-05-31 | 2019-09-10 | 华南理工大学 | A kind of double cluster results of the constraint based on order-preserving submatrix and missing values prediction technique |
CN110222089B (en) * | 2019-05-31 | 2021-05-14 | 华南理工大学 | Constraint double-clustering mining and missing value prediction method based on order preserving submatrix |
CN110928873A (en) * | 2019-10-09 | 2020-03-27 | 中国人民解放军国防科技大学 | Spatial object indexing and querying method based on improved non-dominated sorting genetic algorithm |
CN110928873B (en) * | 2019-10-09 | 2023-05-30 | 中国人民解放军国防科技大学 | Spatial object index and query method based on improved non-dominant ordering genetic algorithm |
CN110955702A (en) * | 2019-11-28 | 2020-04-03 | 江南大学 | Pattern data mining method based on improved genetic algorithm |
CN110955702B (en) * | 2019-11-28 | 2024-03-29 | 江南大学 | Improved genetic algorithm-based mode data mining method |
CN111191076A (en) * | 2019-12-30 | 2020-05-22 | 北京工业大学 | Movie recommendation method based on prospect theory and multi-target evolution |
CN111143685A (en) * | 2019-12-30 | 2020-05-12 | 第四范式(北京)技术有限公司 | Recommendation system construction method and device |
CN111143685B (en) * | 2019-12-30 | 2024-01-26 | 第四范式(北京)技术有限公司 | Commodity recommendation method and device |
CN112765398A (en) * | 2021-01-04 | 2021-05-07 | 珠海格力电器股份有限公司 | Information recommendation method and device and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN106844637B (en) | 2020-01-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106844637A (en) | Method is recommended based on the film for just giving cluster to prune improvement multi-objective genetic algorithm | |
CN106021364B (en) | Foundation, image searching method and the device of picture searching dependency prediction model | |
CN103678672B (en) | Method for recommending information | |
CN107256241A (en) | The film recommendation method for improving multi-objective genetic algorithm is replaced based on grid and difference | |
CN103377250B (en) | Top k based on neighborhood recommend method | |
Li et al. | Using multidimensional clustering based collaborative filtering approach improving recommendation diversity | |
CN102841946B (en) | Commodity data retrieval ordering and Method of Commodity Recommendation and system | |
CN107220365A (en) | Accurate commending system and method based on collaborative filtering and correlation rule parallel processing | |
CN108647996A (en) | A kind of personalized recommendation method and system based on Spark | |
CN104794367B (en) | Medical treatment resource scoring based on hidden semantic model is with recommending method | |
CN106997553A (en) | A kind of method for digging of the grouping of commodities pattern based on multiple-objection optimization | |
CN106846029B (en) | Collaborative filtering recommendation algorithm based on genetic algorithm and novel similarity calculation strategy | |
Moslehi et al. | A genetic algorithm-based framework for mining quantitative association rules without specifying minimum support and minimum confidence | |
Guo et al. | Dynamic feature generation and selection on heterogeneous graph for music recommendation | |
He et al. | SRSH: a social recommender system based on Hadoop | |
CN108681570A (en) | A kind of individualized webpage recommending method based on multi-objective Evolutionary Algorithm | |
CN103186650A (en) | Searching method and device | |
CN107133321A (en) | The analysis method and analytical equipment of the search attribute of the page | |
CN101582085A (en) | Set option method based on distributed information retrieval system | |
Wu et al. | Modeling uncertainty driven curiosity for social recommendation | |
CN105653686A (en) | Domain name network address activeness statistics method and system | |
Xu et al. | Discover dependency pattern among attributes by using a new type of nonlinear multiregression | |
Chen et al. | A new approach to the group ranking problem: finding consensus ordered segments from users’ preference data | |
Daghouri et al. | Information system performance evaluation and optimization using AHP and TOPSIS: Construction industry case | |
CN106991432A (en) | The feature selection approach of forest optimized algorithm based on scoring |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |