CN107545471B - Big data intelligent recommendation method based on Gaussian mixture - Google Patents
Big data intelligent recommendation method based on Gaussian mixture Download PDFInfo
- Publication number
- CN107545471B CN107545471B CN201710844205.1A CN201710844205A CN107545471B CN 107545471 B CN107545471 B CN 107545471B CN 201710844205 A CN201710844205 A CN 201710844205A CN 107545471 B CN107545471 B CN 107545471B
- Authority
- CN
- China
- Prior art keywords
- user
- item
- model
- particle
- 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.)
- Active
Links
- 239000000203 mixture Substances 0.000 title claims abstract description 37
- 238000000034 method Methods 0.000 title claims abstract description 14
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 60
- 238000001914 filtration Methods 0.000 claims abstract description 19
- 239000002245 particle Substances 0.000 claims description 34
- 238000004364 calculation method Methods 0.000 claims description 10
- 238000005457 optimization Methods 0.000 claims description 7
- 230000006835 compression Effects 0.000 claims description 3
- 238000007906 compression Methods 0.000 claims description 3
- 230000001133 acceleration Effects 0.000 claims description 2
- 238000009795 derivation Methods 0.000 claims description 2
- 238000003064 k means clustering Methods 0.000 claims description 2
- 238000012163 sequencing technique Methods 0.000 claims description 2
- 238000007476 Maximum Likelihood Methods 0.000 abstract description 4
- 230000000694 effects Effects 0.000 abstract description 3
- 238000011160 research Methods 0.000 abstract description 3
- 230000006870 function Effects 0.000 description 12
- 238000010586 diagram Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000019771 cognition Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000012886 linear function Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Landscapes
- Other Investigation Or Analysis Of Materials By Electrical Means (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses an intelligent recommendation method based on Gaussian mixture, belonging to the field of intelligent recommendation of big data; the research and application of the collaborative filtering recommendation algorithm GMM-TCF based on the Gaussian mixture model mainly comprise the generation of joint probability of users and projects, how to define a maximum likelihood function in the big data recommendation model, how to initialize and optimize Gaussian mixture parameters in the big data recommendation model, and how to linearly combine the user interest degree model based on the Gaussian mixture with the recommendation model based on the projects. On one hand, the user belongs to a certain cluster and is converted into a plurality of clusters from the beginning, so that the interest of the user is greatly reflected; on the other hand, by adding the project time factors, the similarity between projects is improved, so that a project-based prediction model is established, the recommendation effect of the algorithm is better, and the recommendation precision is higher.
Description
Technical Field
The invention relates to an intelligent recommendation method based on Gaussian mixture, belongs to the field of intelligent recommendation of big data, and relates to combination and use of a big data recommendation model and a group intelligent optimization algorithm.
Background
With the development of electronic commerce, the types and the number of commodities provided by merchants are increased sharply. The user with definite requirement can directly search for the commodity to be purchased through searching. However, in practice, user demand is often subject to uncertainty and ambiguity. According to amazon statistics, only 16% of customers who shop on their web sites have a clear buying intention. If the merchant can actively recommend the commodities meeting the fuzzy requirements of the user to the user from massive commodities, the potential requirements are expected to be converted into actual requirements, the sales volume of the e-commerce website can be increased, and the loyalty of the user to the website can be improved. Under the background, an individualized recommendation system capable of recommending commodities with pertinence according to user characteristics is produced and widely applied. Many web sites, including Amazon, eBay, YouTube and Google, deploy different forms of recommendation systems and generate significant business profits. Statistically, the 2006 recommendation system improves Amazon sales by 30%. The recommendation system has important application value, not only becomes one of challenging research subjects in the computer field, but also attracts researchers from numerous fields such as mathematics, physics, cognition, artificial intelligence, management, marketing and the like.
Therefore, how to effectively utilize the data mining and machine learning methods to recommend valuable personalized information to users becomes a problem with great practical significance. Collaborative filtering is an effective technique capable of generating personalized recommendations, and is widely applied to various recommendation systems, and the basic task is to match users according to similar preferences so as to recommend items that the users may like. Collaborative filtering algorithms can be generally classified into memory-based and model-based. The collaborative filtering based on the memory can be divided into a user-based filtering and a project-based filtering. The former is to calculate the similarity between users to obtain the nearest neighbor similar to the interest preference of the target user, and then to predict and recommend the target user based on the nearest neighbor. The method has no learning capability on original data, has no good expansibility, and has a cold start problem. Whereas item-based recommendations are based on calculating the similarity between items from a commodity perspective, generating recommendations by nearest neighbor searching of the items. Generally speaking, the similarity between items is more stable than the similarity preferred by the user, and the similarity is calculated off line, so that the nearest neighbor search of the user with large magnitude order on line is avoided, and the on-line calculation amount is greatly reduced. Different from the algorithm based on the memory, the algorithm based on the model does not directly calculate and recommend through the user-item raw scoring data, but learns and models the data by using a machine learning method, and predicts the habit preference of the user to the unscored items through the model. Commonly used models include bayesian, Singular Value Decomposition (SVD), Latent Semantic Analysis (LSA), Probabilistic Latent Semantic Analysis (PLSA), and the like.
However, model-based algorithms necessarily consider the following issues: first, the degree of the user's rating of the project is influenced by various factors, including the user's interest level in the subject/content of the project, the user's rating habits, the quality of the project itself, etc. Secondly, the single latent variable hardly reflects the relationships between users, between items and between users and between items at the same time, and the users need to cluster according to similar interests and preferences and the items according to similar subjects and contents. Third, the common "hard" clustering approach, where each user or item belongs to only a single class, makes it difficult to describe the multiple interests that the same user may have and the multiple different attributes of the same item. It is reasonable to perform "soft" clustering, where each user or item belongs to multiple classes according to probability, and the magnitude of the probability value of a user in a certain class reflects the preference degree of the user for that class, and the items are also possible.
Based on the above thought, the invention provides a new Collaborative Filtering algorithm GMM-TCF (Gaussian Mixture Model-Time Collaborative Filtering) recommendation based on a Gaussian Mixture Model, which combines a Gaussian Model and an improved project-based Collaborative Filtering algorithm. The concept of a multi-interest group is led out through the latent variable, the user model is learned to obtain the probability that the user belongs to each interest group and the interest preference of the group to the item. On the other hand, in calculating the recommendation values based on the items, considering that the closer the appearance times of the two items are, the higher the similarity between the items is, the calculation of the similarity introduces a time factor of item similarity. Meanwhile, the algorithm optimizes parameters in the Gaussian mixture model by using a CFPSO (compression Factor Particle Swarm optimization) algorithm, and effectively combines recommendation values based on user interests and items in a linear weighting mode, so that the limitation of interest groups caused by recommendation based on a user interest degree model is made up, the cold start problem of a recommendation system and the limitation of single interest recommendation are solved, and the recommendation accuracy is effectively improved. In addition, the algorithm is modeled using joint probabilities of user interests and items. Meanwhile, the parameters in the Gaussian mixture model are reasonably selected according to initialization, and when the parameters are solved by using an EM (expectation maximization) algorithm in the Gaussian mixture model, a proper likelihood function is adopted.
The main recommendation mechanism of the Gaussian mixture model collaborative filtering recommendation algorithm GMM-TCF is as follows: in the selection of the recommendation algorithm, the users are recommended not only by selecting the similarity among the users, but also by combining the similarity among the items. The recommendation method not only makes up the limitation of a single cluster to which the user belongs, but also improves the similarity calculation formula by using the time factors among the items, thereby effectively improving the recommendation precision of the algorithm.
Disclosure of Invention
The method mainly comprises the research and application of a collaborative filtering recommendation algorithm GMM-TCF based on a Gaussian mixture model, and mainly comprises the generation of joint probability of users and items, the definition of a maximum likelihood function in the big data recommendation model, the initialization and optimization of Gaussian mixture parameters in the big data recommendation model, and the linear combination of a user interest degree model based on Gaussian mixture and a recommendation model based on items. The technical scheme adopted by the method is as follows: 1) performing soft clustering on the users by using a Gaussian mixture model, so as to provide a similarity prediction model based on user interests, and calculating the prediction score of the model; 2) improving a similarity calculation formula by introducing a project time factor, and calculating a prediction score based on a project model; 3) and linearly combining the prediction scores of the two models to serve as a final project prediction score, and recommending the final project prediction score to a target user.
The invention adopts the technical scheme that a big data intelligent recommendation method based on Gaussian mixture is realized by the following steps:
(1) a similarity prediction model based on user interests. Firstly, the joint probability of users and items is built through a PLSA model, and based on the probability form, a proper likelihood function, namely a big data intelligent recommendation model, is built. And then, optimizing the EM algorithm through the CFPSO algorithm so as to solve the parameters in the big data intelligent recommendation model. And finally, solving the prediction scores of the target users on the items by using a big data intelligent recommendation model.
a. A joint probability of the user and the project is constructed. For each three-dimensional directionThe quantity < u, i, v >, where u, i, v are user u, item i, and the user u's score for item i, respectively. Introducing latent variable Z ═ Z1,z2,…,zkIn which z isi(1. ltoreq. i.ltoreq.k) as different cluster groups, using P (z)k| u) indicates that user u belongs to zkProbability of population and
P(v|i,zk) Representing a population zkThe probability of v is scored for item i. Then the joint probability of the user and the item is:
b. and (5) constructing a likelihood function, namely a big data intelligent recommendation model. Assuming that the conditional probability P (v | i, z) of the group z scoring the item i by v conforms to a gaussian distribution, there is P (v | i, z) ═ N (μi,z,σi,z)＝P(v；μi,z,σi,z) In which μi,z,σi,zThe mean and variance of the scores of the group Z for the item i, respectively, then the joint probability of the user and the item is a probability model satisfying Gaussian mixture:
then the log-likelihood function is obtained as:
c. and (4) initializing and selecting parameters. For parameter P (z | u), μi,z，
And initializing the parameters by adopting a K-means clustering algorithm. And simultaneously, optimizing the initialized parameters by using a CFPSO algorithm. Let the number of particle groups be n, and the position of any particle i be Xi＝(x1,x2,…xn) The fitness function is selected as equation (1). The steps for optimizing the EM algorithm based on CFPSO are as follows:
the first step is as follows: a population of particles is initialized. Initializing initial position and velocity of each particle in the particle population given the number of the particle population, the position and velocity formula being as follows:
wherein,
representing the velocity of the ith particle in d-dimensional space at time t +1, k being the compression factor, ω being the inertial weight, c1，c2Is an acceleration constant, r1And r2Is a random number, and is a random number,
and
respectively representing the individual optimal and global optimal values of the d-dimensional space of particle i at time t,
representing the position of the ith particle in d-dimensional space at time t + 1.
The second step is that: calculating the fitness value of each particle in the particle swarm, and updating the current individual optimal position p of the particlebestAnd a population optimal position g of the particle groupbest。
The third step: g for a particle populationbestIt is updated according to the EM algorithm. Comparing the fitness values before and after updating, if g is updatedbestIf the value of fitness function is increased, the optimal position g of population is updatedbestOtherwise, the information is not updated.
The fourth step: for the updated population optimal value gbestChecking is carried out, if the requirements are met, the CFPSO is endedAlgorithm and get gbestThe attribute information of the value is used as an initial parameter of the EM algorithm; otherwise, go to the fifth step.
The fifth step: and (3) updating the speed and position parameters of the particle individuals in the particle swarm according to the formulas (2) and (3) and continuing to execute the second step. The EM algorithm comprises the following steps:
e, step E: from each score vector < u, i, v >, a posterior probability P (Z | u, v, i) is computed for each potential variable Z ∈ Z, as follows:
and M: obtaining P (z | u), mu according to the posterior probability obtained by the calculation in the step E and the partial derivation of the likelihood function by combining the Lagrange optimization extremumi,z，
The values of (a) are as follows:
optimizing initialization parameters of the EM algorithm according to the CFPSO, alternately executing the step E and the step M until convergence, and solving parameters P (z | u), mui,z，
As a parameter set for big data recommendation model.
d. And predicting the scores by using the user interest similarity model. And (3) constructing a user interest similarity model based on Gaussian mixture through the parameter set in the step M, thereby calculating the prediction score of the user u on the item i, wherein the specific formula is as follows:
(2) a project-based predictive model. The more users that score items according to their score, the higher the similarity between items. Meanwhile, as the time of appearance of the same kind of articles is more similar, the similarity between the articles is higher, so that a project time factor is introduced and defined as follows:
wherein, tiAnd tjThe times at which item i and item j occur. Define the similarity of items i and j as sim (i, j):
where U (i) and U (j) are the set of users that score items i and j, respectively, ru,iAnd ru,jRepresenting the user u's score for item i and item j,
and
represents the average rating of all users for item i and item j, with θ being a non-negative number. Then the user's prediction score for the project is defined as follows:
and S (i) is a field set of the item i, wherein the mode of selecting whether the item is added into the neighborhood is to judge whether the similarity of the two items is greater than a certain threshold value, and the calculation mode reduces the comparison sequencing between the items and saves the operation time.
(3) And linear weighted prediction. Combining the prediction scores of the user interest similar prediction model and the project-based prediction model in a linear weighting mode, thereby calculating the final prediction score of the user on the project, wherein the formula is calculated as follows:
rateu,i＝α×rate_uHMMu,i+β×rate_Itemu,i，0＜α＜1,α+β＝1 (12)
in conclusion, the prediction scores of the user on the project in the collaborative filtering algorithm are improved, a prediction model based on the user interest similarity is established through Gaussian mixture, CFPSO and EM algorithms, and a prediction model based on the project is established through adding time factors, so that the prediction scores of the Gaussian mixture, the CFPSO and the EM algorithms are linearly combined to serve as the final prediction score of the user on the project.
By the method, on one hand, the initial user belongs to a certain cluster and is converted into a plurality of clusters, so that the interest of the user is greatly reflected; on the other hand, by adding the project time factors, the similarity between projects is improved, so that a project-based prediction model is established, the recommendation effect of the algorithm is better, and the recommendation precision is higher.
Drawings
FIG. 1: overall framework diagram of collaborative filtering algorithm based on Gaussian mixture model
FIG. 2: gaussian mixture model-based user interest prediction model flow chart
FIG. 3: project-based predictive recommendation model flow chart
FIG. 4: MAE graph under different Item-CF neighborhoods
Detailed Description
The invention is explained and illustrated below with reference to the accompanying drawings:
in order to make the objects, technical solutions and features of the present invention more apparent, the present invention is further described in detail below with reference to the accompanying drawings. The overall framework diagram of the collaborative filtering algorithm based on the Gaussian mixture model is shown in FIG. 1.
The individual steps are illustrated below:
(1) a user interest degree prediction model based on a Gaussian mixture model is provided, so that the user interest is greatly reflected and is not limited in a certain cluster.
(2) A project-based prediction model is provided, and the influence of time factors of the occurrence sequence of projects on recommendation scores is researched.
(3) Compared with the model, the model recommendation accuracy is higher.
The experimental environment was as follows:
the invention verifies the actual effect of the collaborative filtering algorithm based on the Gaussian mixture model, which is provided by the invention, through experiments, the experimental environment is a win7(64 bit) host, an 8G memory and a 1T hard disk, a Netflix data set is adopted, and the recommended MAE value is verified in a cross verification mode.
Firstly, an algorithm establishes a user interest prediction model based on a Gaussian mixture model. And analyzing the interest of the user by adopting PLSA probability latent semantics, and introducing a latent variable Z to introduce multiple interests of the user so as to calculate the joint probability of the user and the item. On the basis, the scores of the user group Z on the items are assumed to be subjected to Gaussian distribution, so that a maximum likelihood function is constructed, the logarithm of the maximum likelihood function is solved, and a general solution of the parameters is solved. Aiming at the initialization of the parameters, the model sets the initial values of the models by adopting a K-mean clustering algorithm, continuously optimizes the initial value parameters by a CFPSO algorithm until the fitness function of the CFPSO algorithm converges or the algorithm reaches the specified iteration times, ends the optimization process, outputs the final value of the EM algorithm parameters, thereby constructing the user interest model and predicting the scoring rate _ uHMM of the user to the projectu,iThe model building flow chart is shown in fig. 2.
The algorithm then builds a project-based predictive recommendation model. By introducing the project time factor, a project time factor function is constructed, so that a similarity calculation mode based on project collaborative filtering recommendation is improved, and the similarity between projects is improved. The project forecast of the user is calculated through a forecast scoring formula by constructing a project-based forecast recommendation modelMeasuring and scoring rate _ Itemu,iThe flow chart of the predictive recommendation model construction is shown in fig. 3.
Finally, the algorithm relies on a user interestingness model based on a Gaussian mixture model and a project-based prediction recommendation model, a collaborative filtering recommendation model based on Gaussian mixture is constructed, namely the algorithm obtains the prediction scores of the project of the user according to the two models, the prediction scores are connected through a linear function, a prediction scoring formula is improved, and then the final prediction scoring rate of the project of the user is calculatedu,i。
From the comparison between the algorithm model provided by the invention and other algorithms in fig. 4, it can be clearly seen that the MAE value of the algorithm GMM-TCF provided herein is significantly smaller than that of other algorithms, and as the number of items in the project neighborhood increases, the MAE value shows a decreasing trend.
By combining the experiments, the collaborative filtering recommendation model based on the Gaussian mixture is used for recommending the user, on one hand, the multi-interest of the user can be effectively utilized, and meanwhile, the selection of the Gaussian mixture parameter value is optimized. On the other hand, the similarity between projects can be improved by introducing the project time factor, the running time of the algorithm can be effectively reduced by a threshold value calculation neighborhood mode, and the score of the project is more accurate by a user by adjusting a prediction score linear combination mode, so that the recommendation precision and recommendation efficiency of the algorithm are improved.
Claims (1)
1. A big data intelligent recommendation method based on Gaussian mixture is characterized by comprising the following steps: the implementation steps of the method are as follows,
(1) a similarity prediction model based on user interests; firstly, constructing joint probability of users and items through a PLSA model, and constructing a proper likelihood function, namely a big data intelligent recommendation model based on the probability form; then, optimizing an EM algorithm through CFPSO (particle swarm optimization) so as to solve parameters in the big data intelligent recommendation model; finally, solving the prediction scores of the target users on the items by using a big data intelligent recommendation model;
a. constructing joint probabilities of users and projects; for each three-dimensional vector<u,ii,v>Wherein u, ii, v are user u, project ii, and the rating of project ii by user u, respectively; introducing latent variable Z ═ Z1,z2,…zi,…zkIn which z isiFor different clustering groups, i is more than or equal to 1 and less than or equal to k, P (z) is usedk| u) indicates that user u belongs to zkProbability of population and
P(v|ii,zi) Representing a population ziScoring the probability of v for item ii; then the joint probability of the user and the item is:
b. constructing a likelihood function, namely a big data intelligent recommendation model; assuming that the conditional probability P (v | ii, z) for the population z to score v for the item ii conforms to a gaussian distribution, with P (v | ii, z) ═ N (μi,z,σi,z)＝P(v；μi,z,σi,z) In which μi,z,σi,zThe mean and variance of the scores of the group Z for the item i, respectively, then the joint probability of the user and the item is a probability model satisfying Gaussian mixture:
then the log-likelihood function is obtained as:
c. initializing and selecting parameters; for parameter P (z | u), μi,z，
Initializing the parameters by adopting a K-means clustering algorithm; meanwhile, optimizing the initialized parameters by using a CFPSO algorithm; the number of particle groups is set to n, and the position of any particle l is represented by Xl＝(x1,x2,…xn) Selecting the formula (1) as a fitness function; thenThe steps for optimizing the EM algorithm based on the CFPSO are as follows:
the first step is as follows: initializing a particle population; initializing initial position and velocity of each particle in the particle population given the number of the particle population, the position and velocity formula being as follows:
wherein,
representing the velocity of the l-th particle in d-dimensional space at time t +1, ε being a compression factor, ω being an inertial weight, c1，c2Is an acceleration constant, r1And r2Is a random number, and is a random number,
and
respectively representing the individual optimal and global optimal values of the d-dimensional space of the particle i at time t,
represents the position of the l-th particle in d-dimensional space at time t + 1;
the second step is that: calculating the fitness value of each particle in the particle swarm, and updating the current individual optimal position p of the particlebestAnd a population optimal position g of the particle groupbest；
The third step: g for a particle populationbestUpdating the EM algorithm; comparing the fitness values before and after updating, if g is updatedbestIf the value of fitness function is increased, the optimal position g of population is updatedbestOtherwise, not updating;
the fourth step: for the updated population optimal value gbestChecking is carried out, if the requirements are met, the CFPSO algorithm is ended, and g is obtainedbestThe attribute information of the value is used as an initial parameter of the EM algorithm; otherwise, turning to the fifth step;
the fifth step: updating the speed and position parameters of the particle individuals in the particle swarm according to the formulas (2) and (3) and transferring to the second step for continuous execution; the EM algorithm comprises the following steps:
e, step E: from each scoring vector < u, i, v >, a posterior probability P (Z | u, v, i) for each potential variable Z ∈ Z is computed as follows:
and M: obtaining P (z | u), mu according to the posterior probability obtained by the calculation in the step E and the partial derivation of the likelihood function by combining the Lagrange optimization extremumi,z，
The values of (a) are as follows:
optimizing initialization parameters of the EM algorithm according to the CFPSO, alternately executing the step E and the step M until convergence, and solving parameters P (z | u), mui,z，
A parameter set as a big data recommendation model;
d. predicting and scoring by a user interest similarity model; and (3) constructing a user interest similarity model based on Gaussian mixture through the parameter set in the step M, thereby calculating the prediction score of the user u on the item i, wherein the specific formula is as follows:
(2) a project-based predictive model; the more users that score the projects, the higher the similarity between the projects; meanwhile, as the time of appearance of the same kind of articles is more similar, the similarity between the articles is higher, so that a project time factor is introduced and defined as follows:
wherein, tiAnd tjTime of occurrence for item i and item j; define the similarity of items i and j as sim (i, j):
where U (i) and U (j) are the set of users that score items i and j, respectively, ru,iAnd ru,jRepresenting the user u's score for item i and item j,
and
representing the average scores of all users on the item i and the item j, and theta is a non-negative number; then the user's prediction score for the project is defined as follows:
s (i) is a neighborhood set of the item i, wherein the mode of selecting whether the item is added into the neighborhood is to judge whether the similarity of the two items is greater than a certain threshold value, and the calculation mode reduces the comparison sequencing between the items and saves the operation time;
(3) linear weighted prediction; combining the prediction scores of the user interest similar prediction model and the project-based prediction model in a linear weighting mode, thereby calculating the final prediction score of the user on the project, wherein the formula is calculated as follows:
rateu,i＝α×rate_uHMMu,i+β×rate_Itemu,i，0＜α＜1,α+β＝1 (12)
in conclusion, the prediction scores of the user on the project in the collaborative filtering algorithm are improved, a prediction model based on the user interest similarity is established through Gaussian mixture, CFPSO and EM algorithms, and a prediction model based on the project is established through adding time factors, so that the prediction scores of the Gaussian mixture, the CFPSO and the EM algorithms are linearly combined to serve as the final prediction score of the user on the project.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710844205.1A CN107545471B (en) | 2017-09-19 | 2017-09-19 | Big data intelligent recommendation method based on Gaussian mixture |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710844205.1A CN107545471B (en) | 2017-09-19 | 2017-09-19 | Big data intelligent recommendation method based on Gaussian mixture |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107545471A CN107545471A (en) | 2018-01-05 |
| CN107545471B true CN107545471B (en) | 2021-06-11 |
Family
ID=60964064
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710844205.1A Active CN107545471B (en) | 2017-09-19 | 2017-09-19 | Big data intelligent recommendation method based on Gaussian mixture |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107545471B (en) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110110214A (en) * | 2018-01-25 | 2019-08-09 | 重庆邮电大学 | Dynamic recommendation and plus method for de-noising based on bidirectional weighting value and user behavior |
| CN108733784B (en) * | 2018-05-09 | 2020-12-29 | 深圳市领点科技有限公司 | Teaching courseware recommendation method, device and equipment |
| CN109034968B (en) * | 2018-07-18 | 2021-11-05 | 江苏中润普达信息技术有限公司 | Artwork recommendation method based on particle swarm algorithm |
| CN109190029B (en) * | 2018-08-22 | 2021-09-28 | 中食安泓(广东)健康产业有限公司 | Working method of cloud intelligent information pushing platform |
| CN109347924A (en) * | 2018-09-20 | 2019-02-15 | 西北大学 | A kind of recommended method based on intelligent perception |
| CN110119974B (en) * | 2019-05-17 | 2022-07-05 | 武汉众诚华鑫科技有限公司 | delta-GMM clustering algorithm-based intelligent pushing method for mobile packages |
| CN111127139B (en) * | 2019-12-06 | 2023-06-27 | 成都理工大学 | Mixed recommendation algorithm based on ProbS and HeatS calculation mode improvement |
| CN112508050A (en) * | 2020-11-06 | 2021-03-16 | 重庆恢恢信息技术有限公司 | Construction engineering construction planning working method based on mass data |
| CN114862486B (en) * | 2021-02-04 | 2024-11-01 | 西安电子科技大学青岛计算技术研究院 | Virtual fitting room implementation method based on user model and collaborative filtering algorithm |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103870604B (en) * | 2014-04-04 | 2017-03-29 | 北京航空航天大学 | Method and apparatus is recommended in tourism |
| SG10201406215YA (en) * | 2014-09-30 | 2016-04-28 | Mentorica Technology Pte Ltd | Systems and methods for automated data analysis and customer relationship management |
| CN106708953A (en) * | 2016-11-28 | 2017-05-24 | 西安电子科技大学 | Discrete particle swarm optimization based local community detection collaborative filtering recommendation method |
-
2017
- 2017-09-19 CN CN201710844205.1A patent/CN107545471B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN107545471A (en) | 2018-01-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107545471B (en) | Big data intelligent recommendation method based on Gaussian mixture | |
| Cai et al. | A hybrid recommendation system with many-objective evolutionary algorithm | |
| Logesh et al. | A hybrid quantum-induced swarm intelligence clustering for the urban trip recommendation in smart city | |
| Nilashi et al. | A recommender system based on collaborative filtering using ontology and dimensionality reduction techniques | |
| Cheng et al. | Learning to recommend accurate and diverse items | |
| Desrosiers et al. | A comprehensive survey of neighborhood-based recommendation methods | |
| Ma et al. | Learning to recommend with trust and distrust relationships | |
| Hu et al. | Movie collaborative filtering with multiplex implicit feedbacks | |
| Yu et al. | Joint user knowledge and matrix factorization for recommender systems | |
| Aghdam et al. | A novel non-negative matrix factorization method for recommender systems | |
| Raghuwanshi et al. | Accelerated singular value decomposition (asvd) using momentum based gradient descent optimization | |
| Sivaramakrishnan et al. | An effective user clustering-based collaborative filtering recommender system with grey wolf optimisation | |
| Mao et al. | Hybrid ecommerce recommendation model incorporating product taxonomy and folksonomy | |
| Nguyen et al. | Tourism recommender system based on cognitive similarity between cross-cultural users | |
| Fareed et al. | A collaborative filtering recommendation framework utilizing social networks | |
| Yin et al. | A survey of learning-based methods for cold-start, social recommendation, and data sparsity in e-commerce recommendation systems | |
| Drakopoulos et al. | Self organizing maps for cultural content delivery | |
| CN108984551A (en) | A kind of recommended method and system based on the multi-class soft cluster of joint | |
| Li et al. | Task recommendation with developer social network in software crowdsourcing | |
| Pang et al. | Efficient deep reinforcement learning-enabled recommendation | |
| Ifada et al. | Do-rank: DCG optimization for learning-to-rank in tag-based item recommendation systems | |
| Xu et al. | Exploiting interactions of review text, hidden user communities and item groups, and time for collaborative filtering | |
| Lu et al. | Artificial immune network with feature selection for bank term deposit recommendation | |
| Deenadayalan et al. | User Feature Similarity Supported Collaborative Filtering for Page Recommendation Using Hybrid Shuffled Frog Leaping Algorithm. | |
| Wang et al. | Multi‐feedback Pairwise Ranking via Adversarial Training for Recommender |
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 | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20240410 Address after: Room 81075, 4th Floor, Building 3, No. 116 Xinhua East Street, Tongzhou District, Beijing, 101100 Patentee after: Beijing Linear Overlay Technology Co.,Ltd. Country or region after: China Address before: 100124 No. 100 Chaoyang District Ping Tian Park, Beijing Patentee before: Beijing University of Technology Country or region before: China |