CN108268464A - A kind of personalized recommendation method and device returned based on collaborative filtering and logistic - Google Patents
A kind of personalized recommendation method and device returned based on collaborative filtering and logistic Download PDFInfo
- Publication number
- CN108268464A CN108268464A CN201611254081.3A CN201611254081A CN108268464A CN 108268464 A CN108268464 A CN 108268464A CN 201611254081 A CN201611254081 A CN 201611254081A CN 108268464 A CN108268464 A CN 108268464A
- Authority
- CN
- China
- Prior art keywords
- user
- unit
- logistic regression
- collaborative filtering
- personalized recommendation
- 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
- 238000001914 filtration Methods 0.000 title claims abstract description 31
- 238000000034 method Methods 0.000 title claims abstract description 29
- 239000011159 matrix material Substances 0.000 claims abstract description 30
- 238000000605 extraction Methods 0.000 claims abstract description 12
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 9
- 238000007477 logistic regression Methods 0.000 claims description 37
- 230000006399 behavior Effects 0.000 claims description 13
- 238000004422 calculation algorithm Methods 0.000 claims description 12
- 238000002790 cross-validation Methods 0.000 claims description 12
- 238000004364 calculation method Methods 0.000 claims description 7
- 238000012163 sequencing technique Methods 0.000 claims description 5
- 238000007476 Maximum Likelihood Methods 0.000 claims description 4
- 238000012821 model calculation Methods 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 abstract description 2
- 206010068052 Mosaicism Diseases 0.000 abstract 1
- HUTDUHSNJYTCAR-UHFFFAOYSA-N ancymidol Chemical compound C1=CC(OC)=CC=C1C(O)(C=1C=NC=NC=1)C1CC1 HUTDUHSNJYTCAR-UHFFFAOYSA-N 0.000 abstract 1
- 210000003765 sex chromosome Anatomy 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 4
- 230000009467 reduction Effects 0.000 description 2
- 230000008672 reprogramming Effects 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 229960001948 caffeine Drugs 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000010790 dilution Methods 0.000 description 1
- 239000012895 dilution Substances 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000004091 panning Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 239000000243 solution Substances 0.000 description 1
- RYYVLZVUVIJVGH-UHFFFAOYSA-N trimethylxanthine Natural products CN1C(=O)N(C)C(=O)C2=C1N=CN2C RYYVLZVUVIJVGH-UHFFFAOYSA-N 0.000 description 1
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
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Physics & Mathematics (AREA)
- Economics (AREA)
- General Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Marketing (AREA)
- Development Economics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention discloses a kind of personalized recommendation device returned based on collaborative filtering and logistic, and including input unit, extraction unit, score unit, predicting unit, computing unit, selection unit, pretreatment unit, model computing unit compare and recommendation unit and output unit.The similarity between user is calculated first with the historical information of user, using the user high with target user's similarity as neighbours, using the commodity that neighbours like as commodity to be recommended, is ranked up according to purchase possibility, is recommended in this order.The present invention is based on arest neighbors collaborations to have found the possible interested commodity of user, singular value decomposition is used in the process, solves the sparse sex chromosome mosaicism of rating matrix, and Recommendations are treated according to purchase possibility and are ranked up, this is more accurate for sorting with respect to user interest, can preferably realize precision marketing.
Description
Technical Field
The invention relates to the field of personalized recommendation, in particular to a personalized recommendation method and device based on collaborative filtering and logistic regression.
Background
With the rapid development of the internet, people can not leave the internet more and more in clothes and eating and residents, online shopping becomes a trend, and personalized recommendation technology becomes more and more important. The online stores have various commodities and large amount of information, and how to realize accurate marketing is very important for sellers and buyers. And a targeted marketing strategy is adopted to replace the sea panning, so that the user can be helped to quickly find the commodities which are really needed by the user in a large number of products. User-based collaborative filtering is a very successful and widely applied recommendation technique, which finds similar (interested) users of a specified user in a user group by analyzing user interests, integrates evaluations of the similar users on certain information, forms a system and predicts the preference degree of the specified user on the information.
However, most recommendation algorithms are recommendations based on user interests, and the recommendation effect is influenced to a certain extent if the possibility of purchase of the user is not considered. If a device is available, a model is provided to predict the rate of purchase of the products that may be of interest to the buyer, and the recommended products are sorted according to probability, such a recommendation manner must better meet the needs of the user.
In view of the above-mentioned drawbacks, the inventors of the present invention have finally obtained the present invention through a long period of research and practice.
Disclosure of Invention
In order to solve the technical defects, the invention adopts a technical scheme that a personalized recommendation device based on collaborative filtering and logistic regression is provided, and the personalized recommendation device comprises:
an input unit: it is used to obtain user information;
an extraction unit: it is used to perform feature extraction on the information;
a scoring unit: which is used for scoring the commodities purchased by the user;
a prediction unit: which is used to predict the scores of unscored goods;
a calculation unit: it is used to calculate the similarity between users;
a selecting unit: it is used to select neighbors and goods to be recommended;
a pretreatment unit: it is used to transform categorical variables and normalize continuous variables;
a model calculation unit: it uses lasso estimation of logistic regression model to get probability estimation model;
a comparison and recommendation unit: the system is used for sequencing predicted values and recommending commodities;
an output unit: which is used to give recommendations for goods to the user.
Preferably, the extraction unit includes a collaborative filtering module and a logistic regression module, and extracts different features respectively.
A personalized recommendation method based on collaborative filtering and logistic regression is characterized by comprising the following steps:
step S1: extracting the characteristics of the collaborative filtering part;
step S2: constructing a scoring matrix;
step S3: singular value decomposition;
step S4: calculating the similarity;
step S5: selecting neighbors and determining commodities to be recommended;
step S6: extracting the characteristics of a logistic regression module;
step S7: preprocessing variables;
step S8: lasso estimation of logistic regression model;
step S9: and obtaining a predicted value, and recommending the commodities according to the sequence of the predicted value.
Preferably, theIn step S2, the different behaviors of the user are quantified by using a scoring matrix Rm×n(m.gtoreq.n) represents the user's score for the good, Ru,iIndicates the rating of the item i by the user u,representing the average rating of the item that user u has purchased.
Preferably, in the step S3, R is transformed by singular valuem×nAnd decomposing the matrix into three matrixes of U, S and V, and satisfying that R is U multiplied by S multiplied by V'.
Preferably, in step S3, the eigenvalues of R' R are calculated, sorted from large to small and set aside to obtain S; solving the eigenvector of R' R, sorting eigenvalues from big to small, and orthogonalizing the eigenvector to obtain V; and solving the feature vector of RR', sorting the feature values from large to small, and orthogonalizing the feature vector to obtain U.
Preferably, in the step S3, dimension S is reduced to k order according to the principle of singular value decomposition, and U can be obtained by the same methodk，VkThe relation is changed to Rk＝Uk×Sk×Vk。
Preferably, in step S3, the score of the non-scored merchandise is calculated according to the following formula:
whereinRepresents the average rating of the items purchased by user U,to representIn the u-th row of (a),is composed ofColumn i.
Preferably, in the step S4, the adjusted cosine measure is used to calculate the similarity between the user and the target user, and the specific algorithm is as follows:
preferably, the maximum likelihood estimation method is selected to estimate the parameters of the regression equation, wherein the coefficients in the lasso-logistic regression model are given by the minima of the convex function, as follows:
where L is the log-likelihood function and λ is the harmonic parameter.
The estimates under the lasso algorithm are:
wherein λ can be determined by a generalized cross-validation method
Order to
The GCV statistic of the generalized cross validation value is
Penalty parameter for optimizing penalty function for minimizing generalized cross-validation value GCV
The estimation of the parameter β can be obtained through Lagrange multiplier algorithm to obtain a probability estimation model
Compared with the prior art, the invention has the beneficial effects that: the personalized recommendation device based on collaborative filtering and logistic regression finds out the commodities which are possibly interested by the user based on nearest neighbor collaborative filtering, is decomposed by Singular Values (SVD) in the process, solves the problem of sparsity of a scoring matrix, and ranks the commodities to be recommended according to the purchase possibility, so that the personalized recommendation device is more accurate to ranking according to the interests of the user and better realizes accurate marketing.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below.
FIG. 1 is a functional block diagram of a personalized recommender based on collaborative filtering and logistic regression;
FIG. 2 is a general flow diagram of a personalized recommendation method based on collaborative filtering and logistic regression.
Detailed Description
The above and further features and advantages of the present invention are described in more detail below with reference to the accompanying drawings.
Example 1
As shown in fig. 1, which is a functional block diagram of the present invention. The personalized recommendation device based on collaborative filtering and logistic regression comprises: the system comprises an input unit 1, an extraction unit 2, a scoring unit 3, a prediction unit 4, a calculation unit 5, a selection unit 6, a preprocessing unit 7, a model calculation unit 8, a comparison and recommendation unit 9 and an output unit 10.
The input unit 1 is used for acquiring various information and data of a user, such as: user basic information, user purchase records, and the like.
The extraction unit 2 is divided into a collaborative filtering module 21 and a logistic regression module 22. The collaborative filtering module 21 performs feature extraction on the information and data of the input unit 1, for example: and browsing, clicking, commenting, collecting and purchasing the commodity by the user. The logistic regression module 22 determines whether the target user purchased or not as a response variable, assuming that y follows the bernoulli distribution, where y-1 represents a purchase behavior and y-0 represents an unpurceful behavior. The dilution variable is X, where X is (X)1,x2,...xp-1) For example: basic information of the user, such as age, income condition, basic information of the commodity, such as category, price, and the like.
The scoring unit 3 scores the commodities according to behaviors of browsing, clicking, commenting, collecting and purchasing and the like of the user. And quantifying different behaviors of the user, and recording the scores of the user on the commodities according to the highest score without accumulating. Denote the set of users by U ═ U1,u2,...,umDenotes a set of commodities by I ═ I1,i2,...,inUsing a scoring matrix Rm×n(m.gtoreq.n) represents the user's score for the good. Wherein R isu,iIndicates the rating of the item i by the user u,representing the average rating that user u has purchased all of the items.
The prediction unit 4 performs dimension reduction, eigenvalue calculation, eigenvector calculation and the like on the scoring matrix of the scoring unit 3, and predicts the score of the unscored commodity through a formula. The specific working principle is as follows: using Singular Values (SVD) to assign the scoring matrix Rm×n(m is more than or equal to n) is decomposed into U, S and V, so that the relation is satisfied: r ═ U × S × V'. The method comprises the steps of obtaining a characteristic value of R ' R, sorting from large to small and squaring to obtain a matrix S, obtaining a characteristic vector of R ' R, sorting from large to small according to the characteristic value, orthogonalizing the characteristic vector to obtain a matrix V, obtaining a characteristic vector of a matrix RR ', sorting from large to small according to the characteristic value, orthogonalizing the characteristic vector to obtain a matrix U.
According to the principle of single-valued decomposition, S is reduced to k-th order (k < r), and Uk, Vk, which satisfies the relation: rk＝Uk×Sk×Vk. The score for an unscored good is calculated using the following formula:
whereinRepresenting the average rating of the item that user u has purchased,to representIn the u-th row of (a),to representColumn i.
The calculating unit 5 calculates the similarity between the user and the target user according to the adjusted cosine measure, and the formula is as follows:
and the selecting unit 6 selects neighbors according to the similarity obtained by the calculating unit, and selects the commodities to be recommended from the commodities with higher scores.
The processing unit 7 is used for reprogramming the classification variables into virtual variables and normalizing the continuous variables. Assuming that a variable has k classes, then (k-1) variables (k ═ 1,2,3 … …) are recoded, which are of the form: x is the number of1＝(1,0,...,0)T...xk-1＝(0,...0,1)T. The normalized formula is:
the model calculation unit 8 adopts lasso estimation of a logistic regression model, and selects a maximum likelihood estimation method to estimate the parameters of the regression equation, so as to finally obtain a probability estimation model, wherein the specific working principle is as follows:
the coefficients in the lasso-logistic regression model are given by the minima of the convex function, as follows:
where L is the log-likelihood function and λ is the harmonic parameter.
The estimates under the lasso algorithm are:
where lambda may be determined by a generalized cross-validation method,
order to
The GCV statistic of the generalized cross validation value is
Penalty parameter for optimizing penalty function for minimizing generalized cross-validation value GCV
The estimation of the parameter β can be obtained through Lagrange multiplier algorithm to obtain a probability estimation model
The comparing and recommending unit 9 predicts the purchase probability of the to-be-recommended commodity through the probability estimation model obtained by the model calculating unit 8 to obtain a predicted valueThe commodities are sorted and recommended according to the ranking sequence.
The output unit 10 outputs the recommended product obtained by the comparison and recommendation unit 9, and gives a recommendation of the product to the user.
Example 2
As shown in the figure, the general flow diagram of the personalized recommendation method based on collaborative filtering and logistic regression of the present invention is shown, and the personalized recommendation method based on collaborative filtering and logistic regression includes the following steps:
step S1: feature extraction of collaborative filtering section
Acquiring various aspects of information and data of a user, such as: user basic information, historical purchase records and the like, and browsing, clicking, commenting, collecting and purchasing behaviors of the feature selection user.
Step S2: construction of scoring matrices
And scoring the commodities according to behaviors of browsing, clicking, commenting, collecting, purchasing and the like of the user. Quantifying the different behaviors of the user, for example: and the browsing score is 2, the clicking score is 4, the collection score is 6, the purchasing score is 8, the comments score is 10, and the scores of the user on the commodities are recorded according to the highest score without accumulation. Let us assume that the set of users is denoted by U, U ═ U1,u2,...,umThe set of goods is denoted by I ═ I1,i2,...,inUsing a scoring matrix Rm×n(m.gtoreq.n) to represent the user's rating of the good. Wherein R isu,iIndicates the rating of the item i by the user u,representing the average rating that user u has purchased all of the items.
Step S3: singular value decomposition to solve the sparsity problem of the scoring matrix
Singular Value (SVD) decomposition is used to predict the user's score for unscored merchandise. Singular Value (SVD) enables the scoring matrix Rm×n(m is more than or equal to n) is decomposed into three matrixes of U, S and V, and the relation R ═ UxS xV'. Where U is an m × n orthogonal matrix, V is an r × n matrix, and S is an r × r matrix. The elements on the non-diagonal of the matrix S are all 0, and the elements on the diagonal satisfy: sigma1≥σ2≥…≥σn≧ 0, called singular value, which indicates how close a given matrix is to matrices of lower rank than it. According to the principle of single-value decomposition, S is reduced to k-order (k < r), and U can be obtained by the same methodk、VkWhich satisfies the relation Rk＝Uk×Sk×Vk。
The score for an unscored good is calculated using the following formula:
whereinRepresents the average rating of the items purchased by user U,to representIn the u-th row of (a),is composed ofColumn i.
S3-1: calculating the matrix S
And calculating R ' and R ' R, solving the characteristic value of the R ' R, sequencing from large to small and squaring to obtain a matrix S.
S3-2: calculating the matrix V
And solving the eigenvector of the matrix R' R, and carrying out orthogonalization on the eigenvector according to the sequence of the eigenvalues from large to small.
S3-3 calculating matrix U
And solving the eigenvector of the matrix RR', and carrying out orthogonalization on the eigenvector according to the sequence of eigenvalues from large to small.
S3-4 matrix dimension reduction
Only the first k largest singular values, the first k columns of U, and the first k rows of V are retained.
S3-5 predicting the score of an unscored commodity
Wherein,representing the prediction of the score of the user u for the unscored item i,represents the average rating of the items purchased by user U,to representIn the u-th row of (a),is composed ofColumn i.
Step S4: similarity calculation
Calculating the similarity between the user and the target user by using the adjusted cosine measure, wherein the specific algorithm is as follows:
step S5: selecting neighbors and commodities to be recommended
And sequencing the neighbor users according to the similarity obtained by calculation in the step S4, selecting the nearest k neighbors as the neighbors, and selecting the front k' commodities as the commodities to be recommended from the commodities which have higher scores and are not purchased by the target user.
Step S6: feature extraction for logistic regression module
Assuming that y follows the bernoulli distribution, y-1 represents a purchase behavior and y-0 represents an unpurceful behavior. The behavior that the user purchases the goods is classified as y being 1, and the goods that the user browses but has not purchased are classified as y being 0. The factors influencing the purchase of the commodity by the consumer are taken as explanatory variables and are represented by X, wherein X is (X)1,x2,...xp-1). For example: basic information of the user, such as age, gender, occupation, income condition, consumption level and the like, basic information of the commodity, such as commodity category, price, style, material and the like, and purchasing behavior of the user.
Step S7: variable pre-processing
S7-1: reprogramming classified variables into virtual variables
For example, if a variable has k classes, then (k-1) variables (k ═ 1,2,3 … …) are recoded, the variable form being: x is the number of1＝(1,0,...,0)T...xk-1＝(0,...0,1)T。
S7-2: normalizing continuous variables
The normalization formula is as follows:
step S8: lasso estimation of logistic regression model
The maximum likelihood estimation method is selected to estimate the parameters of the regression equation, where the coefficients in the lasso-logistic regression model are given by the minima of the convex function, as follows:
where L is the log-likelihood function and λ is the harmonic parameter.
The estimates under the lasso algorithm are:
where lambda may be determined by a generalized cross-validation method,
order to
The GCV statistic of the generalized cross validation value is
Penalty parameter for optimizing penalty function for minimizing generalized cross-validation value GCV
The estimation of the parameter β can be obtained through Lagrange multiplier algorithm to obtain a probability estimation model
Step S9: aiming at the target user, the purchase probability of the commodity to be recommended is predicted to obtain a predicted valueAnd sorting according to the prediction probability, and recommending the commodities according to the order.
And substituting the related data of the to-be-recommended commodities and the basic information of the user, which are obtained in the step S5, into the probability estimation model in the step S8 to obtain the probability estimation value of each commodity being purchased. And sequencing the commodities to be purchased according to the probability estimation value of the purchase, and recommending the commodities to the user according to the sequence.
The foregoing is merely a preferred embodiment of the invention, which is intended to be illustrative and not limiting. It will be understood by those skilled in the art that various changes, modifications and equivalents may be made therein without departing from the spirit and scope of the invention as defined in the appended claims.
Claims (10)
1. A personalized recommendation device based on collaborative filtering and logistic regression, comprising:
an input unit: it is used to obtain user information;
an extraction unit: it is used to perform feature extraction on the information;
a scoring unit: which is used for scoring the commodities purchased by the user;
a prediction unit: which is used to predict the scores of unscored goods;
a calculation unit: it is used to calculate the similarity between users;
a selecting unit: it is used to select neighbors and goods to be recommended;
a pretreatment unit: it is used to transform categorical variables and normalize continuous variables;
a model calculation unit: it uses lasso estimation of logistic regression model to get probability estimation model;
a comparison and recommendation unit: the system is used for sequencing predicted values and recommending commodities;
an output unit: which is used to give recommendations for goods to the user.
2. The personalized recommendation device based on collaborative filtering and logistic regression according to claim 1, wherein the extraction unit comprises a collaborative filtering module and a logistic regression module, and different features are extracted respectively.
3. A personalized recommendation method based on collaborative filtering and logistic regression is characterized by comprising the following steps:
step S1: extracting the characteristics of the collaborative filtering part;
step S2: constructing a scoring matrix;
step S3: singular value decomposition;
step S4: calculating the similarity;
step S5: selecting neighbors and determining commodities to be recommended;
step S6: extracting the characteristics of a logistic regression module;
step S7: preprocessing variables;
step S8: lasso estimation of logistic regression model;
step S9: and obtaining a predicted value, and recommending the commodities according to the sequence of the predicted value.
4. The personalized recommendation method based on collaborative filtering and logistic regression as claimed in claim 3, wherein in step S2, different behaviors of the user are quantified and a scoring matrix R is usedm×n(m.gtoreq.n) represents the user's score for the good, Ru,iIndicates the rating of the item i by the user u,representing the average rating of the item that user u has purchased.
5. The personalized recommendation method based on collaborative filtering and logistic regression according to claim 4, wherein in the step S3, R is transformed by singular valuem×nAnd decomposing the matrix into three matrixes of U, S and V, and satisfying that R is U multiplied by S multiplied by V'.
6. The personalized recommendation method based on collaborative filtering and logistic regression as claimed in claim 5, wherein in step S3, the eigenvalues of R' R are solved, and are ranked from large to small and developed to obtain S; solving the eigenvector of R' R, sorting eigenvalues from big to small, and orthogonalizing the eigenvector to obtain V; and solving the feature vector of RR', sorting the feature values from large to small, and orthogonalizing the feature vector to obtain U.
7. The personalized recommendation method based on collaborative filtering and logistic regression as claimed in claim 6, wherein in step S3, dimension S is reduced to k order according to the principle of single value decomposition, and U is obtained by the same principlek，VkThe relation is changed to Rk＝Uk×Sk×Vk。
8. The personalized recommendation method based on collaborative filtering and logistic regression according to claim 7, wherein in the step S3, the score of the non-scored merchandise is calculated according to the following formula:
whereinRepresents the average rating of the items purchased by user U,to representIn the u-th row of (a),is composed ofColumn i.
9. The personalized recommendation method based on collaborative filtering and logistic regression as claimed in claim 8, wherein the step S4 calculates the similarity between the user and the target user by using the adjusted cosine measure, and the specific algorithm is as follows:
10. the personalized recommendation method based on collaborative filtering and logistic regression according to claim 9, wherein in the step S8:
the maximum likelihood estimation method is selected to estimate the parameters of the regression equation, where the coefficients in the lasso-logistic regression model are given by the minima of the convex function, as follows:
where L is the log-likelihood function and λ is the harmonic parameter.
The estimates under the lasso algorithm are:
wherein λ can be determined by a generalized cross-validation method
Order to
The GCV statistic of the generalized cross validation value is
Penalty parameter for optimizing penalty function for minimizing generalized cross-validation value GCV
The estimation of the parameter β can be obtained through Lagrange multiplier algorithm to obtain a probability estimation model
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611254081.3A CN108268464B (en) | 2016-12-30 | 2016-12-30 | Personalized recommendation method and device based on collaborative filtering and logistic regression |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611254081.3A CN108268464B (en) | 2016-12-30 | 2016-12-30 | Personalized recommendation method and device based on collaborative filtering and logistic regression |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108268464A true CN108268464A (en) | 2018-07-10 |
| CN108268464B CN108268464B (en) | 2021-01-12 |
Family
ID=62754305
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611254081.3A Active CN108268464B (en) | 2016-12-30 | 2016-12-30 | Personalized recommendation method and device based on collaborative filtering and logistic regression |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108268464B (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109284442A (en) * | 2018-11-27 | 2019-01-29 | 浙江工商大学 | Merge the linear regression recommended method of collaborative filtering |
| CN109345348A (en) * | 2018-09-30 | 2019-02-15 | 重庆誉存大数据科技有限公司 | The recommended method of multidimensional information portrait based on travel agency user |
| CN109934681A (en) * | 2019-03-13 | 2019-06-25 | 中国科学技术大学 | The recommended method of user's commodity interested |
| CN111242732A (en) * | 2020-01-09 | 2020-06-05 | 北京慧博科技有限公司 | Commodity recommendation model-based recommendation method |
| CN111639274A (en) * | 2020-06-01 | 2020-09-08 | 重庆锐云科技有限公司 | Online commodity intelligent sorting method and device, computer equipment and storage medium |
| CN111767474A (en) * | 2020-09-01 | 2020-10-13 | 上海蜜度信息技术有限公司 | Method and equipment for constructing user portrait based on user operation behaviors |
| CN113409124A (en) * | 2021-07-08 | 2021-09-17 | 山东大学 | Bulk commodity recommendation method and system based on Bayesian regression analysis |
| CN114417174A (en) * | 2022-03-23 | 2022-04-29 | 腾讯科技（深圳）有限公司 | Content recommendation method, device, equipment and computer storage medium |
| CN115293859A (en) * | 2022-09-28 | 2022-11-04 | 江苏火焰云数据技术有限公司 | Electronic commerce platform commodity intelligence recommendation management system based on artificial intelligence |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150312603A1 (en) * | 2014-04-23 | 2015-10-29 | Netflix, Inc. | Recommending media items based on take rate signals |
| CN105678587A (en) * | 2016-01-12 | 2016-06-15 | 腾讯科技（深圳）有限公司 | Recommendation feature determination method and information recommendation method and device |
| CN105787100A (en) * | 2016-03-18 | 2016-07-20 | 浙江大学 | User session recommendation method based on deep neural network |
| CN106022865A (en) * | 2016-05-10 | 2016-10-12 | 江苏大学 | Goods recommendation method based on scores and user behaviors |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP5454357B2 (en) * | 2010-05-31 | 2014-03-26 | ソニー株式会社 | Information processing apparatus and method, and program |
| CN104991899B (en) * | 2015-06-02 | 2018-06-19 | 广州酷狗计算机科技有限公司 | The recognition methods of user property and device |
-
2016
- 2016-12-30 CN CN201611254081.3A patent/CN108268464B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150312603A1 (en) * | 2014-04-23 | 2015-10-29 | Netflix, Inc. | Recommending media items based on take rate signals |
| CN105678587A (en) * | 2016-01-12 | 2016-06-15 | 腾讯科技（深圳）有限公司 | Recommendation feature determination method and information recommendation method and device |
| CN105787100A (en) * | 2016-03-18 | 2016-07-20 | 浙江大学 | User session recommendation method based on deep neural network |
| CN106022865A (en) * | 2016-05-10 | 2016-10-12 | 江苏大学 | Goods recommendation method based on scores and user behaviors |
Non-Patent Citations (2)
| Title |
|---|
| 周春荣: ""百度团购个性化推荐引擎的设计与实现"", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 雷伟: ""基于多行为融合的电商推荐系统研究与实现"", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109345348A (en) * | 2018-09-30 | 2019-02-15 | 重庆誉存大数据科技有限公司 | The recommended method of multidimensional information portrait based on travel agency user |
| CN109284442A (en) * | 2018-11-27 | 2019-01-29 | 浙江工商大学 | Merge the linear regression recommended method of collaborative filtering |
| CN109934681A (en) * | 2019-03-13 | 2019-06-25 | 中国科学技术大学 | The recommended method of user's commodity interested |
| CN109934681B (en) * | 2019-03-13 | 2022-01-11 | 中国科学技术大学 | Recommendation method for user interested goods |
| CN111242732A (en) * | 2020-01-09 | 2020-06-05 | 北京慧博科技有限公司 | Commodity recommendation model-based recommendation method |
| CN111639274A (en) * | 2020-06-01 | 2020-09-08 | 重庆锐云科技有限公司 | Online commodity intelligent sorting method and device, computer equipment and storage medium |
| CN111639274B (en) * | 2020-06-01 | 2023-05-09 | 重庆锐云科技有限公司 | Online commodity intelligent sorting method, device, computer equipment and storage medium |
| CN111767474A (en) * | 2020-09-01 | 2020-10-13 | 上海蜜度信息技术有限公司 | Method and equipment for constructing user portrait based on user operation behaviors |
| CN113409124A (en) * | 2021-07-08 | 2021-09-17 | 山东大学 | Bulk commodity recommendation method and system based on Bayesian regression analysis |
| CN114417174A (en) * | 2022-03-23 | 2022-04-29 | 腾讯科技（深圳）有限公司 | Content recommendation method, device, equipment and computer storage medium |
| CN115293859A (en) * | 2022-09-28 | 2022-11-04 | 江苏火焰云数据技术有限公司 | Electronic commerce platform commodity intelligence recommendation management system based on artificial intelligence |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108268464B (en) | 2021-01-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108268464B (en) | Personalized recommendation method and device based on collaborative filtering and logistic regression | |
| Kang et al. | Visually-aware fashion recommendation and design with generative image models | |
| Zhang et al. | Forecasting agricultural commodity prices using model selection framework with time series features and forecast horizons | |
| CN103093376B (en) | Clustering collaborative filtering recommendation system based on singular value decomposition algorithm | |
| CN108665311B (en) | Electric commercial user time-varying feature similarity calculation recommendation method based on deep neural network | |
| CN108921602B (en) | User purchasing behavior prediction method based on integrated neural network | |
| Choudhary et al. | SARWAS: Deep ensemble learning techniques for sentiment based recommendation system | |
| CN111724235A (en) | Online commodity recommendation method based on user novelty | |
| CN106157156A (en) | A kind of cooperation recommending system based on communities of users | |
| CN112381627B (en) | Commodity scoring processing recommendation method and device under child-care knowledge | |
| CN116205675B (en) | Data acquisition method and device based on thread division | |
| CN112417294A (en) | Intelligent business recommendation method based on neural network mining model | |
| Chhikara et al. | Wine Quality Prediction Using Machine Learning Techniques | |
| Kashef et al. | Predicting l-CrossSold products using connected components: A clustering-based recommendation system | |
| Ali et al. | [Retracted] Estimation Model for Bread Quality Proficiency Using Fuzzy Weighted Relevance Vector Machine Classifier | |
| Doquire et al. | Graph laplacian for semi-supervised feature selection in regression problems | |
| CN111899063A (en) | Fresh agricultural product online recommendation method considering customer consumption behaviors and preference | |
| CN116757747A (en) | Click rate prediction method based on behavior sequence and feature importance | |
| Idowu et al. | Customer Segmentation Based on RFM Model Using K-Means, Hierarchical and Fuzzy C-Means Clustering Algorithms | |
| Kaur et al. | Measuring accuracy of stock price prediction using machine learning based classifiers | |
| Mootha et al. | A Stacking Ensemble of Multi Layer Perceptrons to Predict Online Shoppers' Purchasing Intention | |
| Sharma et al. | Machine learning based prediction of customer spending score | |
| Wulandari et al. | Comparison of Density-Based Spatial Clustering of Applications with Noise (DBSCAN), K-Means and X-Means Algorithms on Shopping Trends Data | |
| Mehrazma et al. | Customer Clustering by Combining the Particle Swarm And K-Means Algorithms and Analyzing Their Behavior on Commercial Websites | |
| Algawiaz et al. | Predicting a User’s Purchase Intention Using AdaBoost |
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 |