CN112765466B - Equipment personalized recommendation method based on user history - Google Patents
Equipment personalized recommendation method based on user history Download PDFInfo
- Publication number
- CN112765466B CN112765466B CN202110057021.7A CN202110057021A CN112765466B CN 112765466 B CN112765466 B CN 112765466B CN 202110057021 A CN202110057021 A CN 202110057021A CN 112765466 B CN112765466 B CN 112765466B
- Authority
- CN
- China
- Prior art keywords
- equipment
- user
- matrix
- scoring
- score
- 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
Images
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
Abstract
The invention discloses an equipment personalized recommendation method based on user history, which aims at the problem that an inquiry system matched with traditional equipment data is not friendly to related users, introduces a recommendation system into an equipment database, enables users to obtain self recommendation, can also obtain neighborhood users similar to the users, and obtains self interested equipment information from browsing history of the neighborhood users; aiming at the recommendation based on articles in the traditional recommendation system, the invention adds the historical score data of the user into the evaluation index, takes the score characteristics of the user into consideration, realizes the reward and punishment factor on whether the recommendation equipment is in accordance with the taste of the user, simultaneously takes the time stamp of the user score into consideration, adds the weight factor which changes along with the time stamp, and finally completes the personalized equipment recommendation considering the user.
Description
Technical Field
The invention belongs to the technical field of recommendation systems, and particularly relates to an equipment personalized recommendation method based on user history.
Background
The equipment database contains a large amount of data such as equipment types, models and parameters. However, when facing these massive military equipment data, the relevant personnel cannot efficiently acquire key knowledge from the data, and thus cannot guide work according to the data. The first problem of drowsiness of value information in data is that the data lack a compact and effective organization structure and visual query mode, deep data mining and application are difficult to perform, and even if a visual interface is provided, the traditional search engine and query mode can bring great inconvenience to related personnel.
The traditional visual interface is based on query modes such as keyword query, the professional requirement on operators is high (the operators must be familiar with the proprietary names of equipment), and meanwhile, all users are treated with the same idea and only a simple query function is provided, so that the system is not personalized enough and is not friendly to the users.
Disclosure of Invention
Aiming at the defects in the prior art, the equipment personalized recommendation method based on the user history solves the problem that the equipment personalized recommendation is difficult to perform according to the user requirements in the existing equipment database.
In order to achieve the purpose of the invention, the invention adopts the technical scheme that: an equipment personalized recommendation method based on user history comprises the following steps:
s1, introducing the recommendation system into an equipment database;
s2, processing the user historical data of the equipment using database through a recommendation system, and constructing a user-equipment scoring matrix;
s3, decomposing the user-equipment scoring matrix to obtain a corresponding equipment feature matrix;
s4, calculating the similarity of each equipment in the equipment feature matrix, and intercepting the first K equipment with the highest similarity as an initial recommendation list of the target equipment;
s5, setting a reward and punishment factor and a weight factor based on historical data of a user, and generating a rating table of each device in the initial recommendation list;
s6, intercepting a plurality of scoring information with front scores in the scoring table, generating an equipment recommendation list, and realizing the personalized recommendation of the equipment.
Further, the user-equipment scoring matrix R (m, n) in step S2 is:
in the formula, rui∈R(m,n)，ruiFor user uuTo equipment iiWhen the user u scoresuUnpaired equipment iiWhen scoring, ruiTo 0, the user-equipment scoring matrix R (m, n) includes a set of users U ═ U1,u2,…,umAnd set of equipment I ═ I1,i2,…,in}，uu∈U，ii∈I。
Further, the step S3 is specifically:
s31, decomposing and reducing the user-equipment scoring matrix R (m, n) into a form of product of the user characteristic matrix and the equipment characteristic matrix:
R＝W×H
wherein R is a prediction matrix of a user-equipment scoring matrix R (m, n) which can be decomposed into a product of a user characteristic matrix and an equipment characteristic matrix, W is a user characteristic matrix of m x k dimensions, and H is an equipment characteristic matrix of k x n;
s32, dividing the element R in the user-equipment scoring matrix R (m, n)uiExpressed as:
s33, calculating the difference D between the prediction matrix R and the user-equipment scoring matrix R;
d is the minimum euclidean distance between R and W, H characterizing the difference between the prediction matrix R and the user-equipment scoring matrix R, S is the set of scores of users of the user-equipment scoring matrix R (m, n), and S ═ ui | Rui>0}，
For the u-th row of the user feature matrix, HiFor the ith column of the article feature matrix,
r, a regularization parameter, is a small positive value,
is the square of the two-norm of the matrix W,
is the square of the two-norm of matrix H;
and S34, solving the difference D by adopting a gradient descent method, obtaining the final difference D when the specified iteration times or the updating amount reaches a set threshold value, and further decomposing the user-equipment scoring matrix R (m, n) into a form of multiplying the user characteristic matrix and the equipment characteristic matrix to obtain the corresponding equipment characteristic matrix.
The beneficial effects of the above further scheme are: because the user-equipment scoring matrix is a sparse matrix, the quality of the recommendation result can be influenced by directly carrying out subsequent processing on the user-equipment scoring matrix, and the subsequent direct processing on the equipment characteristic matrix is more reasonable by decomposing the user-equipment scoring matrix, so that the recommendation result is optimized.
Further, the step S4 is specifically:
s41, setting the equipment-characteristic evaluation set of the equipment u in the equipment characteristic matrix H as IuThe equipment-feature evaluation set of the equipment v is IvDetermining a pearson similarity Sim (u, v) between equipment u and equipment v:
in the formula IuvFor equipment-feature scores of all IuAnd the equipment-feature score is IvThe set of intersection scores of (a) is,
for equipment u, the equipment-feature scores are all IuThe mean value of the scores in (1),
for equipment v, the equipment-feature scores are all IvThe score mean value of (1), u, v belongs to I;
s42, based on calculated Sim (u, v), constructing similarity matrix S of equipmentwep：
Wherein Sim (u, v) is associated with the similarity matrix SwepCorrespond to each element in (1);
s43 similarity matrix S based on equipmentwepAnd determining the K pieces of equipment with the highest similarity of each type of equipment, and forming a recommendation set KNN (w) of the equipment in the field.
The beneficial effects of the above further scheme are: the equipment is distinguished from the angle of similarity, and the accuracy of subsequent equipment recommendation is improved.
Further, the step S5 is specifically:
s51, determining a user history scoring data set rank (k) in the user history data;
s52, determining the score r of each equipment in the recommendation set KNN (w) based on the score data in the user historical score data set rank (k) and the time weight factor corresponding to the score datarate；
S53, based on the user score difference, the score r of each equipmentrateIn the method, reward and punishment factors are set, and then the final score r of each equipment is obtainedk；
S54 final score r based on each equipmentkAnd sorting according to the grading scores to obtain a recommended grading table of each device.
Further, in the step S52, the score r of the equipment wrateComprises the following steps:
in the formula, simminTo set a similarity threshold, akFor the set temporal weighting factor, ∑ ak1, Sim _ sum is in rank (k), similarity Sim (r)k,rw) Greater than simminTotal number of equipment of rk,rw∈I。
Further, the reward punishment factor set in the step S53 includes a penalty factor recomeCorresponding reward and punishment factor rincoAnd a reward factor incomeCorresponding reward and punishment factor rreco；
For equipment k in the user history scoring data set rank (k):
when (Sim (k, w)>simmin)&&(rk≥vavg) Time, reward factor income1, corresponding reward penalty factor
When (Sim (k, w)>simmin)&&(rk<vavg) Penalty factor recome1, corresponding reward penalty factor
In the formula, vavgAnd (4) scoring a threshold for the set user, wherein k and w are equal to I.
Further, the final score r of each equipment in the step S53kComprises the following steps:
rk＝rrate+rinco-rreco。
the beneficial effects of the above further scheme are: passing similarity threshold simminThe statistics of the aggregation of user attention equipment is realized, a time weight factor is set to measure the proportion of the historical comment of the user, and the referential property of the historical data of the user is increased; under the condition of considering the scoring characteristics of the users, reward and punishment factors are set to balance scoring data of different users; therefore, the factors are considered in determining the final score of each equipment, so that the final score result obtained by the equipment is more objective and accurate.
The invention has the beneficial effects that:
(1) aiming at the problem that a query system matched with traditional equipment data is not friendly to related users, the recommendation system is introduced into the equipment database, so that a user can obtain recommendations of the user and can also obtain neighborhood users similar to the user, and interested equipment information of the user is obtained from the browsing history of the neighborhood users;
(2) the invention aims at the recommendation based on articles in the traditional recommendation system, and makes the recommendation based on articles more suitable for users, the recommendation based on articles in the traditional recommendation system is based on the similarity ranking of the recommendation set after the recommendation set of the current equipment is obtained, topK equipment with the highest degree of identity is selected to complete the recommendation based on articles, only the similarity information of the current grading equipment of the users is considered, but the information of the users grading the equipment is not considered, and the method embodies the inline relation among the equipment and does not comprehensively consider the personalized information of the users. Aiming at the point, the historical scoring data of the user is added into the evaluation index, the scoring characteristics of the user are considered, the reward and punishment factors are realized whether the recommended equipment is in accordance with the taste of the user, meanwhile, the weighting factors changing along with the timestamp are added in consideration of the user scoring timestamp, and finally the personalized equipment recommendation considering the user is completed.
Drawings
Fig. 1 is a flowchart of a method for personalized recommendation of equipment based on user history.
Detailed Description
The following description of the embodiments of the present invention is provided to facilitate the understanding of the present invention by those skilled in the art, but it should be understood that the present invention is not limited to the scope of the embodiments, and it will be apparent to those skilled in the art that various changes may be made without departing from the spirit and scope of the invention as defined and defined in the appended claims, and all matters produced by the invention using the inventive concept are protected.
As shown in fig. 1, a method for personalized recommendation of equipment based on user history includes the following steps:
s1, introducing the recommendation system into an equipment database;
s2, processing the user historical data of the equipment using database through a recommendation system, and constructing a user-equipment scoring matrix;
s3, decomposing the user-equipment scoring matrix to obtain a corresponding equipment feature matrix;
s4, calculating the similarity of each equipment in the equipment feature matrix, and intercepting the first K equipment with the highest similarity as an initial recommendation list of the target equipment;
s5, setting a reward and punishment factor and a weight factor based on historical data of a user, and generating a rating table of each device in the initial recommendation list;
s6, intercepting a plurality of scoring information with front scores in the scoring table, generating an equipment recommendation list, and realizing the personalized recommendation of the equipment.
In the step S1, the recommendation system is first introduced into the equipment database, so as to solve the problem that the traditional visualization interface is not personalized for different people;
in step S2, collecting evaluation information of the equipment by the relevant user in the form of questionnaire or the like as user history data; processing the user historical data to obtain a user-equipment scoring matrix R (m, n) as the basis of the subsequent processing process, wherein the user-equipment scoring matrix R (m, n) is as follows:
in the formula, rui∈R(m,n)，ruiFor user uuTo equipment iiWhen the user u scoresuUnpaired equipment iiWhen scoring, ruiTo 0, the user-equipment scoring matrix R (m, n) includes a set of users U ═ U1,u2,…,umAnd set of equipment I ═ I1,i2,…,in}，uu∈U，ii∈I。
Due to the diversity of the equipment, all the equipment cannot be scored for each user, so the matrix R (m, n) is a sparse matrix, the sparsity of the matrix reaches over 95% in the context of current big data, the sparsity can affect the quality of the recommendation result quite if not processed, the simplest way is to give a fixed score to the missing score, for example, the historical score of the user is used for replacing the missing score, but the optimization of the recommendation result is not improved greatly in this way, and the better way is to decompose the matrix into the product of the user characteristic matrix and the article characteristic matrix. Therefore, the step S3 is specifically:
s31, decomposing and reducing the user-equipment scoring matrix R (m, n) into a form of product of a user characteristic matrix and an equipment characteristic matrix:
R＝W×H
wherein, R is a prediction matrix of a user-equipment scoring matrix R (m, n) which can be decomposed into a product form of a user characteristic matrix and an equipment characteristic matrix, W is a user characteristic matrix with m multiplied by k dimensions, and H is an equipment characteristic matrix with k multiplied by n dimensions, and the physical meaning of the matrix is that k similar characteristics exist between a user and equipment;
s32, dividing the element R in the user-equipment scoring matrix R (m, n)uiExpressed as:
s33, calculating the difference D between the prediction matrix R and the user-equipment scoring matrix R;
in particular, in order to make it possible for the matrix resulting from the multiplication of W and H to only approximate the user-equipment scoring matrix R (m, n), it is necessary to calculate the difference between the prediction matrix and the user-equipment scoring matrix R (m, n), measured in terms of euclidean distance, assuming that the set of scores of users of the user-equipment scoring matrix R (m, n) is S, S ═ { ui | Rui>0, let D be the minimum Euclidean distance between W, H, D be:
in order to solve the problem of overfitting in the process of calculating the minimum D, a specification factor needs to be added to the D
So that the truly minimized D becomes the original D plus the normalization factor, at which time D becomes:
d is the minimum euclidean distance between R and W, H characterizing the difference between the prediction matrix R and the user-equipment scoring matrix R, S is the set of scores of users of the user-equipment scoring matrix R (m, n), and S ═ ui | Rui>0}，
For the u-th row of the user feature matrix, HiFor the ith column of the article feature matrix,
r, a regularization parameter, is a small positive value,
is the square of the two-norm of the matrix W,
is the square of the two-norm of matrix H;
and S34, solving the difference D by adopting a gradient descent method, obtaining the final difference D when the specified iteration times or the updating amount reaches a set threshold value, and further decomposing the user-equipment scoring matrix R (m, n) into a form of multiplying the user characteristic matrix and the equipment characteristic matrix to obtain the corresponding equipment characteristic matrix.
Specifically, the minimum D value is solved by a gradient descent method, which includes the following steps:
the update rule is as follows:
wherein tau is an iteration step length, and when the specified iteration times is reached through the repeated updating rule or the updating amount reaches a threshold epsilon, the final result obtained by updating is stopped. At this time, the score matrix R (m, n) is decomposed into a product of the user feature matrix and the equipment feature matrix.
In the step S4, the similarity between the two pieces of equipment is obtained by simultaneously scoring the two pieces of equipment, and because of the sparsity of the scoring matrix, the equipment feature matrix H subjected to feature decomposition is used for calculation, and meanwhile, the pearson correlation coefficient is used for calculation in terms of measurement, so the step S4 specifically includes:
s41, setting the equipment-characteristic evaluation set of the equipment u in the equipment characteristic matrix H as IuThe equipment-feature evaluation set of the equipment v is IvDetermining a pearson similarity Sim (u, v) between equipment u and equipment v:
in the formula IuvFor equipment-feature scores of all IuAnd the equipment-feature score is IvThe set of intersection scores of (a) is,
for equipment u, the equipment-feature scores are all IuThe mean value of the scores in (1),
for equipment v, the equipment-feature scores are all IvThe score mean value of (1), u, v belongs to I;
s42, constructing similarity matrix S of the equipment based on the calculated Sim (u, v)wep：
Wherein Sim (u, v) is associated with the similarity matrix SwepCorrespond to each element in (1);
s43 similarity matrix S based on equipmentwepAnd determining the top K pieces of equipment with the highest similarity of each type of equipment as an initial recommendation list KNN (w) of the target equipment.
In the step S5, after a certain rating is given to a certain user, the equipment rated by the user at this time is given knn (w) obtained in the step S4, and the obtained historical rating data of the user is rank (k), and for the same user, the equipment concerned by the user in the last period of time certainly has a certain "commonality", that is, an aggregation exists in the similarity of the task in the last period, so as to obtain a specific rating of each equipment;
specifically, the step S5 is specifically:
s51, determining a user history scoring data set rank (k) in the user history data;
s52, determining the score r of each equipment in the initial recommendation list KNN (w) based on the score data in the user historical score data set rank (k) and the time weight factor corresponding to the score datarate；
Specifically, for the equipment in knn (w), the equipment should be reflected in similarity with the historical scoring equipment of the user, and then a similarity threshold sim is setminThe method is a reasonable mode, scoring historical data of a user is a quasi-stable process, the score of the positive closer time has greater significance to a metric set along with the change of a timestamp, and a weight for measuring the time is introduced by considering the conditionIncreased factor akSo that the result is more reasonable, in the final step S52, the score r of the equipment wrateComprises the following steps:
in the formula, simminTo set a similarity threshold, akFor the set temporal weighting factor, ∑ ak1, Sim _ sum is in rank (k), similarity Sim (r)k,rw) Greater than simminTotal number of equipment of rk,rw∈I；akThe influence factor used for measuring the progressive and bigger time stamps in the user historical comment can dynamically change along with the time stamps.
S53, based on the user score difference, the score r of each equipmentrateIn the method, reward and punishment factors are set, and then the final score r of each equipment is obtainedk；
Specifically, for each different user, there is a scoring characteristic of the user, some users are conservative in preference, the scoring may be generally low, some users are optimistic in preference, and the scoring may be generally high. Considering the user's characteristics, a user scoring threshold v can be setavgMeanwhile, a reward and punishment mechanism is added, and when the reward and punishment mechanism is higher than the threshold, equipment which accords with the expectation of the user needs to be provided with a reward factor incomeIf the value is lower than the threshold, the expectation is not met, and a penalty factor re needs to be establishedcome(ii) a Therefore, the reward penalty factor set in step S53 includes a penalty factor recomeCorresponding reward and punishment factor and reward factor incomeA corresponding reward and penalty factor;
for equipment k in the user history scoring data set rank (k):
when (Sim (k, w)>simmin)&&(rk≥vavg) Time, reward factor income1, corresponding reward penalty factor
When (Sim (k, w)>simmin)&&(rk<vavg) Penalty factor recome1, corresponding reward penalty factor
In the formula, vavgAnd (4) scoring a threshold for the set user, wherein k and w are equal to I.
The logarithm of the reward and punishment factors is mainly used for slightly reducing the influence degree of the reward and punishment factors;
for all the recommended equipment, the final evaluation criteria are determined by the two, and the final score r of each equipment is obtainedkComprises the following steps:
rk＝rrate+rinco-rreco
s54 final score r based on each equipmentkAnd sorting according to the grading scores to obtain a grading table of each device.
Claims (6)
1. An equipment personalized recommendation method based on user history is characterized by comprising the following steps:
s1, introducing the recommendation system into an equipment database;
s2, processing the user historical data of the equipment using database through a recommendation system, and constructing a user-equipment scoring matrix;
s3, decomposing the user-equipment scoring matrix to obtain a corresponding equipment feature matrix;
s4, calculating the similarity of each equipment in the equipment feature matrix, and intercepting the first K equipment with the highest similarity as an initial recommendation list of the target equipment;
s5, setting a reward and punishment factor and a weight factor based on the historical data of the user, and generating a rating table of each device in the initial recommendation list;
s6, intercepting a plurality of scoring information with front scores in the scoring table, generating an equipment recommendation list, and realizing personalized recommendation of equipment;
the step S3 specifically includes:
s31, decomposing and reducing the user-equipment scoring matrix R (m, n) into a form of product of a user characteristic matrix and an equipment characteristic matrix:
R＝W×H
wherein R is a prediction matrix of a user-equipment scoring matrix R (m, n) which can be decomposed into a product of a user characteristic matrix and an equipment characteristic matrix, W is a user characteristic matrix of m x k dimensions, and H is an equipment characteristic matrix of k x n;
s32, dividing the element R in the user-equipment scoring matrix R (m, n)uiExpressed as:
s33, calculating the difference D between the prediction matrix R and the user-equipment scoring matrix R;
where D is the minimum euclidean distance between R and W, H characterizing the difference between the prediction matrix R and the user-equipment scoring matrix R, S is the set of scores for users of the user-equipment scoring matrix R (m, n), and S ═ ui | Rui＞0}，
For the u-th row of the user feature matrix, HiFor the ith column of the article feature matrix,
r, a regularization parameter, is a small positive value,
is the square of the two-norm of the matrix W,
is a momentThe square of the two norms of the matrix H;
s34, solving the difference D by adopting a gradient descent method, obtaining the final difference D when the specified iteration times or the updating amount reaches a set threshold value, and further decomposing the user-equipment scoring matrix R (m, n) into a form of multiplying the user characteristic matrix and the equipment characteristic matrix to obtain a corresponding equipment characteristic matrix;
the step S5 specifically includes:
s51, determining a user history scoring data set rank (k) in the user history data;
s52, determining the score r of each equipment in the initial recommendation list KNN (w) based on the score data in the user historical score data set rank (k) and the time weight factor corresponding to the score datarate；
S53, based on the user score difference, the score r of each equipmentrateIn the method, reward and punishment factors are set, and then the final score r of each equipment is obtainedk；
S54 final score r based on each equipmentkAnd sorting according to the grading scores to obtain a grading table of each device.
2. The method for personalized recommendation of equipment based on user history as claimed in claim 1, wherein the user-equipment scoring matrix R (m, n) in step S2 is:
in the formula, rui∈R(m,n)，ruiFor user uuTo equipment iiWhen the user u scoresuUnpaired equipment iiWhen scoring, ruiTo 0, the user-equipment scoring matrix R (m, n) includes a set of users U ═ U1,u2,…,umAnd set of equipment I ═ I1,i2,…,in}，uu∈U，ii∈I。
3. The method for personalized recommendation of equipment based on user history as claimed in claim 1, wherein said step S4 specifically comprises:
s41, setting the equipment-characteristic evaluation set of the equipment u in the equipment characteristic matrix H as IuThe equipment-feature evaluation set of the equipment v is IvDetermining a pearson similarity Sim (u, v) between equipment u and equipment v:
in the formula IuvFor equipment-feature scores of all IuAnd the equipment-feature score is IvThe set of intersection scores of (a) is,
for equipment u, the equipment-feature scores are all IuThe mean value of the scores in (1),
for equipment v, the equipment-feature scores are all IvThe score mean value of (1), u, v belongs to I;
s42, constructing similarity matrix S of the equipment based on the calculated Sim (u, v)wep：
Wherein Sim (u, v) is associated with the similarity matrix SwepCorrespond to each element in (1);
s43 similarity matrix S based on equipmentwepAnd determining the top K pieces of equipment with the highest similarity of each type of equipment as an initial recommendation list KNN (w) of the target equipment.
4. The method for personalized recommendation of equipment based on user history as claimed in claim 1, wherein in step S52, the score r of equipment wrateComprises the following steps:
in the formula, simminTo set a similarity threshold, akFor the set temporal weighting factor, ∑ ak1, Sim _ sum is in rank (k), similarity Sim (r)k,rw) Greater than simminTotal number of equipment of rk,rw∈I。
5. The method for personalized recommendation of equipment based on user history as claimed in claim 4, wherein the reward penalty factor set in step S53 comprises a penalty factor recomeCorresponding reward and punishment factor rincoAnd a reward factor incomeCorresponding reward and punishment factor rreco；
For equipment k in the user history scoring data set rank (k):
when (Sim (k, w) > Simmin)&&(rk≥vavg) Time, reward factor income1, corresponding reward penalty factor
When (Sim (k, w) > Simmin)&&(rk<vavg) Penalty factor recome1, corresponding reward penalty factor
In the formula, vavgAnd (4) scoring a threshold for the set user, wherein k and w are equal to I.
6. The method for personalized recommendation of equipment based on user history as claimed in claim 5, wherein said step S53 is that the final score r of each equipmentkComprises the following steps:
rk＝rrate+rinco-rreco。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110057021.7A CN112765466B (en) | 2021-01-15 | 2021-01-15 | Equipment personalized recommendation method based on user history |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110057021.7A CN112765466B (en) | 2021-01-15 | 2021-01-15 | Equipment personalized recommendation method based on user history |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112765466A CN112765466A (en) | 2021-05-07 |
| CN112765466B true CN112765466B (en) | 2022-05-03 |
Family
ID=75702010
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110057021.7A Active CN112765466B (en) | 2021-01-15 | 2021-01-15 | Equipment personalized recommendation method based on user history |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112765466B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2010134733A (en) * | 2008-12-05 | 2010-06-17 | Dainippon Printing Co Ltd | Information recommendation device, information recommendation method, and program |
| CN107066476A (en) * | 2016-12-13 | 2017-08-18 | 江苏途致信息科技有限公司 | A kind of real-time recommendation method based on article similarity |
| CN107871259A (en) * | 2016-09-26 | 2018-04-03 | 阿里巴巴集团控股有限公司 | A kind of processing method of information recommendation, device and client |
-
2021
- 2021-01-15 CN CN202110057021.7A patent/CN112765466B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2010134733A (en) * | 2008-12-05 | 2010-06-17 | Dainippon Printing Co Ltd | Information recommendation device, information recommendation method, and program |
| CN107871259A (en) * | 2016-09-26 | 2018-04-03 | 阿里巴巴集团控股有限公司 | A kind of processing method of information recommendation, device and client |
| CN107066476A (en) * | 2016-12-13 | 2017-08-18 | 江苏途致信息科技有限公司 | A kind of real-time recommendation method based on article similarity |
Non-Patent Citations (2)
| Title |
|---|
| Diversifying Personalized Recommendation with User-session Context;Liang Hu等;《Proceedings of the Twenty-Sixth International Joint Conference on Artificial Intelligence (IJCAI-17)》;20171231;全文 * |
| 基于偏好度特征构造的个性化推荐算法;黄金超等;《上海交通大学学报》;20180731;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112765466A (en) | 2021-05-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110020128B (en) | Search result ordering method and device | |
| US9535911B2 (en) | Processing a content item with regard to an event | |
| US8612435B2 (en) | Activity based users' interests modeling for determining content relevance | |
| US7685091B2 (en) | System and method for online information analysis | |
| CN108363821A (en) | A kind of information-pushing method, device, terminal device and storage medium | |
| CN107122467B (en) | Search engine retrieval result evaluation method and device and computer readable medium | |
| CN107341268B (en) | Hot searching ranking method and system | |
| CN102779193B (en) | Self-adaptive personalized information retrieval system and method | |
| CN104615741B (en) | Cold-start project recommendation method and device based on cloud computing | |
| CN101556603A (en) | Coordinate search method used for reordering search results | |
| CN114707074B (en) | Content recommendation method, device and system | |
| CN110647678B (en) | Recommendation method based on user character label | |
| CN112052394B (en) | Professional content information recommendation method, system, terminal equipment and storage medium | |
| CN103309894A (en) | User attribute-based search realization method and system | |
| CN112487283A (en) | Method and device for training model, electronic equipment and readable storage medium | |
| CN112612951A (en) | Unbiased learning sorting method for income improvement | |
| CN109508407A (en) | The tv product recommended method of time of fusion and Interest Similarity | |
| CN111915409A (en) | Article recommendation method, device and equipment based on article and storage medium | |
| CN113407729B (en) | Judicial-oriented personalized case recommendation method and system | |
| CN104572915A (en) | User event relevance calculation method based on content environment enhancement | |
| CN116541607B (en) | Intelligent recommendation method based on commodity retrieval data analysis | |
| CN112765466B (en) | Equipment personalized recommendation method based on user history | |
| CN112733006B (en) | User portrait generation method, device and equipment and storage medium | |
| CN116680320A (en) | Mixed matching method based on big data | |
| CN110727867A (en) | Semantic entity recommendation method based on fuzzy mechanism |
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 |