CN109271590B - Recommendation method based on time sequence decision model - Google Patents
Recommendation method based on time sequence decision modelDownload PDFInfo
- Publication number
- CN109271590B CN109271590BCN201811145391.0ACN201811145391ACN109271590BCN 109271590 BCN109271590 BCN 109271590BCN 201811145391 ACN201811145391 ACN 201811145391ACN 109271590 BCN109271590 BCN 109271590B
- Authority
- CN
- China
- Prior art keywords
- user
- parameter
- recommendation
- changed
- role
- 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
- 238000000034methodMethods0.000titleclaimsabstractdescription14
- 238000003066decision treeMethods0.000claimsabstractdescription6
- 230000004048modificationEffects0.000claimsdescription7
- 238000012986modificationMethods0.000claimsdescription7
- 239000011159matrix materialSubstances0.000claimsdescription4
- 230000003542behavioural effectEffects0.000claimsdescription3
- 230000006399behaviorEffects0.000description39
- 230000006870functionEffects0.000description4
- 238000004364calculation methodMethods0.000description2
- 230000009286beneficial effectEffects0.000description1
- 230000007547defectEffects0.000description1
- 238000000605extractionMethods0.000description1
- 238000002347injectionMethods0.000description1
- 239000007924injectionSubstances0.000description1
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
Description
Technical Field
The invention relates to a recommendation method for realizing personalized information retrieval according to conditions such as user personality and characteristics, in particular to a recommendation method based on a time sequence decision model.
Background
With the development and popularization of the internet, information and data from the network are explosively increased, and the requirements of different users on various types of information are fully met. On the other hand, however, the largest detriment caused by an excess of data information is an information overload: in the case of the information flow of the well-injection type, a user needs to face a large amount of useless information, or the user cannot determine which information is useful in a short time, and the information utilization efficiency is reduced. Under such a circumstance, the recommendation system is applied to various fields as it is, and is helpful to improve the retrieval efficiency of the data information.
The recommendation system is established on the basis of mass data and provides personalized decision support and information service for users according to conditions of personal identity, interest, requirements and the like of the users. Through personalized calculation of the user, condition information matched with the user characteristics is screened out, so that the user is guided to find out information which is realized by the user, particularly potential information which is not realized by the user, but can meet conditions such as preference of the user, and the like, wherein the potential information comprises information which is known by the user, forgotten and never found.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a recommendation method based on a time sequence decision model, which realizes individual recommendation of a user through extraction and analysis of user behavior characteristics and behavior association degree analysis, is convenient for the user to quickly find needed information in a large amount of data and can improve the use efficiency of the user on the information.
The purpose of the invention is realized by the following technical scheme: a recommendation method based on a time sequence decision model comprises the following steps:
s1, defining the following historical behavior sequence of the user at certain n continuous time: (x)1，x2，......xn)；
S2, representing the behavior sequence prediction of the user in a recursive mode:
xi+1＝Q1xi+Q2x′i (1)
wherein Q is1、Q2Respectively representing parameters needing to be predicted; x is the number ofiInformation representing a last recommendation to the user; x is the number ofi+1The recommendation information which is about to be given to the user at the next moment is shown, namely the target required by the scheme; x'iRepresenting the user's behavior since receiving the last time recommendation;
s3, judging parameter Q1And Q2If the parameter Q needs to be modified, calculating the parameter Q by adopting a decision tree model1And Q2Otherwise, return to step S2.
Further, the step S3 includes the following sub-steps:
s31, confirming whether the parameter Q needs to be modified1And Q2If the modification is needed, executing the step S32, otherwise, not operating;
s32, determining Q by modeling an optimal objective function1And Q2A value of (d);
the specific implementation method comprises the following steps: the objective function is set as:
min Cost(Id，Action)＝Id(Q1)+Action(Q1，Q2) (2)
wherein,
the second part of the objective function equation, Action () above, is calculated for the value of the behavioral Action:
a represents the complete set of all the different objects that the user may access; assuming a total of i different objects, the set is denoted A1，A2......Ai；R(An+1) Representing a slave object An-1To AnThe correlation value between the two is determined by the object correlation weight matrix;
represents the parameter Q1The value of the n-th prediction,
represents the parameter Q1The value of the (n + 1) th prediction,
represents the parameter Q2The value of the n-th prediction,
represents the parameter Q2The value of the (n + 1) th prediction;
for parameter Q1And Q2Respectively deriving the formula (3) to obtain
And
the values of (A) are as follows:
order to
Obtain a new parameter Q1And Q2The value is obtained.
Further, the parameter Q in the step S311And Q2Whether modification is needed is confirmed by three factors of whether the user role is changed or not, whether the user browsing behavior is changed or not and whether last recommendation is accepted or not;
wherein, the parameter Q1And Q2Conditions that need to be modified include:
(1) changing the role of the user, changing the browsing behavior of the user and accepting the last recommendation;
(2) changing the role of the user, not changing the browsing behavior of the user, and accepting the last recommendation;
(3) the user role is not changed, the user browsing behavior is changed, and the last recommendation is accepted;
(4) the user role is not changed, the user browsing behavior is not changed, and the last recommendation is accepted;
without modifying the parameter Q1And Q2The conditions of (a) include:
(1) changing the role of the user, changing the browsing behavior of the user and not receiving the last recommendation;
(2) changing the role of the user, not changing the browsing behavior of the user, and not receiving the last recommendation;
(3) the role of the user is not changed, the browsing behavior of the user is changed, and the last recommendation is not received;
(4) the role of the user is not changed, the browsing behavior of the user is not changed, and the last recommendation is not received.
The invention has the beneficial effects that: according to the invention, the individual recommendation of the user is realized by extracting and analyzing the user behavior characteristics and analyzing the behavior association degree, the user can conveniently and quickly find needed information in a large amount of data, and the use efficiency of the user on the information can be improved.
Drawings
FIG. 1 is a decision tree model of the present invention.
Detailed Description
The key point is that the behavior actions of the user are recorded in a time sequence arrangement, a relation can be generated between the behavior sequences, and the magnitude of the relation value between the behaviors and the preference value of the user and the behaviors are obtained through the learning calculation of sequence numerical values.
The technical scheme of the invention is further explained by combining the attached drawings.
The invention relates to a recommendation method based on a time sequence decision model, which comprises the following steps:
s1, defining the following historical behavior sequence of the user at certain n continuous time: (x)1，x2，......xn)，xiAnd xj(i ≠ j) may be the same or different; the sequence represents the web sites that the user has visited or clicked on the n time windows, and so on. The scheme calculates the next object x to be accessed according to the identity change of the user, the current behavior characteristics and the object association degree weight matrixi+1And recommend it to the user.
S2, at two points in time that are not far apart, the identity and personality preferences of the user do not change much, and behaviors that occur at closer time intervals are more characteristic of the user. And behavior that occurs over time is of little value for behavior prediction. The recommended predicted items for the user are often associated with a plurality of historical behaviors of the user that are closest to the current time. Therefore, the behavior sequence prediction of the user is expressed in a recursive manner:
xi+1＝Q1xi+Q2x′i (1)
wherein Q is1、Q2Respectively representing parameters needing to be predicted; x is the number ofiInformation representing a last recommendation to the user; x is the number ofi+1The recommendation information which is about to be given to the user at the next moment is shown, namely the target required by the scheme; x'iRepresenting the user's behavior since receiving the last time recommendation;
s3, judging parameter Q1And Q2If the parameter Q needs to be modified, calculating the parameter Q by adopting a decision tree model1And Q2Otherwise, returning to step S2; the method specifically comprises the following substeps:
s31, confirming whether the parameter Q needs to be modified1And Q2If the modification is needed, executing the step S32, otherwise, not operating;
s32, determining Q by modeling an optimal objective function1And Q2A value of (d);
the specific implementation method comprises the following steps: the objective function is set as:
min Cost(Id，Action)＝Id(Q1)+Action(Q1，Q2) (2)
wherein,
the second part of the objective function equation, Action () above, is calculated for the value of the behavioral Action:
a represents the complete set of all the different objects that the user may access; assuming a total of i different objects, the set is denoted A1，A2......Ai；R(An+1) Representing a slave object An-1To AnThe correlation value between the two is determined by the object correlation weight matrix;
represents the parameter Q1The value of the n-th prediction,
represents the parameter Q1The value of the (n + 1) th prediction,
represents the parameter Q2The value of the n-th prediction,
represents the parameter Q2The value of the (n + 1) th prediction;
for parameter Q1And Q2Respectively deriving the formula (3) to obtain
And
the values of (A) are as follows:
order to
Obtain a new parameter Q1And Q2The value is obtained.
Furthermore, the decision tree is mainly formed according to the historical behaviors of the user, and the Q is judged according to the behaviors of the user1And Q2Whether or not modification is required, as shown in fig. 1. The parameter Q in the step S311And Q2Whether the modification is needed is confirmed by three factors of whether the user role is changed or not, whether the user browsing behavior is changed or not and whether the last recommendation is accepted or not.
Wherein, the parameter Q1And Q2Conditions that need to be modified include:
(1) changing the role of the user, changing the browsing behavior of the user and accepting the last recommendation;
(2) changing the role of the user, not changing the browsing behavior of the user, and accepting the last recommendation;
(3) the user role is not changed, the user browsing behavior is changed, and the last recommendation is accepted;
(4) the user role is not changed, the user browsing behavior is not changed, and the last recommendation is accepted;
without modifying the parameter Q1And Q2The conditions of (a) include:
(1) changing the role of the user, changing the browsing behavior of the user and not receiving the last recommendation;
(2) changing the role of the user, not changing the browsing behavior of the user, and not receiving the last recommendation;
(3) the role of the user is not changed, the browsing behavior of the user is changed, and the last recommendation is not received;
(4) the role of the user is not changed, the browsing behavior of the user is not changed, and the last recommendation is not received.
It will be appreciated by those of ordinary skill in the art that the embodiments described herein are intended to assist the reader in understanding the principles of the invention and are to be construed as being without limitation to such specifically recited embodiments and examples. Those skilled in the art can make various other specific changes and combinations based on the teachings of the present invention without departing from the spirit of the invention, and these changes and combinations are within the scope of the invention.
Claims (1)
1. A recommendation method based on a time sequence decision model is characterized by comprising the following steps:
s1, defining the following historical behavior sequence of the user at certain n continuous time: (x)1，x2，......xn)；
S2, representing the behavior sequence prediction of the user in a recursive mode:
xi+1＝Q1xi+Q2x′i (1)
wherein Q is1、Q2Respectively representing parameters needing to be predicted; x is the number ofiInformation representing a last recommendation to the user; x is the number ofi+1The recommendation information which is about to be given to the user at the next moment is shown, namely the target required by the scheme; x'iRepresenting the user's behavior since receiving the last time recommendation;
s3, judging parameter Q1And Q2If the parameter Q needs to be modified, calculating the parameter Q by adopting a decision tree model1And Q2Otherwise, returning to step S2;
the method comprises the following substeps:
s31, confirming whether the parameter Q needs to be modified1And Q2If the modification is needed, executing the step S32, otherwise, not operating; parameter Q1And Q2Whether modification is needed is confirmed by three factors of whether the user role is changed or not, whether the user browsing behavior is changed or not and whether last recommendation is accepted or not;
wherein, the parameter Q1And Q2Conditions that need to be modified include:
(1) changing the role of the user, changing the browsing behavior of the user and accepting the last recommendation;
(2) changing the role of the user, not changing the browsing behavior of the user, and accepting the last recommendation;
(3) the user role is not changed, the user browsing behavior is changed, and the last recommendation is accepted;
(4) the user role is not changed, the user browsing behavior is not changed, and the last recommendation is accepted;
without modifying the parameter Q1And Q2The conditions of (a) include:
(1) changing the role of the user, changing the browsing behavior of the user and not receiving the last recommendation;
(2) changing the role of the user, not changing the browsing behavior of the user, and not receiving the last recommendation;
(3) the role of the user is not changed, the browsing behavior of the user is changed, and the last recommendation is not received;
(4) the role of the user is not changed, the browsing behavior of the user is not changed, and the last recommendation is not received;
s32, determining Q by modeling an optimal objective function1And Q2A value of (d);
the specific implementation method comprises the following steps: the objective function is set as:
min Cost(Id，Action)＝Id(Q1)+Action(Q1，Q2) (2)
wherein,
the second part of the objective function equation, Action () above, is calculated for the value of the behavioral Action:
a represents the complete set of all the different objects that the user may access; assuming a total of i different objects, the set is denoted A1，A2......Ai；R(An+1) Representing a slave object An-1To AnThe correlation value between the two is determined by the object correlation weight matrix;
represents the parameter Q1The value of the n-th prediction,
represents the parameter Q1The value of the (n + 1) th prediction,
represents the parameter Q2The value of the n-th prediction,
represents the parameter Q2The value of the (n + 1) th prediction;
for parameter Q1And Q2Respectively deriving the formula (3) to obtain
And
the values of (A) are as follows:
order to
Obtain a new parameter Q1And Q2The value is obtained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811145391.0ACN109271590B (en) | 2018-09-29 | 2018-09-29 | Recommendation method based on time sequence decision model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811145391.0ACN109271590B (en) | 2018-09-29 | 2018-09-29 | Recommendation method based on time sequence decision model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109271590A CN109271590A (en) | 2019-01-25 |
| CN109271590Btrue CN109271590B (en) | 2021-08-31 |
Family
ID=65198861
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811145391.0AActiveCN109271590B (en) | 2018-09-29 | 2018-09-29 | Recommendation method based on time sequence decision model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109271590B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110727705B (en)* | 2019-10-12 | 2021-05-25 | 腾讯科技（深圳）有限公司 | Information recommendation method and device, electronic equipment and computer-readable storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2011154591A (en)* | 2010-01-28 | 2011-08-11 | Nec Corp | Recommending device, method, and program |
| CN103839167A (en)* | 2012-11-21 | 2014-06-04 | 大连灵动科技发展有限公司 | Commodity candidate set recommendation method |
| CN103902538A (en)* | 2012-12-25 | 2014-07-02 | 中国银联股份有限公司 | Information recommendation device and method based on decision-making tree |
| CN106649884A (en)* | 2017-01-11 | 2017-05-10 | 河南科技大学 | Multimedia content recommendation method based on user situational analysis |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9183497B2 (en)* | 2012-02-23 | 2015-11-10 | Palo Alto Research Center Incorporated | Performance-efficient system for predicting user activities based on time-related features |
- 2018
- 2018-09-29CNCN201811145391.0Apatent/CN109271590B/enactiveActive
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2011154591A (en)* | 2010-01-28 | 2011-08-11 | Nec Corp | Recommending device, method, and program |
| CN103839167A (en)* | 2012-11-21 | 2014-06-04 | 大连灵动科技发展有限公司 | Commodity candidate set recommendation method |
| CN103902538A (en)* | 2012-12-25 | 2014-07-02 | 中国银联股份有限公司 | Information recommendation device and method based on decision-making tree |
| CN106649884A (en)* | 2017-01-11 | 2017-05-10 | 河南科技大学 | Multimedia content recommendation method based on user situational analysis |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109271590A (en) | 2019-01-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Mehrotra et al. | Bandit based optimization of multiple objectives on a music streaming platform | |
| CN113687897B (en) | System and method for proactively providing recommendations to users of computing devices | |
| Jiao et al. | A novel learning rate function and its application on the SVD++ recommendation algorithm | |
| CN105335491A (en) | Method and system for recommending books to users on basis of clicking behavior of users | |
| EP2652909B1 (en) | Method and system for carrying out predictive analysis relating to nodes of a communication network | |
| CN106951436B (en) | Large-scale online recommendation method based on mobile situation | |
| CN111259231A (en) | Recommendation method and device for application program | |
| CN111078858A (en) | Article searching method and device and electronic equipment | |
| CN106779946A (en) | A kind of film recommends method and device | |
| CN119151643A (en) | Commodity recommendation method based on consumer behavior | |
| CN113987363A (en) | Cold start recommendation algorithm based on hidden factor prediction | |
| CN109978575B (en) | Method and device for mining user flow operation scene | |
| CN109271590B (en) | Recommendation method based on time sequence decision model | |
| CN114820085A (en) | User screening method, related device and storage medium | |
| CN110969184A (en) | Directed trajectories through communication decision trees using iterative artificial intelligence | |
| CN115456656A (en) | Method and device for predicting purchase intention of consumer, electronic equipment and storage medium | |
| Mehta et al. | Collaborative personalized web recommender system using entropy based similarity measure | |
| CN114647787B (en) | A user personalized recommendation method based on multimodal data | |
| KR101658714B1 (en) | Method and system for predicting online customer action based on online activity history | |
| CN119295139A (en) | A method, device, terminal and medium for processing product data | |
| CN118313548B (en) | Enterprise energy consumption situation prediction method and system based on time sequence portrait technology | |
| Schmidt | Numerical prediction and sequential process optimization in sheet forming based on genetic algorithm | |
| Kużelewska | Dynamic neighbourhood identification based on multi-clustering in collaborative filtering recommender systems | |
| CN118195672A (en) | Method and device for obtaining multi-target prediction model and product data prediction based on large model | |
| CN115098771B (en) | Recommendation model updating method, recommendation model training method and computing equipment |
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:20210916 Address after:610000 No. 817, floor 8, unit 1, building 12, No. 77, Tianmu Road, high tech Zone, Chengdu, Sichuan Patentee after:Chengdu aoruike Electronic Technology Co.,Ltd. Address before:610041 No.4, Xinhang Road, West Park, high tech Zone, Chengdu, Sichuan Patentee before:SICHUAN LINGLINGQI ROBOT Co.,Ltd. |