CN105824912A - Personalized recommending method and device based on user portrait - Google Patents
Personalized recommending method and device based on user portrait Download PDFInfo
- Publication number
- CN105824912A CN105824912A CN201610147694.0A CN201610147694A CN105824912A CN 105824912 A CN105824912 A CN 105824912A CN 201610147694 A CN201610147694 A CN 201610147694A CN 105824912 A CN105824912 A CN 105824912A
- Authority
- CN
- China
- Prior art keywords
- user
- matrix
- portrait
- shifting
- module
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 30
- 239000011159 matrix material Substances 0.000 claims abstract description 110
- 230000007704 transition Effects 0.000 claims abstract description 26
- 238000007637 random forest analysis Methods 0.000 claims description 33
- 238000004422 calculation algorithm Methods 0.000 claims description 27
- 239000002131 composite material Substances 0.000 claims description 3
- 239000000203 mixture Substances 0.000 description 8
- 238000010586 diagram Methods 0.000 description 4
- 230000003203 everyday effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
Classifications
-
- 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
-
- 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
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
- G06Q30/0269—Targeted advertisements based on user profile or attribute
- G06Q30/0271—Personalized advertisement
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- Strategic Management (AREA)
- Physics & Mathematics (AREA)
- Development Economics (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides a personalized recommendation method based on a user portrait. The method comprises the following steps: acquiring label information of a user, and according to the label information, establishing the user portrait, and acquiring the initial state of the user; according to the user portrait and the initial state, determining a one-step transition matrix of the user, and calculating an interest list of the user according to the one-step transition matrix; performing recommendation according to the interest list. The user portrait and the initial state are used as basis at the same time, so that personalized recommendation is performed for each user, the recommendation accuracy rate is increased, and besides, the problem of cold start by a new user is solved. In addition, the invention further provides a personalized recommendation device based on the user portrait.
Description
Technical field
The present invention relates to computer disposal field, particularly relate to a kind of personalized recommendation method based on user's portrait and device.
Background technology
Along with the development of the Internet, the life of people be more and more closely connected with the Internet together with.In these allegro epoch, user wants to the product being quickly found oneself and being needed by the Internet, but the product data of magnanimity constantly produce every day in the Internet, and this causes Internet user to be difficult to quickly find information oneself needs or interested.In order to allow user quickly find oneself product interested, traditional collaborative filtering is that the history purchase situation according to user goes to recommend, but for new user, often face cold start-up problem, hardly result in accurate recommendation, and the interest for some product user is probably disposably, recommend if simply buying situation according to history, it is recommended that accuracy rate the highest.
Summary of the invention
Based on this, the problem the highest in order to solve above-mentioned recommendation accuracy rate, it is proposed that a kind of based on user portrait personalized recommendation method and device.
A kind of personalized recommendation method based on user's portrait, described method includes: obtain the label information of user；User's portrait is set up according to described label information；Obtain the original state of user；The Matrix of shifting of a step of user is determined according to described user portrait and described original state；The list interested of user is calculated according to described Matrix of shifting of a step；Recommend according to described list interested.
Wherein in an embodiment, described set up the step of user's portrait according to described label information and include: one or more label information of user is formed a text vector；Described text vector is drawn a portrait as the user of user.
Wherein in an embodiment, the step of the described Matrix of shifting of a step determining user according to described user portrait and described original state includes: described user portrait and described original state are combined as an input variable；The algorithm of random forest is used to determine the Matrix of shifting of a step of user according to described input variable.
Wherein in an embodiment, described use the algorithm of random forest to determine according to described input variable the step of Matrix of shifting of a step of user includes: next step transfers to the transition probability of each state to use the algorithm of random forest to calculate user according to described input variable；Transition probability according to each state calculated determines the Matrix of shifting of a step of user.
Wherein in an embodiment, the step of the described list interested calculating user according to described Matrix of shifting of a step includes: the k mated with described user walks transfer matrix to use Markov Chain algorithm to determine according to described Matrix of shifting of a step, wherein, k is the positive integer more than or equal to 1；The list interested of user is calculated according to the described k mated with user step transfer matrix.
A kind of personalized recommendation device based on user's portrait, described device includes: data obtaining module, for obtaining the label information of user；Set up module, for setting up user's portrait according to described label information；State acquisition module, for obtaining the original state of user；First determines module, for determining the Matrix of shifting of a step of user according to described user portrait and described original state；First computing module, for calculating the list interested of user according to described Matrix of shifting of a step；Recommending module, for recommending according to described list interested.
Wherein in an embodiment, described module of setting up is additionally operable to one or more label information of user is formed a text vector, is drawn a portrait as the user of user by described text vector.
Wherein in an embodiment, described first determines that module includes: composite module, for described user portrait and described original state are combined as an input variable；Second determines module, for using the algorithm of random forest to determine the Matrix of shifting of a step of user according to described input variable.
Wherein in an embodiment, described second determines and module is additionally operable to use the algorithm of random forest to calculate user according to described input variable next step transfers to the transition probability of each state, determines the Matrix of shifting of a step of user according to the transition probability of each state calculated.
Wherein in an embodiment, described computing module includes: the 3rd determines module, and the k mated with described user for using Markov Chain algorithm to determine according to described Matrix of shifting of a step walks transfer matrix, and wherein, k is the positive integer more than or equal to 1；Second computing module, for calculating the list interested of user according to the described k mated with user step transfer matrix.
Above-mentioned personalized recommendation method based on user's portrait and device, by obtaining the label information of user, user's portrait is set up according to label information, obtain the original state of user, the Matrix of shifting of a step of user is determined according to user's portrait and original state, then calculate the list interested of user according to Matrix of shifting of a step, recommend finally according to list interested.By determining, with the original state of user, the Matrix of shifting of a step that user is corresponding according to user's portrait, then the list interested of user is determined according to this Matrix of shifting of a step, due to the corresponding exclusive transfer matrix of each user, obtain is also the exclusive list interested of user, the recommendation recommending situation about can be good at according to each user to carry out personalization is carried out according to list interested, improve the accuracy rate of recommendation, simultaneously because this recommendation method is recommended based on user's portrait and original state, new user is also suitable, well solve the cold start-up problem of new user.
Accompanying drawing explanation
Fig. 1 is the flow chart of personalized recommendation method based on user's portrait in an embodiment；
Fig. 2 is the method flow diagram setting up user's portrait in an embodiment；
Fig. 3 is the method flow diagram determining Matrix of shifting of a step in an embodiment；
Fig. 4 is the method flow diagram determining Matrix of shifting of a step in another embodiment；
Fig. 5 is the method flow diagram calculating list interested in an embodiment；
Fig. 6 is the structured flowchart of personalized recommendation device based on user's portrait in an embodiment；
Fig. 7 is the first structured flowchart determining module in an embodiment；
Fig. 8 is the structured flowchart of computing module in an embodiment.
Detailed description of the invention
In order to make the purpose of the present invention, technical scheme and advantage clearer, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein, only in order to explain the present invention, is not intended to limit the present invention.
As it is shown in figure 1, in one embodiment it is proposed that a kind of personalized recommendation method based on user's portrait, the method comprises the following steps:
Step 102, obtains the label information of user.
Concrete, the label information of user can be the build-in attribute of user, it is also possible to be the dynamic attribute of user, it is also possible to be both combinations, can obtain different label informations according to different business scenarios.Wherein, build-in attribute includes the attributes such as the age of user, sex, occupation, and dynamic attribute includes the historical behavior that user buys, the attribute such as record browsing viewing.
Step 104, sets up user's portrait according to label information.
Concrete, user's portrait is a kind of effective tool delineating targeted customer, contact user's demand and design direction.Often with the most plain and closeness to life language, the attribute of user, behavior are connected with expectation during practical operation.In the present embodiment, user's portrait is made up of the multiple label informations obtained, and the multiple label informations obtained are consisted of a text vector, is drawn a portrait as the user of this user by the text vector of composition.
Step 106, obtains the original state of user.
Concrete, the original state of user can be click on certain event, it is also possible to is user's certain event the most browsed, it is also possible to be the behavior state of other users.Such as can being set to original state whether click on certain event, if clicked, being 1, be then 0 without clicking on.Can certainly simultaneously using user to the click situation of multiple events or the situation that browses as original state.
Step 108, determines the Matrix of shifting of a step of user according to user's portrait and original state.
In the present embodiment, the original state user of foundation drawn a portrait and obtain is combined as an input vector, then determines the Matrix of shifting of a step of user according to this input vector.Concrete, user's portrait and original state are combined as an input vector, and using the user of next step click behavior existing as sample, user transfers to the probability of next possible state to use the mode of random forest to predict.Such as, it is assumed that have 100 events (A1 to A100) at present, it would be desirable to predict the transition probability of next step each state of user according to the original state of user's portrait and user.As, investigating whether click the A1 event original state as user, user clicks, and is 1, if user does not click on, is 0.So input variable of user's portrait and original state composition is as shown in table 1:
Table 1
| Label 1 | Label 2 | Label 3 | Label 4 | …… | Label n | Original state A1 | |
| User 1 | Female | 25 years old | White collar | Unmarried | …… | … | 1 |
| User 2 | Man | 40 years old | Blue collar | Married | …… | … | 0 |
Next step user clicking on behavior existing is modeled as sample, it was predicted that user carries out next step probability clicking on each state, and output variable is as shown in table 2:
Table 2
| A2 | A3 | A4 | …… | An | |
| User 1 clicks on probability | 0.1 | 0.2 | 0.2 | …… | 0.1 |
| User 2 clicks on probability | 0.2 | 0.15 | 0.25 | …… | 0.15 |
For each user, all meet P (A2)+P (A3)+...+P (An)=1.Next, consider click probability possible after clicking A2 event successively, generate probability tables as shown above, the like, until setting up the Random Forest model of 100 events (A1-A100), finally obtain the Matrix of shifting of a step of each state according to the Random Forest model set up.In a specific embodiment, user is drawn a portrait and original state is as an input vector, wherein, the number of times that original state can be click on or expose, being predicted according to Random Forest model can essentially be the prediction of a ranking (grade), the part that may constitute for each state is divided into two parts, one is whether to click on, two are click on number of times, and time dimension is taken into account, must be i.e. that the b clicked on again after being a and having clicked on is only state change, could be as transfer matrix.
Step 110, calculates the list interested of user according to Matrix of shifting of a step.
Concrete, the list interested of user is obtained according to Matrix of shifting of a step, here list interested can be the probability that this user is interested in all kinds of article, can also be to screen, through probability interested, the article that the user obtained is interested, it is also possible to be other forms of expression that can embody user interest tendency.
Step 112, recommends according to list interested.
In the present embodiment, after obtaining the list interested of user, recommend according to the list interested of this user.Concrete, the list interested of user describes the probability that user is interested in all kinds of article, such as, endowment insurance 40%, vehicle insurance 90%, accident insurance 80%.So according to this list interested, preferentially recommend vehicle insurance, the accident insurance being most interested in user.
In the present embodiment, above-mentioned personalized recommendation method based on user's portrait, by obtaining the label information of user, user's portrait is set up according to label information, obtain the original state of user, determine the Matrix of shifting of a step of user according to user's portrait and original state, then calculate the list interested of user according to Matrix of shifting of a step, recommend finally according to list interested.By determining, with the original state of user, the Matrix of shifting of a step that user is corresponding according to user's portrait, then the list interested of user is determined according to this Matrix of shifting of a step, due to the corresponding exclusive transfer matrix of each user, obtain is also the exclusive list interested of user, carry out the recommendation recommending situation about can be good at according to each user to carry out personalization according to list interested, improve the accuracy rate of recommendation.Simultaneously because this recommendation method is recommended based on user's portrait and original state, new user is also suitable, well solves the cold start-up problem of new user.
As in figure 2 it is shown, in one embodiment, the step setting up user's portrait according to label information includes:
Step 104a, forms a text vector by one or more label information of user.
Concrete, multiple label informations of the user obtained are formed a long text vector, as shown in table 3:
Table 3
| Label 1 | Label 2 | Label 3 | …… | Label n | |
| User 1 | Man | 28 years old | 6000 yuan | Unmarried | |
| User 2 | Female | 36 years old | 8000 yuan | Married | |
| ……· |
As shown in table 3, the label information of user can include the sex of user, age, income, occupation etc..According to different business scenarios, different label informations can be obtained.
Step 104b, draws a portrait text vector as the user of user.
Concrete, the text vector of user user tag formed is drawn a portrait as the user of user, and user's portrait is as the virtual representations of actual user, and it builds according to product and market often, has reacted feature and the demand of real user.
As it is shown on figure 3, in one embodiment, determine that according to user's portrait and original state the step of the Matrix of shifting of a step of user includes:
Step 108a, is combined as an input variable by user's portrait and original state.
In the present embodiment, determine that according to user's portrait and original state the Matrix of shifting of a step of user is combined as a long text vector particular by by user's portrait together with original state, text vector is substituted into Random Forest model as an input variable, and then prediction user transfers to each shape probability of state.
Step 108b, uses the algorithm of random forest to determine the Matrix of shifting of a step of user according to input variable.
In the present embodiment, according to user's portrait and the input variable of original state composition, use the algorithm of random forest, it was predicted that user next step transfer to the transition probability of each state, obtain the Matrix of shifting of a step of user according to the transition probability obtained.The Matrix of shifting of a step using the algorithm predicts user of random forest is by predicting using next step user clicking on behavior existing as sample, that is, the method is to carry out recommending by the overall probability of crowd of combination, personal attribute and historic state, improves the accuracy rate of recommendation.
As shown in Figure 4, in one embodiment, determine according to the algorithm of input variable employing random forest that the step of the Matrix of shifting of a step of user includes:
Step 402, next step transfers to the transition probability of each state to use the algorithm of random forest to calculate user according to input variable.
In the present embodiment, to have same or analogous user portrait and original state and to have next step user clicking on behavior as sample, next step transfers to the transition probability of each state to use Random Forest model to predict user according to the input variable of user's portrait and original state composition.
Step 404, determines the Matrix of shifting of a step of user according to the transition probability of each state calculated.
Concrete, after the transition probability of each state obtaining user according to Random Forest model, determine, according to transition probability, the Matrix of shifting of a step that user is corresponding.Then the list interested of user is obtained according to Matrix of shifting of a step, recommend finally according to the list interested obtained, in the present embodiment, it is to carry out recommending by the overall probability of crowd of combination, personal attribute and historic state, for the user not having historic state, can also recommend in conjunction with the probability of overall crowd and personal attribute, that is the method is not only applicable to old user and applies also for new user, also solves cold start-up problem while improve recommendation accuracy rate.
As it is shown in figure 5, in one embodiment, the step of the described list interested calculating user according to described Matrix of shifting of a step includes:
Step 110a, the k mated with user walks transfer matrix to use Markov Chain algorithm to determine according to Matrix of shifting of a step, and wherein, k is the positive integer more than or equal to 1.
Concrete, after obtaining the Matrix of shifting of a step of user, Markov Chain algorithm is used to determine the final transfer matrix mated with user, during calculating, the true result of clicking on of predicting the outcome of obtaining and existing sample can be contrasted, determining that the final k mated with user walks transfer matrix, wherein, k is the positive integer more than or equal to 1.The concrete selection by Matrix of shifting of a step being iterated number of times, likely the end-state of some model best suits user's actual click preference, likely some model state after 10 times even 50 times iteration meets user and truly clicks on preference, wherein, two step transfer matrixes be Matrix of shifting of a step square, three step transfer matrixes are the cube of Matrix of shifting of a step, and four step transfer matrixes are the biquadratics of Matrix of shifting of a step, the like.
Step 110b, calculates the list interested of user according to the k step transfer matrix mated with user.
Concrete, after determining the k step transfer matrix finally mated with user, calculate the list interested of user according to this final transfer matrix determined, and then recommend according to the list interested obtained.In the present embodiment, by finding the k mated with legitimate reading to walk transfer matrix, and then calculate list interested according to final k step transfer matrix, recommend finally according to the list interested obtained, further increase the accuracy rate of recommendation.
As shown in Figure 6, in one embodiment it is proposed that a kind of personalized recommendation device based on user's portrait, this device includes:
Data obtaining module 602, for obtaining the label information of user.
Concrete, the label information of user can be the build-in attribute of user, it is also possible to be the dynamic attribute of user, it is also possible to be both combinations, can obtain different label informations according to different business scenarios.Wherein, build-in attribute includes the attributes such as the age of user, sex, occupation, and dynamic attribute includes the historical behavior that user buys, the attribute such as record browsing viewing.
Set up module 604, for setting up user's portrait according to label information.
Concrete, user's portrait is a kind of effective tool delineating targeted customer, contact user's demand and design direction.Often with the most plain and closeness to life language, the attribute of user, behavior are connected with expectation during practical operation.In the present embodiment, user's portrait is made up of the multiple label informations obtained, and the multiple label informations obtained are consisted of a text vector, is drawn a portrait as the user of this user by the text vector of composition.
State acquisition module 606, for obtaining the original state of user.
Concrete, the original state of user can be click on certain event, it is also possible to is user's certain event the most browsed, it is also possible to be the behavior state of other users.Such as can being set to original state whether click on certain event, if clicked, being 1, be then 0 without clicking on.Can certainly simultaneously using user to the click situation of multiple events or the situation that browses as original state.
First determines module 608, for determining the Matrix of shifting of a step of user according to user's portrait and original state.
In the present embodiment, the original state user of foundation drawn a portrait and obtain is combined as an input vector, then determines the Matrix of shifting of a step of user according to this input vector.Concrete, user's portrait and original state are combined as an input vector, and using the user of next step click behavior existing as sample, user transfers to the probability of next possible state to use the mode of random forest to predict.Such as, it is assumed that have 100 events (A1 to A100) at present, it would be desirable to predict the transition probability of next step each state of user according to the original state of user's portrait and user.As, investigating whether click the A1 event original state as user, user clicks, and is 1, if user does not click on, is 0.So input variable of user's portrait and original state composition is as shown in table 1, next step user clicking on behavior existing is modeled as sample, it was predicted that user carries out next step probability clicking on each state, and output variable is as shown in table 2.For each user, all meet P (A2)+P (A3)+...+P (An)=1.Next, consider click probability possible after clicking A2 event successively, generate probability tables as shown above, the like, until setting up the Random Forest model of 100 events (A1-A100), finally obtain the Matrix of shifting of a step of each state according to the Random Forest model set up.In a specific embodiment, user is drawn a portrait and original state is as an input vector, wherein, the number of times that original state can be click on or expose, being predicted according to Random Forest model can essentially be the prediction of a ranking (grade), the part that may constitute for each state is divided into two parts, one is whether to click on, two are click on number of times, and time dimension is taken into account, must be i.e. that the b clicked on again after being a and having clicked on is only state change, could be as transfer matrix.
First computing module 610, for calculating the list interested of user according to Matrix of shifting of a step.
Concrete, the list interested of user is obtained according to Matrix of shifting of a step, here list interested can be the probability that this user is interested in all kinds of article, can also be to screen, through probability interested, the article that the user obtained is interested, it is also possible to be other forms of expression that can embody user interest tendency.
Recommending module 612, for recommending according to list interested.
In the present embodiment, after obtaining the list interested of user, recommend according to the list interested of this user.Concrete, the list interested of user describes the probability that user is interested in all kinds of article, such as, endowment insurance 40%, vehicle insurance 90%, accident insurance 80%.So according to this list interested, preferentially recommend vehicle insurance, the accident insurance being most interested in user.
In the present embodiment, above-mentioned personalized recommendation device based on user's portrait, by obtaining the label information of user, user's portrait is set up according to label information, obtain the original state of user, determine the Matrix of shifting of a step of user according to user's portrait and original state, then calculate the list interested of user according to Matrix of shifting of a step, recommend finally according to list interested.By determining, with the original state of user, the Matrix of shifting of a step that user is corresponding according to user's portrait, then the list interested of user is determined according to this Matrix of shifting of a step, due to the corresponding exclusive transfer matrix of each user, obtain is also the exclusive list interested of user, carry out the recommendation recommending situation about can be good at according to each user to carry out personalization according to list interested, improve the accuracy rate of recommendation.Simultaneously because this recommendation method is recommended based on user's portrait and original state, new user is also suitable, well solves the cold start-up problem of new user.
In one embodiment, set up module 604 and be additionally operable to one or more label information of user is formed a text vector, text vector is drawn a portrait as the user of user.
Concrete, multiple label informations of the user obtained are formed a long text vector, as shown in table 3, the label information of user can include the sex of user, age, income, occupation etc..According to different business scenarios, different label informations can be obtained.The text vector of user user tag formed is drawn a portrait as the user of user, and user's portrait is as the virtual representations of actual user, and it builds according to product and market often, has reacted feature and the demand of real user.
As it is shown in fig. 7, in one embodiment, first determines that module 608 includes:
Composite module 608a, for being combined as an input variable by user's portrait and described original state.
In the present embodiment, determine that according to user's portrait and original state the Matrix of shifting of a step of user is combined as a long text vector particular by by user's portrait together with original state, text vector is substituted into Random Forest model as an input variable, and then prediction user transfers to each shape probability of state.
Second determines module 608b, for using the algorithm of random forest to determine the Matrix of shifting of a step of user according to input variable.
In the present embodiment, according to user's portrait and the input variable of original state composition, use the algorithm of random forest, it was predicted that user next step transfer to the transition probability of each state, obtain the Matrix of shifting of a step of user according to the transition probability obtained.The Matrix of shifting of a step using the algorithm predicts user of random forest is by predicting using next step user clicking on behavior existing as sample, that is, the method is to carry out recommending by the overall probability of crowd of combination, personal attribute and historic state, improves the accuracy rate of recommendation.
In one embodiment, second determines and module 608b is additionally operable to use the algorithm of random forest to calculate user according to input variable next step transfers to the transition probability of each state, determines the Matrix of shifting of a step of user according to the transition probability of each state calculated.
In the present embodiment, to have same or analogous user portrait and original state and to have next step user clicking on behavior as sample, next step transfers to the transition probability of each state to use Random Forest model to predict user according to the input variable of user's portrait and original state composition.After the transition probability of each state obtaining user according to Random Forest model, determine, according to transition probability, the Matrix of shifting of a step that user is corresponding.Then the list interested of user is obtained according to Matrix of shifting of a step, recommend finally according to the list interested obtained, in the present embodiment, it is to carry out recommending by the overall probability of crowd of combination, personal attribute and historic state, for the user not having historic state, can also recommend in conjunction with the probability of overall crowd and personal attribute, that is the method is not only applicable to old user and applies also for new user, also solves cold start-up problem while improve recommendation accuracy rate.
As shown in Figure 8, in one embodiment, computing module 610 includes:
3rd determines module 610a, and the k mated with user for using Markov Chain algorithm to determine according to Matrix of shifting of a step walks transfer matrix, and wherein, k is the positive integer more than or equal to 1.
Concrete, after obtaining the Matrix of shifting of a step of user, Markov Chain algorithm is used to determine the final transfer matrix mated with user, during calculating, the true result of clicking on of predicting the outcome of obtaining and existing sample can be contrasted, determining that the final k mated with user walks transfer matrix, wherein, k is the positive integer more than or equal to 1.The concrete selection by Matrix of shifting of a step being iterated number of times, likely the end-state of some model best suits user's actual click preference, likely some model state after 10 times even 50 times iteration meets user and truly clicks on preference, wherein, two step transfer matrixes be Matrix of shifting of a step square, three step transfer matrixes are the cube of Matrix of shifting of a step, and four step transfer matrixes are the biquadratics of Matrix of shifting of a step, the like.
Second computing module 610b, for calculating the list interested of user according to the k step transfer matrix mated with user.
Concrete, after determining the k step transfer matrix finally mated with user, calculate the list interested of user according to this final transfer matrix determined, and then recommend according to the list interested obtained.In the present embodiment, by finding the k mated with legitimate reading to walk transfer matrix, and then calculate list interested according to final k step transfer matrix, recommend finally according to the list interested obtained, further increase the accuracy rate of recommendation.
Embodiment described above only have expressed the several embodiments of the present invention, and it describes more concrete and detailed, but therefore can not be interpreted as the restriction to the scope of the claims of the present invention.It should be pointed out that, for the person of ordinary skill of the art, without departing from the inventive concept of the premise, it is also possible to make some deformation and improvement, these broadly fall into protection scope of the present invention.Therefore, the protection domain of patent of the present invention should be as the criterion with claims.
Claims (10)
1. a personalized recommendation method based on user's portrait, said method comprising the steps of:
Obtain the label information of user；
User's portrait is set up according to described label information；
Obtain the original state of user；
The Matrix of shifting of a step of user is determined according to described user portrait and described original state；
The list interested of user is calculated according to described Matrix of shifting of a step；
Recommend according to described list interested.
Method the most according to claim 1, it is characterised in that described according to described label information set up user portrait step include:
One or more label information of user is formed a text vector；
Described text vector is drawn a portrait as the user of user.
Method the most according to claim 1, it is characterised in that the step of the described Matrix of shifting of a step determining user according to described user portrait and described original state includes:
Described user portrait and described original state are combined as an input variable；
The algorithm of random forest is used to determine the Matrix of shifting of a step of user according to described input variable.
Method the most according to claim 3, it is characterised in that the described algorithm according to described input variable employing random forest determines that the step of the Matrix of shifting of a step of user includes:
Next step transfers to the transition probability of each state to use the algorithm of random forest to calculate user according to described input variable；
Transition probability according to each state calculated determines the Matrix of shifting of a step of user.
Method the most according to claim 1, it is characterised in that the step of the described list interested calculating user according to described Matrix of shifting of a step includes:
The k mated with described user walks transfer matrix to use Markov Chain algorithm to determine according to described Matrix of shifting of a step, and wherein, k is the positive integer more than or equal to 1；
The list interested of user is calculated according to the described k mated with user step transfer matrix.
6. a personalized recommendation device based on user's portrait, it is characterised in that described device includes:
Data obtaining module, for obtaining the label information of user；
Set up module, for setting up user's portrait according to described label information；
State acquisition module, for obtaining the original state of user；
First determines module, for determining the Matrix of shifting of a step of user according to described user portrait and described original state；
First computing module, for calculating the list interested of user according to described Matrix of shifting of a step；
Recommending module, for recommending according to described list interested.
Device the most according to claim 6, it is characterised in that described module of setting up is additionally operable to one or more label information of user is formed a text vector, is drawn a portrait as the user of user by described text vector.
Device the most according to claim 6, it is characterised in that described first determines that module includes:
Composite module, for being combined as an input variable by described user portrait and described original state；
Second determines module, for using the algorithm of random forest to determine the Matrix of shifting of a step of user according to described input variable.
Device the most according to claim 8, it is characterized in that, described second determines and module is additionally operable to use the algorithm of random forest to calculate user according to described input variable next step transfers to the transition probability of each state, determines the Matrix of shifting of a step of user according to the transition probability of each state calculated.
Device the most according to claim 6, it is characterised in that described computing module includes:
3rd determines module, and the k mated with described user for using Markov Chain algorithm to determine according to described Matrix of shifting of a step walks transfer matrix, and wherein, k is the positive integer more than or equal to 1；
Second computing module, for calculating the list interested of user according to the described k mated with user step transfer matrix.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610147694.0A CN105824912A (en) | 2016-03-15 | 2016-03-15 | Personalized recommending method and device based on user portrait |
| PCT/CN2017/074400 WO2017157146A1 (en) | 2016-03-15 | 2017-02-22 | User portrait-based personalized recommendation method and apparatus, server, and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610147694.0A CN105824912A (en) | 2016-03-15 | 2016-03-15 | Personalized recommending method and device based on user portrait |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105824912A true CN105824912A (en) | 2016-08-03 |
Family
ID=56987258
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610147694.0A Pending CN105824912A (en) | 2016-03-15 | 2016-03-15 | Personalized recommending method and device based on user portrait |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN105824912A (en) |
| WO (1) | WO2017157146A1 (en) |
Cited By (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106097024A (en) * | 2016-08-18 | 2016-11-09 | 北京京东尚科信息技术有限公司 | Data digging method, device and system |
| CN106339505A (en) * | 2016-09-27 | 2017-01-18 | 电子科技大学 | Music recommendation method based on Markov chain |
| CN106354519A (en) * | 2016-09-30 | 2017-01-25 | 乐视控股（北京）有限公司 | Method and device for generating label for user portrait |
| CN106383895A (en) * | 2016-09-27 | 2017-02-08 | 北京金山安全软件有限公司 | Information recommendation method and device and terminal equipment |
| CN106875278A (en) * | 2017-01-19 | 2017-06-20 | 浙江工商大学 | Social network user portrait method based on random forest |
| WO2017157146A1 (en) * | 2016-03-15 | 2017-09-21 | 平安科技（深圳）有限公司 | User portrait-based personalized recommendation method and apparatus, server, and storage medium |
| CN107464186A (en) * | 2017-08-04 | 2017-12-12 | 缪骁 | Claims Resolution pattern based on vehicle insurance customer portrait recommends method |
| CN107679904A (en) * | 2017-09-28 | 2018-02-09 | 山东浪潮云服务信息科技有限公司 | A kind of method and device of Recommendations |
| CN107977864A (en) * | 2017-12-07 | 2018-05-01 | 北京贝塔智投科技有限公司 | A kind of customer insight method and system suitable for financial scenario |
| CN108769159A (en) * | 2018-05-16 | 2018-11-06 | 北京豆果信息技术有限公司 | A kind of electronic cookbook intelligent recommendation method |
| CN108920682A (en) * | 2018-07-11 | 2018-11-30 | 厦门盈趣科技股份有限公司 | Social user's recommended method and device based on machine learning and user's Portrait brand technology |
| CN108960975A (en) * | 2018-06-15 | 2018-12-07 | 广州麦优网络科技有限公司 | Personalized Precision Marketing Method, server and storage medium based on user's portrait |
| CN109710842A (en) * | 2018-12-17 | 2019-05-03 | 泰康保险集团股份有限公司 | Method for pushing, device and the readable storage medium storing program for executing of business information |
| CN109788056A (en) * | 2019-01-10 | 2019-05-21 | 四川新网银行股份有限公司 | User's theme message method for pushing and system based on clustering |
| CN110335060A (en) * | 2019-05-20 | 2019-10-15 | 微民保险代理有限公司 | Product information method for pushing, device, storage medium and computer equipment |
| CN110413882A (en) * | 2019-07-15 | 2019-11-05 | 阿里巴巴集团控股有限公司 | Information-pushing method, device and equipment |
| CN110852852A (en) * | 2019-11-15 | 2020-02-28 | 成都航天科工大数据研究院有限公司 | Industrial Internet product recommendation system and method |
| CN110852825A (en) * | 2019-08-12 | 2020-02-28 | 陈柏臻 | Method for customizing product fragrance formula based on user preference |
| CN111191142A (en) * | 2018-11-14 | 2020-05-22 | 腾讯科技（深圳）有限公司 | Electronic resource recommendation method and device and readable medium |
| CN111522828A (en) * | 2020-04-23 | 2020-08-11 | 中国农业银行股份有限公司 | User portrait label value analysis method and device |
| CN111767455A (en) * | 2019-05-09 | 2020-10-13 | 北京京东尚科信息技术有限公司 | Information pushing method and device |
| CN111767290A (en) * | 2019-04-01 | 2020-10-13 | 北京百度网讯科技有限公司 | Method and apparatus for updating a user representation |
| CN112685641A (en) * | 2020-12-31 | 2021-04-20 | 五八有限公司 | Information processing method and device |
| CN112862567A (en) * | 2021-02-25 | 2021-05-28 | 华侨大学 | Exhibit recommendation method and system for online exhibition |
| CN113313545A (en) * | 2021-04-19 | 2021-08-27 | 深圳市竹芒信息技术有限公司 | Information recommendation method and device, computer equipment and storage medium |
| CN113672818A (en) * | 2020-05-13 | 2021-11-19 | 中南大学 | Method and system for obtaining user portrait of social media |
| CN114168862A (en) * | 2021-12-14 | 2022-03-11 | 平安国际智慧城市科技股份有限公司 | Product recommendation method, device, equipment and storage medium based on collaborative filtering |
| CN116258356A (en) * | 2023-05-16 | 2023-06-13 | 深圳市万物云科技有限公司 | Work order dispatching method and device based on WaveNet and related medium |
Families Citing this family (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108521439B (en) * | 2018-03-08 | 2022-01-21 | 北京三快在线科技有限公司 | Message pushing method and device |
| CN110866040B (en) * | 2018-08-27 | 2024-05-17 | 京东科技控股股份有限公司 | User portrait generation method, device and system |
| CN109493199A (en) * | 2019-01-04 | 2019-03-19 | 深圳壹账通智能科技有限公司 | Products Show method, apparatus, computer equipment and storage medium |
| CN111666503B (en) * | 2019-03-05 | 2023-08-18 | 百度在线网络技术（北京）有限公司 | Travel mode recommendation method and device, computer equipment and readable storage medium |
| CN110222272B (en) * | 2019-04-18 | 2022-10-14 | 广东工业大学 | Potential customer mining and recommending method |
| CN110334274A (en) * | 2019-05-30 | 2019-10-15 | 平安科技（深圳）有限公司 | Information-pushing method, device, computer equipment and storage medium |
| CN110543594B (en) * | 2019-08-06 | 2022-03-25 | 山东科技大学 | Knowledge base-based personalized evidence-based correction recommendation method for prisoners |
| CN112579876A (en) * | 2019-09-30 | 2021-03-30 | 北京京东尚科信息技术有限公司 | Information pushing method, device and system based on user interest and storage medium |
| CN112752256B (en) * | 2019-10-31 | 2022-06-03 | 中移(苏州)软件技术有限公司 | Client portrait label determination method, device, equipment and storage medium |
| CN112749323B (en) * | 2019-10-31 | 2024-07-23 | 北京沃东天骏信息技术有限公司 | Method and device for constructing user portrait |
| CN111091351A (en) * | 2019-12-16 | 2020-05-01 | 北京政信1890智能科技有限公司 | User portrait construction method and device, electronic equipment and readable storage medium |
| CN111611483B (en) * | 2020-05-11 | 2022-07-22 | 腾讯科技（深圳）有限公司 | Object portrait construction method, device and equipment and storage medium |
| CN111753151B (en) * | 2020-06-24 | 2023-09-15 | 广东科杰通信息科技有限公司 | Service recommendation method based on Internet user behavior |
| CN111782959B (en) * | 2020-07-20 | 2023-07-14 | 深圳市与飞科技有限公司 | User portrait updating method, device, computer equipment and storage medium |
| CN112182391A (en) * | 2020-09-30 | 2021-01-05 | 北京神州泰岳智能数据技术有限公司 | User portrait drawing method and device |
| CN112270350B (en) * | 2020-10-23 | 2023-11-21 | 泰康保险集团股份有限公司 | Method, apparatus, device and storage medium for portraying organization |
| CN112395331B (en) * | 2020-11-17 | 2023-10-10 | 平安科技（深圳）有限公司 | User portrait method, device, equipment and medium for credit card customer |
| CN112417301A (en) * | 2020-12-07 | 2021-02-26 | 中国建设银行股份有限公司 | Information pushing method, device, server and storage medium |
| CN112417315A (en) * | 2020-12-15 | 2021-02-26 | 深圳壹账通智能科技有限公司 | User portrait generation method, device, equipment and medium based on website registration |
| CN112581246A (en) * | 2020-12-23 | 2021-03-30 | 上海永骁智能技术有限公司 | Differentiation method and device based on deep learning tax service |
| CN112910953B (en) * | 2021-01-14 | 2023-02-17 | 中国工商银行股份有限公司 | Business data pushing method and device and server |
| CN113010795A (en) * | 2021-04-12 | 2021-06-22 | 北京明略软件系统有限公司 | User dynamic portrait generation method, system, storage medium and electronic device |
| CN113592306A (en) * | 2021-07-30 | 2021-11-02 | 北京壹心壹翼科技有限公司 | Intelligent matching method, device, equipment and medium based on full-flow user portrait |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6655963B1 (en) * | 2000-07-31 | 2003-12-02 | Microsoft Corporation | Methods and apparatus for predicting and selectively collecting preferences based on personality diagnosis |
| CN101826114B (en) * | 2010-05-26 | 2012-05-09 | 南京大学 | Multi Markov chain-based content recommendation method |
| CN103353872B (en) * | 2013-06-03 | 2017-03-01 | 大连理工大学 | A kind of teaching resource personalized recommendation method based on neutral net |
| CN105824912A (en) * | 2016-03-15 | 2016-08-03 | 平安科技（深圳）有限公司 | Personalized recommending method and device based on user portrait |
-
2016
- 2016-03-15 CN CN201610147694.0A patent/CN105824912A/en active Pending
-
2017
- 2017-02-22 WO PCT/CN2017/074400 patent/WO2017157146A1/en active Application Filing
Cited By (41)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017157146A1 (en) * | 2016-03-15 | 2017-09-21 | 平安科技（深圳）有限公司 | User portrait-based personalized recommendation method and apparatus, server, and storage medium |
| CN106097024B (en) * | 2016-08-18 | 2021-11-02 | 北京京东尚科信息技术有限公司 | Data mining method, device and system |
| CN106097024A (en) * | 2016-08-18 | 2016-11-09 | 北京京东尚科信息技术有限公司 | Data digging method, device and system |
| CN106339505A (en) * | 2016-09-27 | 2017-01-18 | 电子科技大学 | Music recommendation method based on Markov chain |
| CN106383895A (en) * | 2016-09-27 | 2017-02-08 | 北京金山安全软件有限公司 | Information recommendation method and device and terminal equipment |
| CN106339505B (en) * | 2016-09-27 | 2019-09-27 | 电子科技大学 | A kind of music recommended method based on markov chain |
| CN106383895B (en) * | 2016-09-27 | 2019-08-27 | 北京金山安全软件有限公司 | Information recommendation method and device and terminal equipment |
| CN106354519A (en) * | 2016-09-30 | 2017-01-25 | 乐视控股（北京）有限公司 | Method and device for generating label for user portrait |
| CN106875278A (en) * | 2017-01-19 | 2017-06-20 | 浙江工商大学 | Social network user portrait method based on random forest |
| CN106875278B (en) * | 2017-01-19 | 2020-11-03 | 浙江工商大学 | Social network user image drawing method based on random forest |
| CN107464186A (en) * | 2017-08-04 | 2017-12-12 | 缪骁 | Claims Resolution pattern based on vehicle insurance customer portrait recommends method |
| CN107679904A (en) * | 2017-09-28 | 2018-02-09 | 山东浪潮云服务信息科技有限公司 | A kind of method and device of Recommendations |
| CN107977864A (en) * | 2017-12-07 | 2018-05-01 | 北京贝塔智投科技有限公司 | A kind of customer insight method and system suitable for financial scenario |
| CN108769159A (en) * | 2018-05-16 | 2018-11-06 | 北京豆果信息技术有限公司 | A kind of electronic cookbook intelligent recommendation method |
| CN108960975A (en) * | 2018-06-15 | 2018-12-07 | 广州麦优网络科技有限公司 | Personalized Precision Marketing Method, server and storage medium based on user's portrait |
| CN108920682A (en) * | 2018-07-11 | 2018-11-30 | 厦门盈趣科技股份有限公司 | Social user's recommended method and device based on machine learning and user's Portrait brand technology |
| CN111191142A (en) * | 2018-11-14 | 2020-05-22 | 腾讯科技（深圳）有限公司 | Electronic resource recommendation method and device and readable medium |
| CN111191142B (en) * | 2018-11-14 | 2023-03-14 | 腾讯科技（深圳）有限公司 | Electronic resource recommendation method and device and readable medium |
| CN109710842B9 (en) * | 2018-12-17 | 2022-08-19 | 泰康保险集团股份有限公司 | Method and device for pushing service information and readable storage medium |
| CN109710842A (en) * | 2018-12-17 | 2019-05-03 | 泰康保险集团股份有限公司 | Method for pushing, device and the readable storage medium storing program for executing of business information |
| CN109710842B (en) * | 2018-12-17 | 2021-06-29 | 泰康保险集团股份有限公司 | Business information pushing method and device and readable storage medium |
| CN109788056A (en) * | 2019-01-10 | 2019-05-21 | 四川新网银行股份有限公司 | User's theme message method for pushing and system based on clustering |
| CN111767290A (en) * | 2019-04-01 | 2020-10-13 | 北京百度网讯科技有限公司 | Method and apparatus for updating a user representation |
| CN111767290B (en) * | 2019-04-01 | 2024-04-19 | 北京百度网讯科技有限公司 | Method and apparatus for updating user portraits |
| CN111767455A (en) * | 2019-05-09 | 2020-10-13 | 北京京东尚科信息技术有限公司 | Information pushing method and device |
| CN110335060A (en) * | 2019-05-20 | 2019-10-15 | 微民保险代理有限公司 | Product information method for pushing, device, storage medium and computer equipment |
| CN110413882B (en) * | 2019-07-15 | 2023-10-31 | 创新先进技术有限公司 | Information pushing method, device and equipment |
| CN110413882A (en) * | 2019-07-15 | 2019-11-05 | 阿里巴巴集团控股有限公司 | Information-pushing method, device and equipment |
| CN110852825A (en) * | 2019-08-12 | 2020-02-28 | 陈柏臻 | Method for customizing product fragrance formula based on user preference |
| CN110852852A (en) * | 2019-11-15 | 2020-02-28 | 成都航天科工大数据研究院有限公司 | Industrial Internet product recommendation system and method |
| CN111522828B (en) * | 2020-04-23 | 2023-08-01 | 中国农业银行股份有限公司 | User portrait tag value analysis method and device |
| CN111522828A (en) * | 2020-04-23 | 2020-08-11 | 中国农业银行股份有限公司 | User portrait label value analysis method and device |
| CN113672818A (en) * | 2020-05-13 | 2021-11-19 | 中南大学 | Method and system for obtaining user portrait of social media |
| CN113672818B (en) * | 2020-05-13 | 2023-11-14 | 中南大学 | Method and system for acquiring social media user portraits |
| CN112685641A (en) * | 2020-12-31 | 2021-04-20 | 五八有限公司 | Information processing method and device |
| CN112862567A (en) * | 2021-02-25 | 2021-05-28 | 华侨大学 | Exhibit recommendation method and system for online exhibition |
| CN113313545A (en) * | 2021-04-19 | 2021-08-27 | 深圳市竹芒信息技术有限公司 | Information recommendation method and device, computer equipment and storage medium |
| CN113313545B (en) * | 2021-04-19 | 2024-04-23 | 深圳市竹芒信息技术有限公司 | Information recommendation method, device, computer equipment and storage medium |
| CN114168862A (en) * | 2021-12-14 | 2022-03-11 | 平安国际智慧城市科技股份有限公司 | Product recommendation method, device, equipment and storage medium based on collaborative filtering |
| CN116258356A (en) * | 2023-05-16 | 2023-06-13 | 深圳市万物云科技有限公司 | Work order dispatching method and device based on WaveNet and related medium |
| CN116258356B (en) * | 2023-05-16 | 2023-08-25 | 深圳市万物云科技有限公司 | Work order dispatching method and device based on WaveNet and related medium |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2017157146A1 (en) | 2017-09-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105824912A (en) | Personalized recommending method and device based on user portrait | |
| Ghose et al. | Modeling consumer footprints on search engines: An interplay with social media | |
| US10783361B2 (en) | Predictive analysis of target behaviors utilizing RNN-based user embeddings | |
| Adomavicius et al. | Personalization technologies: a process-oriented perspective | |
| US10275818B2 (en) | Next generation improvements in recommendation systems | |
| Acilar et al. | A collaborative filtering method based on artificial immune network | |
| CN110222272A (en) | A kind of potential customers excavate and recommended method | |
| US20180357669A1 (en) | System and method for information processing | |
| Gogna et al. | DiABlO: Optimization based design for improving diversity in recommender system | |
| CN104317945A (en) | E-commerce website commodity recommending method on basis of search behaviors | |
| Liu et al. | Mobile commerce product recommendations based on hybrid multiple channels | |
| CN105809558A (en) | Social network based recommendation method and apparatus | |
| CN103824213A (en) | Individualized recommendation method based on user preferences and commodity properties | |
| CN108073667B (en) | Method for generating user browsing attributes, and non-transitory computer readable medium | |
| US20180232794A1 (en) | Method for collaboratively filtering information to predict preference given to item by user of the item and computing device using the same | |
| Aggarwal et al. | Context-sensitive recommender systems | |
| Koniew | Classification of the User's Intent Detection in Ecommerce systems-Survey and Recommendations. | |
| US20140200996A1 (en) | Information distribution device and information distribution method | |
| US20120004957A1 (en) | Forecasting supply for advertisements according to a non-parametric supply model | |
| US20100049586A1 (en) | Method for determining an advertising slate based on an expected utility | |
| Kim et al. | RETRACTED: An empirical test to measure the effectiveness of online advertising in online marketplaces using a hierarchical Bayes model | |
| Zhang et al. | A context-aware recommendation system based on latent factor model | |
| Michalis et al. | Recommendation systems: Bridging technical aspects with marketing implications | |
| Wang et al. | Meta-analyses using information reweighting: An application to online advertising | |
| Park | Recommender technologies and emerging applications |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20160803 |
|
| RJ01 | Rejection of invention patent application after publication |