CN105809464A - Method and device for information delivery - Google Patents
Method and device for information delivery Download PDFInfo
- Publication number
- CN105809464A CN105809464A CN201410850655.8A CN201410850655A CN105809464A CN 105809464 A CN105809464 A CN 105809464A CN 201410850655 A CN201410850655 A CN 201410850655A CN 105809464 A CN105809464 A CN 105809464A
- Authority
- CN
- China
- Prior art keywords
- user
- interest
- category
- information
- information type
- 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
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method and device for information delivery and belongs to the communication field. The method comprises steps that keywords of users are extracted according to the total data of the users; a cluster algorithm is employed to determined corresponding interest categories of the users according to the keywords of the users; information delivery to the users is carried out according to the corresponding interest categories of the users and a relation data table between the interest categories and information types. According to the method, interests of the users can be accurately determined according to the total data of the users, so information delivery to the users can be more accurately carried out, and information delivery precision is improved.
Description
Technical field
The present invention relates to the communications field, particularly to a kind of information distribution method and device.
Background technology
Internet advertising is put on market huge, but the accurate advertisement jettison system of personalization is relatively fewer, and especially the accurate advertisement jettison system of mobile Internet is less.Mostly the advertisement putting that current Internet firm realizes is to rely on the user behavior data that search engine is huge, and user's internet behavior is carried out depth analysis, and then specific user colony throws in different advertisement.
Although prior art can throw in different advertisement for specific user colony, but the fineness ratio being to discriminate between is relatively thick, is still unsatisfactory for the demand that advertisement accurately is thrown in.
Summary of the invention
One purpose of the embodiment of the present invention is: improve the precision that information is thrown in.
An aspect according to embodiments of the present invention, it is provided that a kind of information distribution method, including: according to user's full dose data extract user key word；Key word according to user adopts clustering algorithm to determine the category of interest that user is corresponding；The relation database table of the category of interest corresponding according to user and category of interest and information type is to this user's impression information.
The present invention can determine the interest of user more accurately according to the full dose data of user, and then more accurately to this user's impression information, can improve the precision that information is thrown in.
In one embodiment, user's full dose data include user behavior data, location data, customer consumption information, user base information, third-party external data.
Throwing in advertisement relative to prior art only in accordance with data such as user's internet behaviors, the present invention can realize the accurate input of information according to more more fully data of aforementioned list.Owing to operator can utilize intrinsic advantage can get these full dose data deriving from user, the information being therefore especially suitable for operation aspect is thrown in, good market prospects.
In one embodiment, the key word according to user's full dose data extraction user includes: user's full dose data are converted to textual form；User's full dose data of textual form are carried out participle；Word segmentation result is screened the key word obtaining user.Textual form can make the process of user's full dose data be more prone to and efficiently.
In one embodiment, word segmentation result is screened the key word obtaining user to include: from word segmentation result, remove stop words；And/or, from word segmentation result, remove semantic dittograph language；And/or, calculate the weight of each word segmentation result according to tf-idf-chi algorithm, filter out, according to weight, the key word that each category of interest can be represented.Thus selecting more accurate and representational key word, being conducive to reducing data processing amount, and making the user interest determined more accurate, improving the precision that information is thrown in further.
In one embodiment, it is determined that category of interest corresponding to user also includes: calculate the similarity between each category of interest that user is corresponding, if the similarity of different category of interest is more than the threshold value set, a category of interest is therefrom selected.So that the interest of the user determined is more representative and accurate, it is to avoid information is unnecessary, avoids to a certain extent and be repeated to the user the information that propelling movement is identical.
In one embodiment, the method also includes: adopt tf-idf-chi algorithm to determine the preference value of each category of interest corresponding to user according to the frequency of the key word of user.Preference value can reflect the size of user's class interest.
In one embodiment, include to this user's impression information with the relation database table of information type according to category of interest corresponding to user and category of interest: the category of interest corresponding according to user and the relation database table of preference value and category of interest and information type thereof are to this user's impression information.Come to user's impression information in conjunction with interest and preference size, it is possible to improve the precision that information is thrown in further.
In one embodiment, include to this user's impression information: the relation database table of the category of interest corresponding according to user and category of interest and information type determines the information type of recommendation；Preference value according to the information type recommended and at least one category of interest corresponding to this information type, it is determined that the recommendation of this information type；Recommendation according to each information type is to this user's impression information.According to can the recommendation of recommendation information come to user's impression information, it is possible to improve the precision that information is thrown in further.
In one embodiment, further can according to the preference value of category of interest corresponding to the similarity of the corresponding category of interest of information type recommended and this information type, it is determined that the recommendation of this information type.The method makes the recommendation calculated more accurate, thus improving the precision that information is thrown in further.
In one embodiment, the method also includes: for the key word of product class, if product be there occurs purchasing behavior by user, judge that this product is whether in default life cycle, if this product is in default life cycle, from the category of interest that this user is corresponding, remove the category of interest belonging to this product.For the particularity of product, in its life cycle, no longer push the information of this product, it is to avoid push out-of-date demand to user, improve precision and real-time that information is thrown in.If the key word of non-electrical business's series products (such as news category, academic space key word), then all need not remove from the interest list of user in spite of in life cycle.
In one embodiment, the information of input includes advertisement.Internet advertising is put on market huge, has good economic benefit, and the information distribution method of the present invention can apply to advertisement.
Another further aspect according to embodiments of the present invention, it is provided that a kind of information delivery device, including: key word information acquisition module, for extracting the key word of user according to user's full dose data；Interest information determines module, adopts clustering algorithm to determine the category of interest that user is corresponding for the key word according to user；Information putting module, for the relation database table according to category of interest corresponding to user and category of interest and information type to this user's impression information.
In one embodiment, user's full dose data include user behavior data, location data, customer consumption information, user base information, third-party external data.
In one embodiment, key word information acquisition module includes: text conversion units, for user's full dose data are converted to textual form；Participle unit, for carrying out participle to user's full dose data of textual form；Screening unit, for screening the key word obtaining user to word segmentation result.
In one embodiment, screen unit, specifically for: from word segmentation result, remove stop words；And/or, from word segmentation result, remove semantic dittograph language；And/or, calculate the weight of each word segmentation result according to tf-idf-chi algorithm, filter out, according to weight, the key word that each category of interest can be represented.
In one embodiment, interest information determines module, is additionally operable to: calculate the similarity between each category of interest that user is corresponding, if the similarity of different category of interest is more than the threshold value set, therefrom selects a category of interest.
In one embodiment, interest information determines module, is additionally operable to: adopt tf-idf-chi algorithm to determine the preference value of each category of interest corresponding to user according to the frequency of the key word of user.
In one embodiment, information putting module, specifically for: the relation database table of the category of interest corresponding according to user and preference value and category of interest and information type to this user's impression information.
In one embodiment, information putting module, the information type of recommendation is determined specifically for the relation database table of the category of interest corresponding according to user and category of interest and information type；Preference value according to the information type recommended and at least one category of interest corresponding to this information type, it is determined that the recommendation of this information type；Recommendation according to each information type is to this user's impression information.
In one embodiment, information putting module is when determining recommendation, specifically for the preference value of category of interest corresponding to the similarity according to the corresponding category of interest of information type recommended and this information type, it is determined that the recommendation of this information type.
In one embodiment, this device also includes: feedback module, for the key word for product class, if product be there occurs purchasing behavior by user, judge that this product is whether in default life cycle, if this product is in default life cycle, illustrate that user will not again buy this product in this life cycle, then can remove the category of interest belonging to this product from category of interest corresponding to this user.If the key word of non-electrical business's series products (such as news category, academic space key word), then all need not remove from the interest list of user in spite of in life cycle.
By referring to the accompanying drawing detailed description to the exemplary embodiment of the present invention, the further feature of the present invention and advantage thereof will be made apparent from.
Accompanying drawing explanation
In order to be illustrated more clearly that the embodiment of the present invention or technical scheme of the prior art, the accompanying drawing used required in embodiment or description of the prior art will be briefly described below, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, for those of ordinary skill in the art, under the premise not paying creative work, it is also possible to obtain other accompanying drawing according to these accompanying drawings.
Fig. 1 is the schematic flow sheet of one embodiment of information distribution method of the present invention.
Fig. 2 is a kind of schematic flow sheet realizing method that the present invention extracts the key word of user according to user's full dose data.
Fig. 3 is the present invention according to the relation database table of category of interest corresponding to user and preference value and category of interest and information type to a kind of exemplary schematic flow sheet realizing method of this user's impression information.
Fig. 4 be information distribution method further embodiment of the present invention realize schematic diagram.
Fig. 5 is the structural representation of one embodiment of information delivery device of the present invention.
Fig. 6 is the structural representation of information delivery device further embodiment of the present invention.
Detailed description of the invention
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is clearly and completely described, it is clear that described embodiment is only a part of embodiment of the present invention, rather than whole embodiments.Description only actually at least one exemplary embodiment is illustrative below, never as any restriction to the present invention and application or use.Based on the embodiment in the present invention, the every other embodiment that those of ordinary skill in the art obtain under not making creative work premise, broadly fall into the scope of protection of the invention.
Unless specifically stated otherwise, the parts otherwise set forth in these embodiments and positioned opposite, the numerical expression of step and numerical value do not limit the scope of the invention.
Simultaneously, it should be appreciated that for the ease of describing, the size of the various piece shown in accompanying drawing is not draw according to actual proportionate relationship.
The known technology of person of ordinary skill in the relevant, method and apparatus are likely to be not discussed in detail, but in the appropriate case, described technology, method and apparatus should be considered to authorize a part for description.
Shown here with in all examples discussed, any occurrence should be construed as merely exemplary, not as restriction.Therefore, other example of exemplary embodiment can have different values.
It should also be noted that similar label and letter below figure represent similar terms, therefore, once a certain Xiang Yi accompanying drawing is defined, then it need not be further discussed in accompanying drawing subsequently.
Fig. 1 is the flow chart of one embodiment of information distribution method of the present invention.As it is shown in figure 1, the method for this embodiment includes:
S102, extracts the key word of user according to user's full dose data.Key word can reflect the interest preference in other words of user.
S104, adopts clustering algorithm according to the key word of user, such as K-means algorithm, it is determined that the category of interest that user is corresponding.
S106, the relation database table of the category of interest corresponding according to user and category of interest and information type is to this user's impression information.
The embodiment of the present invention can determine the interest of user more accurately according to the full dose data of user, and then more accurately to this user's impression information, can improve the precision that information is thrown in.Wherein, the information of input includes but not limited to advertisement.Internet advertising is put on market huge, has good economic benefit, and the information distribution method of the present invention can apply to advertisement or other need the field to user's pushed information.
In one embodiment, user's full dose data such as can include user behavior data, location data, customer consumption information, user base information, third-party external data etc..These Data Sources are in DPI (DeepPacketInspection, deep packet resolve) system, CRM (CustomerRelationshipManagement, customer relation management) system, EDA (EnterpriseDataArchitects, enterprise data architecture) system etc..In addition, operator can also depend on sms center and get the propelling movement short message of third party (such as some electricity business).
Throwing in advertisement relative to prior art only in accordance with data such as user's internet behaviors, the present invention can realize the accurate input of information according to more more fully data of aforementioned list.Owing to operator can utilize intrinsic advantage can get these full dose data deriving from user, the information being therefore especially suitable for operation aspect is thrown in, good market prospects.
In above-mentioned steps S102, with reference to Fig. 2, realize method according to the one of the key word of user's full dose data extraction user and include:
S1022, the data form of user's full dose data is likely to difference, for instance, it is possible to there are the data of the forms such as textual form, graphic form, voice and video, are therefore converted to textual form by unified for user's full dose data.Textual form can make the process of user's full dose data be more prone to and efficiently.
Wherein, marked content according to the content of its artificial mark or automatic marking, can be converted to text by the data of the form such as picture, voice and video.It addition, picture can also adopt Gauss model to be labeled, marked content is also converted into text.
User's full dose data of textual form are carried out participle by S1024.It is for instance possible to use ICTCLAS (Chinese lexical analysis system) Chinese word segmentation.
S1026, screens the key word obtaining user to word segmentation result.
In above-mentioned steps S1026, word segmentation result is screened the one of the key word obtaining user and realizes method and include: step S1026A, and/or, step S1026B, and/or, step S1026C.
S1026A, removes stop words from word segmentation result.For example, it is possible to remove incoherent auxiliary words of mood etc. can not represent the vocabulary of user interest preference.
S1026B, removes semantic dittograph language from word segmentation result.Such as, electric fan, fan etc., remove semantic dittograph language, only retain one of which saying.
S1026C, calculates the weight of each word segmentation result according to tf-idf-chi algorithm, filters out, according to weight, the key word that can represent each category of interest.
The key word filtered out can add corpus, it is possible to improves corpus further, for instance key word is added or deletion etc. operates.So far, it is possible to set up an one-dimensional lists of keywords for each user, user interest preference can be described very subtly by the key word in list, depict " portrait " of each user.
Key word screening operation can select more accurate and representational key word, is conducive to reducing data processing amount, and makes the user interest determined more accurate, improves the precision that information is thrown in further.
In above-mentioned steps S104, determine that category of interest corresponding to user also includes: calculate the similarity between each category of interest that user is corresponding, if the similarity-rough set of different category of interest is big, for instance more than a certain threshold value set, then therefrom select a category of interest.So that the interest of the user determined is more representative and accurate, it is to avoid information is unnecessary, avoids to a certain extent and be repeated to the user the information that propelling movement is identical.
Wherein, the similarity between each category of interest that user is corresponding can according to key word corresponding to category of interest and adopt similarity algorithm to calculate.Similarity algorithm such as can adopt correction cosine relevancy algorithm (that is, Cosine similarity), and formula is expressed as follows:
Wherein, the key word vector of category of interest A be set to [A1, A2 ..., An], the key word vector of category of interest B be set to [B1, B2 ..., Bn].
In above-mentioned steps S106, the category of interest corresponding according to user and category of interest with the relation database table of information type to the first exemplary method that realizes of this user's impression information be: according to category of interest corresponding to user from category of interest with the relation database table of information type finds corresponding information type, according to this information type to this user's impression information.
In above-mentioned steps S106, the category of interest corresponding according to user and category of interest with the relation database table of information type to the method that realizes that the second of this user's impression information is exemplary be: the category of interest corresponding according to user and the relation database table of preference value and category of interest and information type thereof are to this user's impression information.Come to user's impression information in conjunction with interest and preference size, it is possible to improve the precision that information is thrown in further.Wherein, tf-idf-chi algorithm is adopted to may determine that the preference value of each category of interest corresponding to user according to the frequency of the key word of user.Preference value can reflect the size of user's class interest.
With reference to Fig. 3, the relation database table of the above-mentioned category of interest corresponding according to user and preference value and category of interest and information type specifically includes to the exemplary method that realizes of the one of this user's impression information:
S1062, the relation database table of the category of interest corresponding according to user and category of interest and information type determines the information type of recommendation.Specifically, utilize category of interest corresponding to user from category of interest with the relation database table of information type finds corresponding information type.
S1064, the preference value according to the information type recommended and at least one category of interest corresponding to this information type, it is determined that the recommendation of this information type.
S1066, according to the recommendation of each information type to this user's impression information.Specifically, it is possible to the recommendation of each information type is ranked up, corresponding information is thrown according to the recommendation after sequence to this user.
Above-mentioned basis can the recommendation of recommendation information be come to user's impression information, it is possible to improves the precision that information is thrown in further.
In above-mentioned steps S1064, determine that the first illustrative methods of recommendation includes: for all category of interest that the information type item recommended is corresponding, the preference value of each category of interest corresponding with this information type for the information type item of recommendation is done multiplying, then adds and be averaged the recommendation that can obtain this information type.
In above-mentioned steps S1064, it is determined that the second illustrative methods of recommendation includes: the preference value according to category of interest corresponding to the similarity of the corresponding category of interest of information type recommended and this information type, it is determined that the recommendation of this information type.Specifically, for all category of interest that the information type item recommended is corresponding, the similarity of each category of interest corresponding for information type item and the preference value of this category of interest are done multiplying, then adds and be averaged the recommendation that can obtain this information type.This method makes the recommendation calculated more accurate, thus improving the precision that information is thrown in further.
Wherein, information type such as can adopt Euclidean distance (EuclideanDistance), Cosine similarity (CosineSimilarity), Pearson correlation coefficients (PearsonCorrelationCoefficient) etc. to calculate with the similarity of category of interest.The computational methods of Cosine similarity, with reference to aforementioned, repeat no more here.
Euclidean distance is originally used for calculating the distance of point-to-point transmission in Euclidean space, and the present invention, when representing similarity with it, adopts below equation to change:
Above-mentioned distance d is more little, and similarity sim is more big.
Pearson correlation coefficients is generally used for and calculates the tightness degree of contact between two spacing variablees, and value is in [-1,1], and formula is expressed as follows:
In one embodiment, information distribution method also includes: for the key word of product class, if product be there occurs purchasing behavior by user, judge that this product is whether in default life cycle, if this product is in default life cycle, from the category of interest that this user is corresponding, remove the category of interest belonging to this product.Particularity for product, user will not again buy this product in the life cycle of product, therefore no longer pushes the information of this product in its life cycle, it is to avoid push out-of-date demand to user, improve precision and real-time that information is thrown in, be conducive to improving the clicking rate of advertisement.If the key word of non-electrical business's series products (such as news category, academic space key word), then all need not remove from the interest list of user in spite of in life cycle.
For example, for buying the user of mobile phone, if in the life cycle of half a year, the advertisement that this user's recommending mobile phone is relevant is not just given；Having crossed after life cycle duration just can recommending mobile phone class is relevant again advertisement.
It addition, different life cycle can be arranged for different products, for instance if user's purchase is mobile phone, then life cycle is likely half a year；If what user bought is furniture sofa, it is 3 years that life cycle is likely.
Fig. 4 be information distribution method further embodiment of the present invention realize schematic diagram.As shown in Figure 4, the method main process of this embodiment includes:
Obtain users's full dose data (S402) such as user base information, position data, consumption data and user behavior data (as the record that surfs the web, search records, advertisement is mutual)；After converting the data into text, File Format Data is carried out pretreatment (S404), pretreatment mainly includes word segmentation processing (S4042), extracts key word (S4044), keyword weight sequence (S4046) etc., it is also possible to include stop words or duplicate removal etc.；The key word that can reflect user interest or preference is added corpus, thus setting up corpus (S406)；Using the lists of keywords corresponding for the user input (S408) as certain clustering algorithm (such as K-means algorithm), the category of interest that output user is corresponding, and can also determine that the preference value of category of interest is in the lump as output according to user's key word frequency, as the foundation (S410) that accurate information is thrown in.Using the relation database table of user interest tables of data and interest and information as inputting (S412), collaborative filtering (with specific reference to S106 and implement) based on user interest carries out processing (S414), output is by the information type (S416) of recommendation sequence, and throws in the information such as advertisement (S418) accordingly to user.In combination with feedback mechanism, avoid repeating this user to throw in same advertisement as far as possible.This mechanism by feedback user to the click of thrown in advertisement or purchasing behavior (S420), judge whether to have searched for the key word (S422) of this product in electricity business or search engine, if, then further determine whether in life cycle (S424), if, the interest pattern (S426) belonging to this product is removed from the category of interest recommended, otherwise, if not searching for the key word of this product or not in life cycle, then still can recommend the interest pattern belonging to this product.
Shown in the user interest preference table reference table 1 related in the various embodiments described above.
Table 1 user interest preference table
Shown in the relation database table reference table 2 of the category of interest related in the various embodiments described above and information type.
The relation database table of table 2 category of interest and information type
It addition, the user profile related in the various embodiments described above (table 3) and interest (table 4) can also adopt the form record of tables of data to store.
Table 3 user message table
| ID | User content |
| … | … |
Table 4 interest table
| Interest ID | Category of interest |
| … | … |
Push process based on above-mentioned each data table information to specifically include that
(a) input active user ID, first from user interest preference table, isolate the interest ID of recommended interest ID and active user, be separately added into recommendation interest set and the current user interest set (including the preference value of interest corresponding to active user) of correspondence
B () does calculated as below for all recommended interest set elements: the Similarity value inquiring about between the user interest of recommendation and information from the relation database table of category of interest and information type, and the preference value corresponding with user interest is multiplied, its meansigma methods is as the recommendation of current recommendation information, the recommendation of all recommendation informations can be obtained, it is ranked up according to recommendation, it is thus achieved that the information recommendation list of active user.Then, corresponding information is pushed according to the information recommendation list of user to user.
Another further aspect according to embodiments of the present invention, it is provided that a kind of information delivery device.With reference to Fig. 5, the device 500 of the present embodiment includes:
Key word information acquisition module 502, for extracting the key word of user according to user's full dose data；
Interest information determines module 504, adopts clustering algorithm to determine the category of interest that user is corresponding for the key word according to user；
Information putting module 506, for the relation database table according to category of interest corresponding to user and category of interest and information type to this user's impression information.
In one embodiment, user's full dose data include user behavior data, location data, customer consumption information, user base information, third-party external data.
In one embodiment, key word information acquisition module 502 includes: text conversion units, for user's full dose data are converted to textual form；Participle unit, for carrying out participle to user's full dose data of textual form；Screening unit, for screening the key word obtaining user to word segmentation result.
In one embodiment, screen unit, specifically for: from word segmentation result, remove stop words；And/or, from word segmentation result, remove semantic dittograph language；And/or, calculate the weight of each word segmentation result according to tf-idf-chi algorithm, filter out, according to weight, the key word that each category of interest can be represented.
In one embodiment, interest information determines module 504, is additionally operable to: calculate the similarity between each category of interest that user is corresponding, if the similarity of different category of interest is more than the threshold value set, therefrom selects a category of interest.
In one embodiment, interest information determines module 504, is additionally operable to: adopt tf-idf-chi algorithm to determine the preference value of each category of interest corresponding to user according to the frequency of the key word of user.
In one embodiment, information putting module 506, specifically for: the relation database table of the category of interest corresponding according to user and preference value and category of interest and information type to this user's impression information.
In one embodiment, information putting module 506, the information type of recommendation is determined specifically for the relation database table of the category of interest corresponding according to user and category of interest and information type；Preference value according to the information type recommended and at least one category of interest corresponding to this information type, it is determined that the recommendation of this information type；Recommendation according to each information type is to this user's impression information.
In one embodiment, information putting module 506 is when determining recommendation, specifically for the preference value of category of interest corresponding to the similarity according to the corresponding category of interest of information type recommended and this information type, it is determined that the recommendation of this information type.
In one embodiment, with reference to Fig. 6, this device 500 also includes: feedback module 608, for the key word for product class, if product be there occurs purchasing behavior by user, it is judged that whether this product is in default life cycle, if this product is in default life cycle, illustrate that user will not again buy this product in this life cycle, then can remove the category of interest belonging to this product from category of interest corresponding to this user.If the key word of non-electrical business's series products (such as news category, academic space key word), then all need not remove from the interest list of user in spite of in life cycle.
One of ordinary skill in the art will appreciate that all or part of step realizing above-described embodiment can be completed by hardware, can also be completed by the hardware that program carrys out instruction relevant, described program can be stored in a kind of computer-readable recording medium, storage medium mentioned above can be read only memory, disk or CD etc..
The foregoing is only presently preferred embodiments of the present invention, not in order to limit the present invention, all within the spirit and principles in the present invention, any amendment of making, equivalent replacement, improvement etc., should be included within protection scope of the present invention.
Claims (22)
1. an information distribution method, it is characterised in that including:
The key word of user is extracted according to user's full dose data；
Key word according to user adopts clustering algorithm to determine the category of interest that user is corresponding；
The relation database table of the category of interest corresponding according to user and category of interest and information type is to this user's impression information.
2. method according to claim 1, it is characterised in that described user's full dose data include user behavior data, location data, customer consumption information, user base information, third-party external data.
3. method according to claim 1, it is characterised in that the described key word according to user's full dose data extraction user includes:
User's full dose data are converted to textual form；
User's full dose data of textual form are carried out participle；
Word segmentation result is screened the key word obtaining user.
4. method according to claim 3, it is characterised in that the described key word obtaining user that word segmentation result is screened includes:
Stop words is removed from word segmentation result；And/or,
Semantic dittograph language is removed from word segmentation result；And/or,
Calculate the weight of each word segmentation result according to tf-idf-chi algorithm, filter out, according to weight, the key word that each category of interest can be represented.
5. method according to claim 1, it is characterised in that determine that category of interest corresponding to user also includes:
Calculate the similarity between each category of interest that user is corresponding, if the similarity of different category of interest is more than the threshold value set, therefrom select a category of interest.
6. method according to claim 1, it is characterised in that also include:
The frequency of the key word according to user adopts tf-idf-chi algorithm to determine the preference value of each category of interest corresponding to user.
7. method according to claim 6, it is characterised in that the relation database table of the described category of interest corresponding according to user and category of interest and information type includes to this user's impression information:
The relation database table of the category of interest corresponding according to user and preference value and category of interest and information type is to this user's impression information.
8. method according to claim 7, it is characterised in that include to this user's impression information:
The relation database table of the category of interest corresponding according to user and category of interest and information type determines the information type of recommendation；
Preference value according to the information type recommended and at least one category of interest corresponding to this information type, it is determined that the recommendation of this information type；
Recommendation according to each information type is to this user's impression information.
9. method according to claim 8, it is characterised in that the preference value according to category of interest corresponding to the similarity of the corresponding category of interest of information type recommended and this information type, it is determined that the recommendation of this information type.
10. method according to claim 1, it is characterised in that also include:
For the key word of product class, if product be there occurs purchasing behavior by user, it is judged that this product is whether in default life cycle, if this product is in default life cycle, from the category of interest that this user is corresponding, remove the category of interest belonging to this product.
11. according to the method described in any one of claim 1-10, it is characterised in that the information of input includes advertisement.
12. an information delivery device, it is characterised in that including:
Key word information acquisition module, for extracting the key word of user according to user's full dose data；
Interest information determines module, adopts clustering algorithm to determine the category of interest that user is corresponding for the key word according to user；
Information putting module, for the relation database table according to category of interest corresponding to user and category of interest and information type to this user's impression information.
13. device according to claim 12, it is characterised in that described user's full dose data include user behavior data, location data, customer consumption information, user base information, third-party external data.
14. device according to claim 12, it is characterised in that described key word information acquisition module includes:
Text conversion units, for being converted to textual form by user's full dose data；
Participle unit, for carrying out participle to user's full dose data of textual form；
Screening unit, for screening the key word obtaining user to word segmentation result.
15. device according to claim 14, it is characterised in that described screening unit, specifically for:
Stop words is removed from word segmentation result；And/or,
Semantic dittograph language is removed from word segmentation result；And/or,
Calculate the weight of each word segmentation result according to tf-idf-chi algorithm, filter out, according to weight, the key word that each category of interest can be represented.
16. device according to claim 12, it is characterised in that described interest information determines module, is additionally operable to:
Calculate the similarity between each category of interest that user is corresponding, if the similarity of different category of interest is more than the threshold value set, therefrom select a category of interest.
17. device according to claim 12, it is characterised in that described interest information determines module, is additionally operable to:
The frequency of the key word according to user adopts tf-idf-chi algorithm to determine the preference value of each category of interest corresponding to user.
18. device according to claim 17, it is characterised in that described information putting module, specifically for:
The relation database table of the category of interest corresponding according to user and preference value and category of interest and information type is to this user's impression information.
19. device according to claim 18, it is characterised in that described information putting module, specifically for:
The relation database table of the category of interest corresponding according to user and category of interest and information type determines the information type of recommendation；
Preference value according to the information type recommended and at least one category of interest corresponding to this information type, it is determined that the recommendation of this information type；
Recommendation according to each information type is to this user's impression information.
20. device according to claim 19, it is characterized in that, described information putting module is when determining recommendation, specifically for the preference value of category of interest corresponding to the similarity according to the corresponding category of interest of information type recommended and this information type, it is determined that the recommendation of this information type.
21. device according to claim 12, it is characterised in that also include:
Feedback module, for the key word for product class, if product be there occurs purchasing behavior by user, it is judged that whether this product is in default life cycle, if this product is in default life cycle, from the category of interest that this user is corresponding, remove the category of interest belonging to this product.
22. according to the device described in any one of claim 12-21, it is characterised in that the information of input includes advertisement.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410850655.8A CN105809464A (en) | 2014-12-31 | 2014-12-31 | Method and device for information delivery |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410850655.8A CN105809464A (en) | 2014-12-31 | 2014-12-31 | Method and device for information delivery |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105809464A true CN105809464A (en) | 2016-07-27 |
Family
ID=56420548
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410850655.8A Pending CN105809464A (en) | 2014-12-31 | 2014-12-31 | Method and device for information delivery |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105809464A (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106529993A (en) * | 2016-10-20 | 2017-03-22 | 天脉聚源（北京）科技有限公司 | Intelligent advertisement delivery method and device |
| CN107480124A (en) * | 2017-07-05 | 2017-12-15 | 小草数语（北京）科技有限公司 | Advertisement placement method and device |
| CN107563810A (en) * | 2017-08-31 | 2018-01-09 | 江苏省公用信息有限公司 | A kind of advertisement placement method based on broadband account |
| CN108156228A (en) * | 2017-12-15 | 2018-06-12 | 维沃移动通信有限公司 | A kind of ad data method for pushing and device |
| CN108280689A (en) * | 2018-01-30 | 2018-07-13 | 浙江省公众信息产业有限公司 | Advertisement placement method, device based on search engine and search engine system |
| CN108446964A (en) * | 2018-03-30 | 2018-08-24 | 中南大学 | A kind of user's recommendation method based on mobile flow DPI data |
| CN108494849A (en) * | 2018-03-20 | 2018-09-04 | 石家庄正和网络有限公司 | A kind of accurate method for pushing in recycling website and system |
| CN108572966A (en) * | 2017-03-09 | 2018-09-25 | 神州数码系统集成服务有限公司 | A kind of method and device creating personal portrait |
| CN109840820A (en) * | 2017-11-28 | 2019-06-04 | 北京京东尚科信息技术有限公司 | A kind of method and apparatus of determining user to object type preference |
| CN109872195A (en) * | 2019-03-08 | 2019-06-11 | 北京搜狐新媒体信息技术有限公司 | A kind of data processing method and device |
| CN110020200A (en) * | 2019-03-15 | 2019-07-16 | 微梦创科网络科技（中国）有限公司 | A kind of personalized recommendation method and system based on history material |
| CN110428285A (en) * | 2019-07-31 | 2019-11-08 | 安徽抖范视频科技有限公司 | A kind of advertisement filter method and filter device based on web advertisement type |
| CN110706021A (en) * | 2019-09-12 | 2020-01-17 | 微梦创科网络科技（中国）有限公司 | Advertisement putting method and system |
| CN111178919A (en) * | 2018-11-09 | 2020-05-19 | 北京沃东天骏信息技术有限公司 | Information delivery method, device, equipment and computer readable storage medium |
| CN117216803A (en) * | 2023-11-09 | 2023-12-12 | 成都乐超人科技有限公司 | Intelligent finance-oriented user information protection method and system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101505284A (en) * | 2008-11-12 | 2009-08-12 | 中国移动通信集团广东有限公司 | Intelligent information dynamic pushing method and system |
| CN101520878A (en) * | 2009-04-03 | 2009-09-02 | 华为技术有限公司 | Method, device and system for pushing advertisements to users |
| CN103399883A (en) * | 2013-07-19 | 2013-11-20 | 百度在线网络技术（北京）有限公司 | Method and system for performing personalized recommendation according to user interest points/concerns |
| CN103678710A (en) * | 2013-12-31 | 2014-03-26 | 同济大学 | Information recommendation method based on user behaviors |
| CN104111946A (en) * | 2013-04-19 | 2014-10-22 | 腾讯科技（深圳）有限公司 | Clustering method and device based on user interests |
-
2014
- 2014-12-31 CN CN201410850655.8A patent/CN105809464A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101505284A (en) * | 2008-11-12 | 2009-08-12 | 中国移动通信集团广东有限公司 | Intelligent information dynamic pushing method and system |
| CN101520878A (en) * | 2009-04-03 | 2009-09-02 | 华为技术有限公司 | Method, device and system for pushing advertisements to users |
| CN104111946A (en) * | 2013-04-19 | 2014-10-22 | 腾讯科技（深圳）有限公司 | Clustering method and device based on user interests |
| CN103399883A (en) * | 2013-07-19 | 2013-11-20 | 百度在线网络技术（北京）有限公司 | Method and system for performing personalized recommendation according to user interest points/concerns |
| CN103678710A (en) * | 2013-12-31 | 2014-03-26 | 同济大学 | Information recommendation method based on user behaviors |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106529993A (en) * | 2016-10-20 | 2017-03-22 | 天脉聚源（北京）科技有限公司 | Intelligent advertisement delivery method and device |
| CN108572966A (en) * | 2017-03-09 | 2018-09-25 | 神州数码系统集成服务有限公司 | A kind of method and device creating personal portrait |
| CN107480124A (en) * | 2017-07-05 | 2017-12-15 | 小草数语（北京）科技有限公司 | Advertisement placement method and device |
| CN107563810A (en) * | 2017-08-31 | 2018-01-09 | 江苏省公用信息有限公司 | A kind of advertisement placement method based on broadband account |
| CN109840820A (en) * | 2017-11-28 | 2019-06-04 | 北京京东尚科信息技术有限公司 | A kind of method and apparatus of determining user to object type preference |
| CN108156228A (en) * | 2017-12-15 | 2018-06-12 | 维沃移动通信有限公司 | A kind of ad data method for pushing and device |
| CN108280689A (en) * | 2018-01-30 | 2018-07-13 | 浙江省公众信息产业有限公司 | Advertisement placement method, device based on search engine and search engine system |
| CN108494849A (en) * | 2018-03-20 | 2018-09-04 | 石家庄正和网络有限公司 | A kind of accurate method for pushing in recycling website and system |
| CN108446964A (en) * | 2018-03-30 | 2018-08-24 | 中南大学 | A kind of user's recommendation method based on mobile flow DPI data |
| CN108446964B (en) * | 2018-03-30 | 2022-03-22 | 中南大学 | User recommendation method based on mobile traffic DPI data |
| CN111178919A (en) * | 2018-11-09 | 2020-05-19 | 北京沃东天骏信息技术有限公司 | Information delivery method, device, equipment and computer readable storage medium |
| CN111178919B (en) * | 2018-11-09 | 2023-12-08 | 北京沃东天骏信息技术有限公司 | Information delivery method, device, equipment and computer readable storage medium |
| CN109872195A (en) * | 2019-03-08 | 2019-06-11 | 北京搜狐新媒体信息技术有限公司 | A kind of data processing method and device |
| CN110020200A (en) * | 2019-03-15 | 2019-07-16 | 微梦创科网络科技（中国）有限公司 | A kind of personalized recommendation method and system based on history material |
| CN110428285A (en) * | 2019-07-31 | 2019-11-08 | 安徽抖范视频科技有限公司 | A kind of advertisement filter method and filter device based on web advertisement type |
| CN110706021A (en) * | 2019-09-12 | 2020-01-17 | 微梦创科网络科技（中国）有限公司 | Advertisement putting method and system |
| CN117216803A (en) * | 2023-11-09 | 2023-12-12 | 成都乐超人科技有限公司 | Intelligent finance-oriented user information protection method and system |
| CN117216803B (en) * | 2023-11-09 | 2024-02-09 | 成都乐超人科技有限公司 | Intelligent finance-oriented user information protection method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105809464A (en) | Method and device for information delivery | |
| CN104394118B (en) | A kind of method for identifying ID and system | |
| CN102279851B (en) | Intelligent navigation method, device and system | |
| CN102591854B (en) | For advertisement filtering system and the filter method thereof of text feature | |
| CN101853295B (en) | Image search method | |
| CN103116588A (en) | Method and system for personalized recommendation | |
| CN102215300A (en) | Telecommunication service recommendation method and system | |
| CN107025310A (en) | A kind of automatic news in real time recommends method | |
| CN103049440A (en) | Recommendation processing method and processing system for related articles | |
| CN105868267B (en) | A kind of modeling method of mobile social networking user interest | |
| CN106874374A (en) | A kind of recommendation method for pushing based on user's history behavior interaction analysis | |
| CN105844424A (en) | Product quality problem discovery and risk assessment method based on network comments | |
| CN107632984A (en) | A kind of cluster data table shows methods, devices and systems | |
| CN104035927A (en) | User behavior-based search method and system | |
| CN105740468A (en) | Individuation recommendation method and system combined with content publisher information | |
| CN104484377B (en) | Replace dictionary generating method and device | |
| CN104008184A (en) | Method and device for pushing information | |
| CN101980199A (en) | Method and system for discovering network hot topic based on situation assessment | |
| CN103729360A (en) | Interest label recommendation method and system | |
| CN101609459A (en) | A kind of extraction system of affective characteristic words | |
| CN107330050A (en) | A kind of personalized recommendation method and system | |
| CN102880687A (en) | Personal interactive data retrieval method and system based on tag technology | |
| CN106168980A (en) | Multimedia resource recommends sort method and device | |
| CN104636371A (en) | Information recommendation method and device | |
| CN103377249A (en) | Keyword putting method and system |
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: 20160727 |
|
| RJ01 | Rejection of invention patent application after publication |