CN112148973B - Data processing method and device for information push - Google Patents
Data processing method and device for information push Download PDFInfo
- Publication number
- CN112148973B CN112148973B CN202010921701.4A CN202010921701A CN112148973B CN 112148973 B CN112148973 B CN 112148973B CN 202010921701 A CN202010921701 A CN 202010921701A CN 112148973 B CN112148973 B CN 112148973B
- Authority
- CN
- China
- Prior art keywords
- information
- user
- prediction model
- determining
- target user
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000003672 processing method Methods 0.000 title claims abstract description 36
- 238000000034 method Methods 0.000 claims abstract description 25
- 230000000694 effects Effects 0.000 claims description 48
- 238000012549 training Methods 0.000 claims description 30
- 238000004891 communication Methods 0.000 claims description 15
- 238000012545 processing Methods 0.000 claims description 15
- 238000006243 chemical reaction Methods 0.000 claims description 12
- 238000012795 verification Methods 0.000 claims description 10
- 238000004590 computer program Methods 0.000 claims description 5
- 238000004458 analytical method Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 4
- 239000000463 material Substances 0.000 abstract description 2
- 230000008901 benefit Effects 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 230000009471 action Effects 0.000 description 3
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000003066 decision tree Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- PCHJSUWPFVWCPO-UHFFFAOYSA-N gold Chemical compound [Au] PCHJSUWPFVWCPO-UHFFFAOYSA-N 0.000 description 2
- 239000010931 gold Substances 0.000 description 2
- 229910052737 gold Inorganic materials 0.000 description 2
- 238000007477 logistic regression Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000002093 peripheral effect Effects 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 230000000153 supplemental effect Effects 0.000 description 2
- RWSOTUBLDIXVET-UHFFFAOYSA-N Dihydrogen sulfide Chemical compound S RWSOTUBLDIXVET-UHFFFAOYSA-N 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000000802 evaporation-induced self-assembly Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The application relates to a data processing method and a device for information push, wherein the method comprises the following steps: determining a target user needing information pushing by using a determining module; the method comprises the steps that a record acquisition module is used for acquiring first history record information corresponding to a target user; determining the user type and the user characteristics of the target user according to the first historical record information by using a user information determining module; the application information acquisition module acquires candidate push information corresponding to the target user according to the user characteristics; a model acquisition module is used for acquiring a pre-trained prediction model corresponding to the user type; and obtaining a predicted recommendation result corresponding to the candidate push information by using a result obtaining module. According to the application, all characteristic information of the user can be combined during recommendation, rather than sorting according to one attribute, a more accurate recommendation result can be obtained, and the information click rate and the order forming rate of the pushing user are improved; meanwhile, grouping and sorting are not needed, so that the recommendation efficiency can be greatly improved, and the cost of manpower and material resources is saved.
Description
Technical Field
The application relates to the technical field of Internet, in particular to a data processing method and device for information push.
Background
The activity recommendation, namely, carrying out activity display related to products or contents on different resource positions aiming at a user browsing the APP, so as to achieve a certain operation conversion purpose.
At present, the activity display needs to be configured and adjusted by special operators, after the users are classified into different types according to priority, the users are manually matched and displayed during the activity, and different activities are displayed to different layered users. When large-scale activities are required to be held, more users are involved, the complexity of products and the diversity of activities are involved, and a great deal of manpower and time are required to discuss the activity display priority and the front crowd layering. When the activity is replaced and iterated, the process is carried out again, so that the efficiency of the activity display iteration is low and the display error is easy to cause because of logic loopholes.
In addition, although the campaign recommendation has some more mature development on the advertisement exposure flow and some APP advertisement positions, in the non-advertisement field and some scenes needing to combine the finance and financial properties of users, the campaign recommendation related technologies mostly adopt front-end user grouping, are shown at the front end of the resource positions after being manually ordered according to the priority, and have not yet appeared the technology for automatically recommending the campaign.
As can be seen from the above, in the prior art, the front-end users are used for sorting in groups, which requires a lot of manpower, so that the efficiency is low; in addition, users are grouped only through different attribute information of the users, so that all characteristics of the users cannot be considered in the advertisement pushing mode, the pertinence is low, and the problem of low click rate of advertisements is further caused.
Aiming at a plurality of technical problems existing in the related art, no effective solution is provided at present.
Disclosure of Invention
In order to solve the technical problems or at least partially solve the technical problems, the application provides a data processing method and device for information push.
In a first aspect, an embodiment of the present application provides a data processing method for information push, including:
determining a target user needing information pushing by using a determining module;
the application record acquisition module acquires first history record information corresponding to the target user;
determining the user type and the user characteristics of the target user according to the first historical record information by using a user information determining module;
an application information acquisition module acquires candidate push information corresponding to the target user according to the user characteristics;
A model acquisition module is used for acquiring a pre-trained prediction model corresponding to the user type;
the application result acquisition module is used for acquiring an expected recommendation result corresponding to the candidate push information according to the user characteristics, the push information characteristics and the prediction model; the push information features are used for characterizing the candidate push information.
Optionally, as with the data processing methods previously described,
the application information acquisition module acquires candidate push information corresponding to the target user according to the user characteristics, and the application information acquisition module comprises:
matching by using a matching unit according to the user characteristics to obtain an associated user which meets the preset matching degree with the target user;
acquiring second history information corresponding to the associated user by using a first acquisition unit;
and determining the candidate push information interacted with the associated user according to the second historical record information by using a candidate push information determining unit.
Optionally, in the foregoing data processing method, the determining, by using a user information determining module, a user type of the target user according to the first history information includes:
a second obtaining unit is used for obtaining the feature type and the activity degree of the target user according to the first history information; the feature types include: demographic attributes, financial characteristics, click-to-transaction conversion characteristics, historical browsing characteristics, real-time search characteristics, real-time browsing characteristics; the liveness includes: transaction frequency, click frequency;
Analyzing the feature types and the liveness by using an analysis unit through a preset user type division strategy to obtain a life cycle corresponding to the target user;
and determining the user type according to the life cycle corresponding to the target user by using a user type determining unit.
Optionally, in the foregoing data processing method, the candidate push information includes at least two, and the method further includes:
calculating the estimated recommendation result corresponding to each candidate push information according to a preset calculation strategy by using a score acquisition module to obtain a recommendation score corresponding to each candidate push information; the predicted recommended results include: click probability, transaction probability, and predicted transaction amount;
pushing the N candidate push information with the highest recommendation scores to the target user by using a push module; wherein N is an integer not less than 1.
Optionally, in the foregoing data processing method, the pushing module is configured to push the N candidate push messages with the highest recommendation scores to the target user, where the pushing module includes:
uploading target pushing information to an activity pool corresponding to the target user by using a resource bit display unit; the target user obtains target pushing information in the activity pool through a user side; the N candidate push information with the highest recommendation score comprises the target push information;
And respectively determining keyword information in each target push information by using a transmitting unit, writing the keyword information into a corresponding position in a preset template, generating push information corresponding to each target push information, and transmitting the push information to the target user.
Optionally, in the foregoing data processing method, the application result obtaining module obtains an estimated recommendation result of the candidate push information according to the user feature, the push information feature and the prediction model, including:
a target prediction model determining unit is used for configuring the prediction model according to the push information characteristics to obtain a target prediction model corresponding to the candidate push information;
and the application result determining unit inputs the user characteristics into the target prediction model to predict, so as to obtain the predicted recommendation result.
Optionally, the data processing method as described above further includes:
a third acquisition unit is used for acquiring third historical record information corresponding to the candidate user according to a preset training data acquisition strategy; the third history information is not used for training the prediction model, the candidate user comprises the target user, and the candidate user is consistent with the user type of the target user;
Randomly dividing the third historical record information by using a dividing unit to obtain training set supplementary data and verification set supplementary data;
and training the prediction model through the training set supplementary data by using a first updating unit, and obtaining a first updated prediction model when the trained prediction model is verified through the verification set supplementary data and the preset performance requirement is met.
Optionally, the data processing method as described above further includes:
acquiring newly uploaded information to be processed by using a fourth acquisition unit;
determining the characteristics to be processed in the information to be processed by using a characteristics to be processed determining unit;
determining new features which are not in the prediction model in the features to be processed by using a new feature determination unit;
and applying a second updating unit to perform feature addition on the prediction model according to the new features to obtain a second updated prediction model.
Optionally, a data processing method as described above further includes
In a second aspect, an embodiment of the present application provides a data processing apparatus for information push, including:
the determining module is used for determining a target user needing information pushing;
the record acquisition module is used for acquiring first history record information corresponding to the target user;
The user information determining module is used for determining the user type and the user characteristics of the target user according to the first historical record information;
the information acquisition module is used for acquiring candidate push information corresponding to the target user according to the user characteristics;
the model acquisition module is used for acquiring a pre-trained prediction model corresponding to the user type;
the result acquisition module is used for acquiring a predicted recommendation result corresponding to the candidate push information according to the user characteristics, the push information characteristics and the prediction model; the push information features are used for characterizing the candidate push information.
Optionally, in the foregoing data processing apparatus, the information acquisition module includes:
the matching unit is used for matching according to the user characteristics to obtain an associated user which meets the preset matching degree with the target user;
the first acquisition unit is used for acquiring second history record information corresponding to the associated user;
and the candidate push information determining unit is used for determining the candidate push information interacted with the associated user according to the second historical record information.
Optionally, the data processing apparatus as described above, the user information determining module includes:
The second acquisition unit is used for acquiring the feature type and the liveness of the target user according to the first history information; the feature types include: demographic attributes, financial characteristics, click-to-transaction conversion characteristics, historical browsing characteristics, real-time search characteristics, real-time browsing characteristics; the liveness includes: transaction frequency, click frequency;
the analysis unit is used for analyzing the characteristic types and the liveness through a preset user type division strategy to obtain a life cycle corresponding to the target user;
and the user type determining unit is used for determining the user type according to the life cycle corresponding to the target user.
Optionally, in the foregoing data processing apparatus, the candidate push information includes at least two candidate push information, and further includes:
the score acquisition module is used for calculating the expected recommendation result corresponding to each candidate push information according to a preset calculation strategy to obtain a recommendation score corresponding to each candidate push information; the predicted recommended results include: click probability, transaction probability, and predicted transaction amount;
the pushing module is used for pushing the N candidate pushing information with the highest recommendation scores to the target user; wherein N is an integer not less than 1.
Optionally, the data processing apparatus as described above, the result obtaining module includes:
the target prediction model determining unit is used for configuring the prediction model according to the push information characteristics to obtain a target prediction model corresponding to the candidate push information;
and the result determining unit is used for inputting the user characteristics into the target prediction model for prediction to obtain the predicted recommendation result.
Optionally, the data processing apparatus as described above further includes: a first model updating module; the first model updating module includes:
the third acquisition unit is used for acquiring third historical record information corresponding to the candidate user according to a preset training data acquisition strategy; the third history information is not used for training the prediction model, the candidate user comprises the target user, and the candidate user is consistent with the user type of the target user;
the dividing unit is used for randomly dividing the third historical record information to obtain training set supplementary data and verification set supplementary data;
the first updating unit is used for training the prediction model through the training set supplementary data, and obtaining a first updated prediction model when the trained prediction model is verified through the verification set supplementary data and the preset performance requirement is met.
Optionally, the data processing apparatus as described above further includes: a second model updating module; the second model updating module includes:
a fourth obtaining unit, configured to obtain newly uploaded information to be processed;
the to-be-processed feature determining unit is used for determining to-be-processed features in the to-be-processed information;
a new feature determining unit configured to determine new features that are not present in the prediction model among the features to be processed;
and the second updating unit is used for adding the characteristics of the prediction model according to the new characteristics to obtain a second updated prediction model.
In a third aspect, an embodiment of the present application provides an electronic device, including: the device comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory are communicated with each other through the communication bus;
the memory is used for storing a computer program;
the processor is configured to implement a processing method according to any one of the preceding claims when executing the computer program.
In a fourth aspect, an embodiment of the present application provides a non-transitory computer readable storage medium, wherein the non-transitory computer readable storage medium stores computer instructions that cause the computer to perform the processing method according to any one of the preceding claims.
Compared with the prior art, the technical scheme provided by the embodiment of the application has the following advantages: all characteristic information of the users can be combined during recommendation, rather than sorting according to one attribute, more accurate recommendation results can be obtained, and the information click rate and the order forming rate of the pushing users are improved; meanwhile, grouping and sorting are not needed, so that the recommendation efficiency can be greatly improved, and the cost of manpower and material resources is saved.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and together with the description, serve to explain the principles of the application.
In order to more clearly illustrate the embodiments of the application or the technical solutions of the prior art, the drawings which are used in the description of the embodiments or the prior art will be briefly described, and it will be obvious to a person skilled in the art that other drawings can be obtained from these drawings without inventive effort.
Fig. 1 is a flow chart of a data processing method for information push according to an embodiment of the present application;
fig. 2 is a schematic flow chart of a data processing method for information push according to another embodiment of the present application;
Fig. 3 is a schematic flow chart of a data processing method for information push according to another embodiment of the present application;
FIG. 4 is a block diagram of a data processing device for pushing information according to an embodiment of the present application;
fig. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present application more apparent, the technical solutions of the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application, and it is apparent that the described embodiments are some embodiments of the present application, but not all embodiments of the present application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
Fig. 1 is a data processing method for information push provided in an embodiment of the present application, including steps S1 to S6 as follows:
s1, determining a target user needing information pushing by using a determining module.
Specifically, the target user is a user who needs to push information by the method in this embodiment, and in general, since the features of each user are different, the information to be pushed is different.
S2, acquiring first history information corresponding to the target user by using a record acquisition module.
Specifically, the operation performed by the user on the pushed information may be recorded and stored in a data table corresponding to the user according to a time sequence, and the history information of each user may include, but is not limited to: login time, offline time, browse records, order records, and so forth. The first history information is history information corresponding to the target user, and thus the first history information may also include, but is not limited to, the above-described information.
And S3, determining the user type and the user characteristics of the target user according to the first historical record information by using a user information determining module.
Specifically, after the first history information is obtained, the first history information may be analyzed and categorized according to the following user characteristics, including but not limited to: for related page view records of different types of activities, user liveness (e.g. number of log-in per day), mall transaction attributes (e.g. average transaction amount per unit), basic attributes (e.g. region, gender, age, etc.), probability of clicking on a pushed activity (e.g. average number of activity information pushed followed by one click), click-to-transaction conversion probability (e.g. number of activity information clicked followed by one transaction). And, the feature image of the target user can be obtained through the user features, and then the user type of the target user can be determined through the feature image.
And S4, acquiring candidate push information corresponding to the target user according to the user characteristics by using an information acquisition module.
Specifically, after the user characteristics are obtained, candidate push information corresponding to the target user can be obtained through user characteristic matching.
The candidate push information is push information obtained through preliminary screening; and in general, the candidate push information needs to be further filtered to obtain at least one push information with the highest probability of being clicked or made into a list.
S5, a model acquisition module is used for acquiring a pre-trained prediction model corresponding to the user type.
That is, each user type has a predictive model corresponding thereto; each prediction model can be classified and stored in a database according to different user types in advance, and a mapping table between the prediction model and the user types is established so as to quickly retrieve the required model from the corresponding storage position.
S6, obtaining a predicted recommendation result corresponding to the candidate push information according to the user characteristics, the push information characteristics and the prediction model by using a result obtaining module; the push information feature is used to characterize the candidate push information.
Specifically, after each newly generated push information to be pushed to the user is obtained, the push information can be normalized and labeled to obtain a normalized push information database. When the push information is active, one of the alternative ways of tagging may be to tag according to the type of activity, the point of interest of the activity, and the goal of the activity, where the type of activity may include: 1. small vault, 2. Collection, 3. Bank, 4. East, 5. Fund, 6. Gold, 7. Comprehensive, 8. Others; the activity benefit points may include: 1. the method comprises the steps of ticket collecting/filling, 2, buying gifts (red package/additional information/Beijing paste), 3, putting aside, 4, paying lottery, 5, ranking, 6, explosion products, 7, function recommendation and the like; the activity targets may include: 1. the method comprises the following steps of first trading of the large financial resource, 2, new drawing of the large financial resource thousand-yuan, 3, new drawing of the large financial resource thousand-yuan, 4, 5, recall of the large financial resource thousand-yuan, 6, first trading of the class, 7, GMV lifting of the thousand-yuan user (Gross Merchandise Volume, total amount of the transaction (within a certain period of time)), 8, GMV lifting of the ten thousand-yuan user, 9, GMV lifting of the high net value user, 10, active lifting of the user and the like. The first transaction may be a first order transaction of a user who has registered with the corresponding platform but has not performed a transaction; the pulling can be that the user is not registered in the corresponding platform and is attracted to register or access; recall may be when a user has not accessed or transacted a platform for a preset period of time (e.g., one year, etc.), attracting him to access or transact again.
Wherein, the prediction model generally comprises all the features, but one piece of push information does not necessarily comprise all the features, so that the push information features can influence the coefficients of different factors in the prediction model; by way of example: when an activity has no benefit point and is only used for carrying out pre-consumption education on a user, and telling the user what to do with the product, the type features and the size features of the related benefit point are missing, and during training, missing values of the related features in the prediction model can be filled according to actual conditions, wherein the missing values can be filled into null values, modes, median and the like according to specific conditions.
Therefore, according to the scheme in the embodiment, different prediction models are matched for different types of users, so that the prediction model with the best prediction effect can be matched for the different types of users, and further the push information which is the most in line with the characteristics of the users can be finally screened out, and the purposes of improving the information click rate, single rate and the like of each user are achieved.
As shown in fig. 2, in some embodiments, as the foregoing data processing method, step S4 uses an information obtaining module to obtain candidate push information corresponding to the target user according to the user characteristics, including steps S41 to S43 as follows:
And S41, using a matching unit, and matching according to the user characteristics to obtain the associated user which meets the preset matching degree with the target user.
Specifically, the user features may include all features related to the target user, and after determining the features included in each user through the history information, a mapping relationship between the user and the features may be established and stored in the user information database.
One of the alternative implementations of this step may be: inquiring in a user information database through various characteristics included in the user characteristics to obtain an associated user meeting a preset matching degree with a target user; the preset matching degree may be: similarity among all the features respectively included by each user; by way of example: when user I includes the features: A. b, C, D, E; when user II includes the features: A. b, C, D, F; the similarity between user I and user II is 80%; when the preset matching degree is 70%, the user II is the associated user of the user I.
S42, acquiring second history information corresponding to the associated user by using the first acquisition unit.
Specifically, the second history information is history information corresponding to the associated user; as history information of each user is recorded under normal conditions; therefore, the corresponding second history information can be obtained by inquiring and then reading the identification information of the associated user from the preset database.
And S43, utilizing a candidate push information determining unit to determine candidate push information interacted with the associated user according to the second historical record information.
Specifically, according to the second history record information, history push information that the associated user clicks or forms a list can be obtained; and taking the historical push information as candidate push information. Further, because the life cycle of the user can be changed, although the current user characteristics meet the requirements, the feature similarity between the associated user and the target user still meets the requirements at the moment of pushing a certain historical pushing information cannot be guaranteed; therefore, the pushing time of the candidate pushing information can be limited, so that the candidate pushing information capable of improving the click rate or the single rate is screened.
In some embodiments, as the foregoing data processing method, step S3 uses the user information determining module to determine the user type of the target user according to the first history information, including steps S31 to S33 as follows:
s31, acquiring the feature type and the activity degree of the target user according to the first history information by using a second acquisition unit; the feature types include: demographic attributes, financial characteristics, click-to-transaction conversion characteristics, historical browsing characteristics, real-time search characteristics, real-time browsing characteristics; the liveness includes: transaction frequency, click frequency.
Specifically, the feature types included in the history information may be selectively recorded according to the needs of the actual application or generated according to the recorded data. By way of example: information such as age, gender, occupation, income and the like in the population attribute can be selected through recorded requirements; the financial characteristics can be generated by processing the recorded information according to the amount of the bill, the financial type (such as insurance, gold, etc.), etc.
Liveness may be derived from the total number of clicks or transactions between segments between units. To characterize the activity of the target user in each time period.
S32, analyzing the feature types and the liveness by using an analysis unit through a preset user type division strategy to obtain a life cycle corresponding to the target user.
S33, determining the user type according to the life cycle corresponding to the target user by using a user type determining unit.
Specifically, the lifecycle is used to characterize the state of a target user in a particular platform. By way of example: for users in lead-in period, the main aim is to promote the first transaction of the users on the platform due to insufficient transaction behavior; for a user in a growth period, the platform has certain use viscosity, and the main targets are to promote the transaction times and activity; users for maturity are then encouraged to grow GMV and stay with AUM (Asset Under Management, asset management scale). Thus, the obtained feature types may include: transaction amount, first transaction information, etc.; liveness may include: transaction frequency.
Alternatively, the user types of the target user may be divided according to the lifecycle of the target user. The partitioning may be followed by the derivation for different user types: and different targets such as user account opening, platform first transaction, transaction amount improvement, user liveness and the like are promoted.
As shown in fig. 3, in some embodiments, as the foregoing data processing method, the candidate push information includes at least two candidate push information, and further includes steps S7 to S8 as follows:
s7, calculating the estimated recommendation result corresponding to each candidate push information by using a score acquisition module according to a preset calculation strategy to obtain a recommendation score corresponding to each candidate push information; the predicted recommended results include: click probability, transaction probability, and predicted transaction amount.
That is, a plurality of models may be included in the preset model for predicting the click probability, the transaction probability, the predicted transaction amount, and the like, respectively. And the number and the types of the models can be selected according to the required results. Alternatively, the importance ranking of a feature can be obtained by a tree splitting algorithm according to the optimal splitting feature, the optimal splitting point of the feature and the splitting times of the feature, and the importance ranking is shown in the following formula:
Wherein,for the traditional linear regression model formula->The relation between any two features can be considered on the basis of a regression model; v i Is with feature x i Corresponding auxiliary vector, v j Is with feature x j A corresponding auxiliary vector; x is x i x j Representing characteristic x i And x j When x is a combination of i And x j All non-zero, combined characteristic x i x j Meaning it is only significant; n represents the feature dimension, and the result y (x) obtained by the method can solve the regression problem and the classification problem, and avoid the influence of data sparseness on model learning.
After each predicted result is obtained, a recommendation score for pushing each candidate push information to the target user is obtained by processing (such as normalization, weighting and the like) each result; in general, the higher the recommendation score, the higher the probability that the candidate push information is clicked or transacted successfully.
S8, pushing N candidate push information with highest recommendation scores to a target user by using a push module; wherein N is an integer not less than 1.
Specifically, the number N may be selected according to the number of information to be pushed. The method comprises the following steps: and screening the highest N candidate push information in all the candidate push information according to the recommendation scores, and pushing the N candidate push information to the target user.
For example, when the push information is activity information, a CTR (Click Through Rate) estimation method may be used to input the user characteristics of the target user into a prediction model for prediction, so as to obtain the user characteristics corresponding to each activity information: the activity exposes to click probabilities, click to transaction conversion probabilities. Multiplying the normalized activity exposure-to-click probability and the click-to-transaction conversion probability corresponding to each activity information to obtain a final recommendation score for each activity information, and further, adjusting coefficients of the activity exposure-to-click probability and the click-to-transaction conversion probability according to specific operation requirements to realize combination of a model and an operation idea, and finally automatically displaying the optimal recommendation combination on a target user side.
Common algorithms for CTR prediction include LR (Logistic Regression, logistic regression analysis), FM (Factorization Machines, factorizer; a matrix-decomposition-based machine learning algorithm proposed by Steffen render, which is most characterized by good learning ability for sparse data, GBDT (Gradient Boosting Decision Tree gradient-lifted iterative decision tree), and the like.
Optionally, in the optional data processing method, step S8 uses a pushing module to push N candidate pushing information with highest recommendation scores to the target user, including:
uploading target pushing information to an activity pool corresponding to a target user by using a resource bit display unit; so that the target user can acquire target push information in the activity pool through the user side; the N candidate push information with the highest recommendation score comprises target push information;
and respectively determining keyword information in each target push information by using a transmitting unit, writing the keyword information into a corresponding position in a preset template, generating push information corresponding to each target push information, and transmitting the push information to a target user. Specifically, when the push information is the activity information, the purpose of pushing the activity information to the target user may be accomplished as follows:
1. resource bit recommendation display: and uploading target pushing information to an activity pool by related personnel through a resource bit display unit, outputting optimal activities based on a single user dimension at two layouts of a main target and an auxiliary target through a CTR model and other conversion models, and displaying at a user side.
2. Offline Push/short message recommendation display: a sending unit is used for obtaining keyword information in each target push information according to a preset keyword extraction strategy, determining the corresponding relation between each keyword information and the position in the preset template, and finally writing the keyword information into the preset template; and the optimal activities can be output through a unified recommendation model and linked with an interface template Push or template short message, so that thousands of people and thousands of faces of documents and connection display are performed for activity recommendation. The text and the links in the short message are unified standardized configuration when the operation uploading activities are carried out, the conversion model is called at the bottom layer, the text and the links matched with the activities are called in real time according to the activities with the highest scores corresponding to different users, the short message or push is reached, and the probability of successful conversion of the activities is further improved.
In some embodiments, as in the foregoing data processing method, the step S6 uses a result obtaining module to obtain, according to the user characteristics, the push information characteristics and the prediction model, a predicted recommended result corresponding to the candidate push information, and includes the following steps S61 and S62:
and S61, configuring the prediction model according to the push information characteristics by using a target prediction model determining unit to obtain a target prediction model corresponding to the candidate push information.
And S62, inputting the user characteristics into a target prediction model by using a result determining unit to predict, and obtaining a predicted recommendation result.
Specifically, all feature items are preset in a prediction model in general, and it can be determined that features a included in the push information features configure each feature in the prediction model to obtain a target prediction model corresponding to the candidate push information. And then the predicted recommendation result of the pushing information for the target user can be predicted according to the configured target prediction model.
For example, the missing values of the relevant features in the prediction model can be filled according to actual conditions, wherein the missing values can be filled into null values, modes, median and the like according to specific conditions; after filling, a target prediction model corresponding to the candidate push information can be obtained; furthermore, before prediction, integral feature processing, such as data normalization, data cleaning, and model prediction after filling in missing values, can be performed.
After the target prediction model is obtained, the user features can be input into the target prediction model to obtain a final predicted recommendation result.
In some embodiments, the data processing method as described above further includes the following steps P1 to P3:
Step P1, a third acquisition unit is used for acquiring third historical record information corresponding to a candidate user according to a preset training data acquisition strategy; the third history information is not used to train the predictive model, the candidate user includes the target user, and the candidate user is consistent with the user type of the target user.
Specifically, the training data acquisition strategy is used for acquiring history information which is not used for training the model; and is noted as third history information for distinguishing from the history information in the foregoing embodiment.
The training data acquisition strategy may be: determining the last training time of the prediction model, and acquiring third historical record information acquired after the training time; or, each training is performed at intervals of a preset time period, and then third history information in a latest preset time period in time is acquired; other acquisition methods are also possible, and are not listed here.
And step P2, randomly dividing the third historical record information by using a dividing unit to obtain training set supplementary data and verification set supplementary data.
And step P3, training the prediction model through the training set supplementary data by using a first updating unit, and obtaining a first updated prediction model when the trained prediction model is verified through the verification set supplementary data and the preset performance requirement is met.
That is, the third history information is divided into training set supplemental data and verification set supplemental data.
After training the prediction model through the training set supplementary data, verifying the trained prediction model through the verification set supplementary data, and obtaining a first updated prediction model when the performance requirement is met; the first updated prediction model is a prediction model trained by real-time latest training data.
Therefore, the model can be automatically optimized by adopting the method in the embodiment, and the accuracy of prediction can be continuously improved while prediction is performed.
In some embodiments, the data processing method as described above further includes steps Q1 to Q4 as follows:
step Q1. uses the fourth obtaining unit to obtain the newly uploaded information to be processed.
In particular, the information to be processed is newly uploaded and can be used for information pushed to the user. Further, the information to be processed may be stored in the corresponding database according to the time sequence after being uploaded, so the manner of obtaining the information to be processed may be: and reading the information to be processed from the database according to a certain period (day/week/month).
And step Q2, determining the characteristics to be processed in the information to be processed by using a characteristics to be processed determining unit.
Specifically, the feature to be processed is a feature corresponding to the information to be processed; one alternative implementation may be: when the information to be processed is generated, the contents and the fields in the information are normalized, so that the purpose of obtaining the corresponding characteristics of the contents by analyzing the contents and the fields is achieved.
Step Q3. uses a new feature determination unit to determine new features of the features to be processed that are not present in the predictive model.
Specifically, all the features in the prediction model can be obtained, and then the features to be processed are traversed one by one among the features included in the model to obtain new features.
And step Q4, using a second updating unit to perform feature addition on the prediction model according to the new features to obtain a second updated prediction model.
Specifically, after the new feature is obtained, the prediction model needs to be correspondingly modified according to the new feature, so that the prediction model can be used for predicting push information including the new feature. One method of feature addition may be: a feature factor corresponding to the new feature is obtained and then added to the predictive model.
Further, after the second updated prediction model is obtained, the result of the prediction of the model may be monitored for a certain period (day/week/month) to timely feed back and adjust the recommended result.
By the method in the embodiment, the prediction model can be optimized and adjusted regularly according to the updating of the push information, so that the problem that good prediction performance cannot be achieved for some push information comprising new features is avoided, and the prediction model can guarantee that accurate prediction results are given out from time to time.
As shown in fig. 4, according to another aspect of the present application, an embodiment of the present application further provides a data processing apparatus for information push, including:
the determining module 1 is used for determining a target user needing information pushing;
the record acquisition module 2 is used for acquiring first history record information corresponding to the target user;
a user information determining module 3, configured to determine a user type and a user characteristic of the target user according to the first history information;
the information acquisition module 4 is used for acquiring candidate push information corresponding to the target user according to the user characteristics;
the model acquisition module 5 is used for acquiring a pre-trained prediction model corresponding to the user type;
the result acquisition module 6 is used for acquiring a predicted recommendation result corresponding to the candidate push information according to the user characteristics, the push information characteristics and the prediction model; the push information feature is used to characterize the candidate push information.
In particular, the specific process of implementing the functions of each module in the apparatus of the embodiment of the present application may be referred to the related description in the method embodiment, which is not repeated herein.
According to another embodiment of the present application, there is also provided an electronic apparatus including: as shown in fig. 5, the electronic device may include: the device comprises a processor 1501, a communication interface 1502, a memory 1503 and a communication bus 1504, wherein the processor 1501, the communication interface 1502 and the memory 1503 are in communication with each other through the communication bus 1504.
A memory 1503 for storing a computer program;
the processor 1501 is configured to execute the program stored in the memory 1503, thereby implementing the steps of the method embodiment described above.
The buses mentioned for the above electronic devices may be peripheral component interconnect standard (Peripheral Component Interconnect, PCI) buses or extended industry standard architecture (Extended Industry Standard Architecture, EISA) buses, etc. The bus may be classified as an address bus, a data bus, a control bus, etc. For ease of illustration, the figures are shown with only one bold line, but not with only one bus or one type of bus.
The communication interface is used for communication between the electronic device and other devices.
The Memory may include random access Memory (Random Access Memory, RAM) or may include Non-Volatile Memory (NVM), such as at least one disk Memory. Optionally, the memory may also be at least one memory device located remotely from the aforementioned processor.
The processor may be a general-purpose processor, including a central processing unit (Central Processing Unit, CPU), a network processor (Network Processor, NP), etc.; but also Digital signal processors (Digital SignalProcessing, DSP), application specific integrated circuits (APPlication Specific Integrated Circuit, ASIC), field-programmable gate arrays (Field-Programmable Gate Array, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.
The embodiment of the application also provides a non-transitory computer readable storage medium, which stores computer instructions that cause a computer to execute the steps of the method embodiment.
It should be noted that in this document, relational terms such as "first" and "second" and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
The foregoing is only a specific embodiment of the invention to enable those skilled in the art to understand or practice the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. The data processing method for information push is characterized by comprising the following steps:
determining a target user needing information pushing by using a determining module;
the application record acquisition module acquires first history record information corresponding to the target user;
a user information determining module is used for determining the user type and the user characteristics of the target user according to the first historical record information;
an application information acquisition module acquires candidate push information corresponding to the target user according to the user characteristics;
a model acquisition module is used for acquiring a pre-trained prediction model corresponding to the user type;
The application result acquisition module is used for acquiring an expected recommendation result corresponding to the candidate push information according to the user characteristics, the push information characteristics and the prediction model; the push information features are used for characterizing the candidate push information.
2. The data processing method according to claim 1, wherein the application information obtaining module obtains candidate push information corresponding to the target user according to the user characteristics, including:
a matching unit is used for matching according to the user characteristics to obtain an associated user which meets the preset matching degree with the target user;
acquiring second history information corresponding to the associated user by using a first acquisition unit;
and determining the candidate push information interacted with the associated user according to the second historical record information by using a candidate push information determining unit.
3. The data processing method according to claim 1, wherein the employing user information determining module determines the user type of the target user based on the first history information, comprising:
a second obtaining unit is used for obtaining the feature type and the activity degree of the target user according to the first history information; the feature types include: demographic attributes, financial characteristics, click-to-transaction conversion characteristics, historical browsing characteristics, real-time search characteristics, real-time browsing characteristics; the liveness includes: transaction frequency, click frequency;
Analyzing the feature types and the liveness by using an analysis unit through a preset user type division strategy to obtain a life cycle corresponding to the target user;
and determining the user type according to the life cycle corresponding to the target user by using a user type determining unit.
4. The data processing method according to claim 1, wherein the candidate push information includes at least two, the method further comprising:
calculating the estimated recommendation result corresponding to each candidate push information according to a preset calculation strategy by using a score acquisition module to obtain a recommendation score corresponding to each candidate push information; the predicted recommended results include: click probability, transaction probability, and predicted transaction amount;
pushing the N candidate push information with the highest recommendation scores to the target user by using a push module; wherein N is an integer not less than 1.
5. The data processing method according to claim 1, wherein the application result obtaining module obtains an estimated recommendation result with the candidate push information according to the user feature, the push information feature, and the prediction model, and includes:
A target prediction model determining unit is used for configuring the prediction model according to the push information characteristics to obtain a target prediction model corresponding to the candidate push information;
and the application result determining unit inputs the user characteristics into the target prediction model to predict, so as to obtain the predicted recommendation result.
6. The data processing method according to claim 1, characterized by further comprising:
a third acquisition unit is used for acquiring third historical record information corresponding to the candidate user according to a preset training data acquisition strategy; the third history information is not used for training the prediction model, the candidate user comprises the target user, and the candidate user is consistent with the user type of the target user;
randomly dividing the third historical record information by using a dividing unit to obtain training set supplementary data and verification set supplementary data;
and training the prediction model through the training set supplementary data by using a first updating unit, and obtaining a first updated prediction model when the trained prediction model is verified through the verification set supplementary data and the preset performance requirement is met.
7. The data processing method according to claim 1, characterized by further comprising:
acquiring newly uploaded information to be processed by using a fourth acquisition unit;
determining the characteristics to be processed in the information to be processed by using a characteristics to be processed determining unit;
determining new features which are not in the prediction model in the features to be processed by using a new feature determination unit;
and applying a second updating unit to perform feature addition on the prediction model according to the new features to obtain a second updated prediction model.
8. A data processing apparatus for information push, comprising:
the determining module is used for determining a target user needing information pushing;
the record acquisition module is used for acquiring first history record information corresponding to the target user;
the user information determining module is used for determining the user type and the user characteristics of the target user according to the first historical record information;
the information acquisition module is used for acquiring candidate push information corresponding to the target user according to the user characteristics;
the model acquisition module is used for acquiring a pre-trained prediction model corresponding to the user type;
The result acquisition module is used for acquiring a predicted recommendation result corresponding to the candidate push information according to the user characteristics, the push information characteristics and the prediction model; the push information features are used for characterizing the candidate push information.
9. An electronic device, comprising: the device comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory are communicated with each other through the communication bus;
the memory is used for storing a computer program;
the processor being adapted to implement the processing method of any of claims 1-7 when executing the computer program.
10. A non-transitory computer readable storage medium storing computer instructions that cause the computer to perform the processing method of any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010921701.4A CN112148973B (en) | 2020-09-04 | 2020-09-04 | Data processing method and device for information push |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010921701.4A CN112148973B (en) | 2020-09-04 | 2020-09-04 | Data processing method and device for information push |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112148973A CN112148973A (en) | 2020-12-29 |
| CN112148973B true CN112148973B (en) | 2023-11-07 |
Family
ID=73889780
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010921701.4A Active CN112148973B (en) | 2020-09-04 | 2020-09-04 | Data processing method and device for information push |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112148973B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113763112A (en) * | 2021-02-25 | 2021-12-07 | 北京沃东天骏信息技术有限公司 | Information pushing method and device |
| CN115225702B (en) * | 2021-04-19 | 2023-08-08 | 北京达佳互联信息技术有限公司 | Information pushing method and device, electronic equipment and storage medium |
| CN114285896B (en) * | 2021-12-28 | 2023-07-07 | 建信金融科技有限责任公司 | Information pushing method, device, equipment, storage medium and program product |
| CN115082132B (en) * | 2022-08-18 | 2022-11-04 | 网娱互动科技(北京)股份有限公司 | Promotion method and system based on point reward |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109460513A (en) * | 2018-10-31 | 2019-03-12 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating clicking rate prediction model |
| CN109460512A (en) * | 2018-10-25 | 2019-03-12 | 腾讯科技（北京）有限公司 | Recommendation information processing method, device, equipment and storage medium |
| CN110263242A (en) * | 2019-01-04 | 2019-09-20 | 腾讯科技（深圳）有限公司 | Content recommendation method, device, computer readable storage medium and computer equipment |
| CN110276446A (en) * | 2019-06-26 | 2019-09-24 | 北京百度网讯科技有限公司 | The method and apparatus of model training and selection recommendation information |
| CN110399550A (en) * | 2019-03-14 | 2019-11-01 | 腾讯科技（深圳）有限公司 | A kind of information recommendation method and device |
| CN110400169A (en) * | 2019-07-02 | 2019-11-01 | 阿里巴巴集团控股有限公司 | A kind of information-pushing method, device and equipment |
| CN110413877A (en) * | 2019-07-02 | 2019-11-05 | 阿里巴巴集团控股有限公司 | A kind of resource recommendation method, device and electronic equipment |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9465876B2 (en) * | 2013-09-09 | 2016-10-11 | International Business Machines Corporation | Managing content available for content prediction |
| US10713716B2 (en) * | 2017-08-04 | 2020-07-14 | Airbnb, Inc. | Verification model using neural networks |
| CN108921221B (en) * | 2018-07-04 | 2022-11-18 | 腾讯科技（深圳）有限公司 | User feature generation method, device, equipment and storage medium |
-
2020
- 2020-09-04 CN CN202010921701.4A patent/CN112148973B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109460512A (en) * | 2018-10-25 | 2019-03-12 | 腾讯科技（北京）有限公司 | Recommendation information processing method, device, equipment and storage medium |
| CN109460513A (en) * | 2018-10-31 | 2019-03-12 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating clicking rate prediction model |
| CN110263242A (en) * | 2019-01-04 | 2019-09-20 | 腾讯科技（深圳）有限公司 | Content recommendation method, device, computer readable storage medium and computer equipment |
| CN110399550A (en) * | 2019-03-14 | 2019-11-01 | 腾讯科技（深圳）有限公司 | A kind of information recommendation method and device |
| CN110276446A (en) * | 2019-06-26 | 2019-09-24 | 北京百度网讯科技有限公司 | The method and apparatus of model training and selection recommendation information |
| CN110400169A (en) * | 2019-07-02 | 2019-11-01 | 阿里巴巴集团控股有限公司 | A kind of information-pushing method, device and equipment |
| CN110413877A (en) * | 2019-07-02 | 2019-11-05 | 阿里巴巴集团控股有限公司 | A kind of resource recommendation method, device and electronic equipment |
Non-Patent Citations (1)
| Title |
|---|
| 基于序列特征的点击率预测模型;朱思涵;浦剑;;华东师范大学学报(自然科学版)(第04期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112148973A (en) | 2020-12-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112148973B (en) | Data processing method and device for information push | |
| US10991048B1 (en) | Curated sentiment analysis in multi-layer, machine learning-based forecasting model using customized, commodity-specific neural networks | |
| CN105989004B (en) | Information delivery preprocessing method and device | |
| CN109783730A (en) | Products Show method, apparatus, computer equipment and storage medium | |
| WO2017190610A1 (en) | Target user orientation method and device, and computer storage medium | |
| US20190213194A1 (en) | System and method for information recommendation | |
| WO2019072128A1 (en) | Object identification method and system therefor | |
| CN111095330B (en) | Machine learning method and system for predicting online user interactions | |
| JP2019215778A (en) | Service provision system, advertisement related service provision system, user side facility and user side advertisement facility | |
| CN111429214B (en) | Transaction data-based buyer and seller matching method and device | |
| CN111582932A (en) | Inter-scene information pushing method and device, computer equipment and storage medium | |
| CN111429161B (en) | Feature extraction method, feature extraction device, storage medium and electronic equipment | |
| CN111680213B (en) | Information recommendation method, data processing method and device | |
| CN113424207B (en) | System and method for efficiently training understandable models | |
| CN110209944B (en) | Stock analyst recommendation method and device, computer equipment and storage medium | |
| Boz et al. | Reassessment and monitoring of loan applications with machine learning | |
| CN112084376A (en) | Map knowledge based recommendation method and system and electronic device | |
| CN113822390B (en) | User portrait construction method and device, electronic equipment and storage medium | |
| Ram et al. | Fake reviews detection using supervised machine learning | |
| WO2018002664A1 (en) | Data aggregation and performance assessment | |
| CN110162714A (en) | Content delivery method, calculates equipment and computer readable storage medium at device | |
| Gulhane et al. | TensorFlow Based Website Click through Rate (CTR) Prediction Using Heat maps | |
| Sharma | Identifying Factors Contributing to Lead Conversion Using Machine Learning to Gain Business Insights | |
| US11887168B2 (en) | Predicting the value of an asset using machine-learning techniques | |
| US20230306279A1 (en) | Guided feedback loop for automated information categorization |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Address after: Room 221, 2 / F, block C, 18 Kechuang 11th Street, Daxing District, Beijing, 100176 Applicant after: Jingdong Technology Holding Co.,Ltd. Address before: Room 221, 2 / F, block C, 18 Kechuang 11th Street, Daxing District, Beijing, 100176 Applicant before: Jingdong Digital Technology Holding Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |