CN111340522B - Resource recommendation method, device, server and storage medium - Google Patents
Resource recommendation method, device, server and storage medium Download PDFInfo
- Publication number
- CN111340522B CN111340522B CN201911398736.8A CN201911398736A CN111340522B CN 111340522 B CN111340522 B CN 111340522B CN 201911398736 A CN201911398736 A CN 201911398736A CN 111340522 B CN111340522 B CN 111340522B
- Authority
- CN
- China
- Prior art keywords
- candidate resource
- resource object
- features
- feature
- intersection
- 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
- 238000000034 method Methods 0.000 title claims abstract description 45
- 238000013528 artificial neural network Methods 0.000 claims description 17
- 238000004590 computer program Methods 0.000 claims description 11
- 230000010354 integration Effects 0.000 claims description 5
- 238000000605 extraction Methods 0.000 claims description 3
- 239000010410 layer Substances 0.000 description 16
- 239000013598 vector Substances 0.000 description 14
- 238000010586 diagram Methods 0.000 description 13
- 230000008569 process Effects 0.000 description 10
- 230000006870 function Effects 0.000 description 9
- 238000012545 processing Methods 0.000 description 8
- 238000012986 modification Methods 0.000 description 5
- 230000004048 modification Effects 0.000 description 5
- 230000006399 behavior Effects 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 3
- 235000013305 food Nutrition 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 241000531116 Blitum bonus-henricus Species 0.000 description 1
- 235000008645 Chenopodium bonus henricus Nutrition 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000003542 behavioural effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000004566 building material Substances 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000008267 milk Substances 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000007115 recruitment Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 239000002356 single layer Substances 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
- G06Q30/0269—Targeted advertisements based on user profile or attribute
- G06Q30/0271—Personalized advertisement
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0207—Discounts or incentives, e.g. coupons or rebates
- G06Q30/0224—Discounts or incentives, e.g. coupons or rebates based on user history
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0207—Discounts or incentives, e.g. coupons or rebates
- G06Q30/0239—Online discounts or incentives
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
- G06Q30/0255—Targeted advertisements based on user history
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
- G06Q30/0259—Targeted advertisements based on store location
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
- G06Q30/0261—Targeted advertisements based on user location
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Strategic Management (AREA)
- Finance (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Economics (AREA)
- Marketing (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the specification provides a resource recommendation method, a device, a server and a storage medium: and carrying out feature intersection on the resource object features of the candidate resource objects and the user features through a combined prediction model to obtain combined features containing multi-order intersection features, carrying out user click rate prediction on the candidate resource objects according to the combined features to obtain CTR prediction results of the candidate resource objects, determining target resource objects from M candidate resource objects according to the CTR prediction results of each candidate resource object in the M candidate resource objects, and recommending the determined target resource objects to a target user.
Description
Technical Field
The embodiment of the specification relates to the technical field of Internet, in particular to a resource recommendation method, a resource recommendation device, a server and a storage medium.
Background
In the computer field, there are a variety of electronic resources that can be used by corresponding objects to implement the functions of the electronic resources. The interests of different users are different and thus the preferences for the electronic resources are different. Such as: shared memory space, electronic data (e.g., video, audio, documents), electronic coupons, and the like.
For example, taking coupons as an example, online commerce such as e-commerce and O2O is vigorous, and various platforms build rich basic capabilities to support better operation of merchants. The e-commerce platform generally stimulates consumer consumption through coupons. And the electronic commerce platform issues the coupon activity to the online merchant for the recruitment. On-line merchants apply for coupons by doing online campaigns. Different users have different interests and preferences, so that different coupons need to be individually recommended to different users to recommend the most suitable coupons to the users.
Disclosure of Invention
The embodiment of the specification provides a resource recommendation method, a device, a server and a storage medium, so that the accuracy and the rationality of resource recommendation are improved, repeated resource recommendation to a user caused by inaccurate resource recommendation is avoided, and network transmission resources are saved.
In a first aspect, an embodiment of the present disclosure provides a resource recommendation method, including: acquiring user characteristics of a target user and resource object characteristics of M candidate resource objects, wherein M is a positive integer; CTR estimation is respectively carried out on the M candidate resource objects, so that a CTR estimation result of each candidate resource object in the M candidate resource objects is obtained; wherein the performing CTR estimation on the M candidate resource objects includes: for each candidate resource object, performing feature intersection on the resource object features of the candidate resource object and the user features through a combined prediction model to obtain combined features of the candidate resource object, and performing CTR prediction on the candidate resource object according to the combined features to obtain a CTR prediction result of the candidate resource object, wherein the combined features comprise linear features, intra-group intersection features and inter-group intersection features which are obtained according to a multi-layer factorizer sub-model of the combined prediction model, and higher-order intersection features which are obtained according to a neural network sub-model of the combined prediction model; and determining a target resource object from the M candidate resource objects according to the CTR estimated result of each candidate resource object in the M candidate resource objects, and recommending the target resource object to the target user.
In a second aspect, embodiments of the present disclosure provide a resource recommendation device, including: the feature extraction unit is used for acquiring user features of a target user and resource object features of M candidate resource objects, wherein M is a positive integer; a CTR estimation unit, configured to perform CTR estimation on the M candidate resource objects, to obtain a CTR estimation result of each candidate resource object in the M candidate resource objects; wherein the performing CTR estimation on the M candidate resource objects includes: for each candidate resource object, performing feature intersection on the resource object features of the candidate resource object and the user features through a combined prediction model to obtain combined features of the candidate resource object, and performing CTR prediction on the candidate resource object according to the combined features to obtain a CTR prediction result of the candidate resource object, wherein the combined features comprise linear features, intra-group intersection features and inter-group intersection features which are obtained according to a multi-layer factorizer sub-model of the combined prediction model, and higher-order intersection features which are obtained according to a neural network sub-model of the combined prediction model; and the resource recommending unit is used for determining a target resource object from the M candidate resource objects according to the CTR estimated result of each candidate resource object in the M candidate resource objects and recommending the target resource object to the target user.
In a third aspect, embodiments of the present disclosure provide a server, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor implements the steps of the resource recommendation method described in any of the above when the program is executed.
In a fourth aspect, embodiments of the present disclosure provide a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the resource recommendation method of any of the above.
The technical scheme provided by the embodiment of the specification at least realizes the following beneficial effects or advantages:
performing feature intersection on the resource object features of the candidate resource objects and the user features through a combined pre-estimation model, wherein the obtained combined features comprise linear features, intra-group intersection features and inter-group intersection features which are obtained according to a multi-layer factorizer sub-model of the combined pre-estimation model, and high-order intersection features which are obtained according to a neural network sub-model of the combined pre-estimation model; CTR prediction is carried out according to the combined characteristics of the candidate resource objects, CTR prediction results of the candidate resource objects are obtained, and target resource objects are recommended to target users from M candidate resource objects according to the CTR prediction results of each candidate resource object in the M candidate resource objects. Through the technical scheme, CTR prediction of the candidate resource object is simultaneously influenced by the first-order linear characteristic, the second-order intra-group cross characteristic, the third-order inter-group cross characteristic and the fourth-order inter-group cross characteristic and the higher-order combination characteristic. The method has the advantages that the characteristic information for predicting the click rate of the candidate resource object is enriched, more-order cross characteristics can reflect the characteristics of the candidate resource object, so that the click rate of the candidate resource object can be predicted more accurately, the accuracy and rationality of recommending resources to a user can be improved, the resources can be recommended to the user accurately at one time, the situation that the resources are recommended to the same user for multiple times is avoided, and network transmission resources are saved.
Drawings
FIG. 1 is a diagram of a system architecture to which a resource recommendation method according to an embodiment of the present disclosure is applied;
FIG. 2 is a flowchart of a resource recommendation method according to a first aspect of the embodiments of the present disclosure;
FIG. 3 is a schematic diagram of an FM sub-model used in an embodiment of a resource recommendation method according to the present disclosure;
FIG. 4 is a schematic structural diagram of a resource recommendation device according to a second aspect of the embodiments of the present disclosure;
fig. 5 is a schematic structural diagram of a resource recommendation server according to a third aspect of the embodiments of the present disclosure.
Detailed Description
In order to better understand the technical solutions described above, the technical solutions of the embodiments of the present specification are described in detail below through the accompanying drawings and the specific embodiments, and it should be understood that the specific features of the embodiments of the present specification and the specific features of the embodiments of the present specification are detailed descriptions of the technical solutions of the embodiments of the present specification, and not limit the technical solutions of the present specification, and the technical features of the embodiments of the present specification may be combined without conflict.
Referring to fig. 1, fig. 1 is a system architecture diagram to which a resource recommendation method in an embodiment of the present disclosure is applied, where the resource recommendation method includes:
merchant device 10 the merchant device 10 comprises a plurality (2 merchant devices 10 are illustrated in fig. 1), each merchant device 10 providing a candidate resource object. User equipment 20, the user equipment 20 includes a plurality of (schematic diagram 4 user equipment in fig. 1), each user equipment 20 obtains candidate resource objects and uses the candidate resource objects. The server 30 may be a server of an e-commerce platform. The server 30 obtains candidate resource objects for the target user, performs feature intersection according to the user features of the target user and the resource object features of the candidate resource objects to obtain combined features, performs CTR evaluation on the candidate resource objects according to the combined features, and determines whether to recommend the candidate resource objects to the target user according to the candidate resource object CTR evaluation result.
The merchant device 10 and the user device 20 in the embodiments of the present disclosure may be specific types of entity devices such as smart phones, desktop computers, tablet computers, notebook computers, digital assistants, intelligent wearable devices, and the operating systems that run may include, but are not limited to, android systems, IOS systems, linux, windows, and the like. The server 30 may comprise a single independently operating server, or a distributed server, or a server cluster consisting of a plurality of servers.
In this specification, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or server that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed or inherent to such process, method, article, or apparatus, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
In a first aspect, an embodiment of the present disclosure provides a resource recommendation method applied to a server 30 of a system architecture diagram shown in fig. 1. Next, a resource recommendation method provided in the embodiment of the present disclosure will be described with reference to fig. 2.
As shown in fig. 2, the resource recommendation method provided in the embodiment of the present disclosure includes the following steps:
s201, obtaining user characteristics of a target user and resource object characteristics of M candidate resource objects, wherein M is a positive integer.
In the present embodiment, the M candidate resource objects are resource objects recalled for the target user. Specifically, it may be based on an LBS (Location Based Service, location-based service) recall policy to recall one or more candidate resource objects for provision to a target user. The number M of recall candidate resource objects is determined according to the number of resource objects around the current real-time position of the target user and a preset distance value in the LBS recall strategy. For example, it may be a candidate resource object recalled to within 10km of the current real-time location of the target user.
Specifically, M candidate resource objects are obtained by: and recalling the candidate resource objects provided by the merchant equipment within the preset distance range of the target user according to the LBS recall strategy to obtain M candidate resource objects for providing to the target user.
In the embodiment of the present specification, different application scenarios correspond to different candidate resource objects. As for the application scenario being a merchant recommendation scenario for an APP of an offline merchant recommendation type, the candidate resource object may specifically be: the offline merchant provides the user with a rights resource object, such as a coupon, redemption ticket, discount coupon, etc., provided by the offline merchant to the user. Coupons provided by merchant devices within a preset distance range of the current real-time location of the target user are recalled. For example, recall coupons offered by merchant devices within 1000m of the current real-time location of the target user.
Of course, the candidate resource object may also be an introduction information file provided by an offline merchant, and then recall the introduction information file sent by one or more merchant devices within a preset distance range of the current real-time location of the target user. As far as the application scenario is a function point recommendation on a map of the map APP, the candidate resource object is a certain or more types of function points within a preset distance range of the target user. Such as gas stations, high-speed service stations, supermarkets, hotels, etc. within a preset distance of the target user. For the short video recommendation scenario, if the candidate resource object is a short video, the short videos issued by other users within a preset distance range of the current real-time position of the target user are recalled.
In a specific implementation, the multi-dimensional user characteristics of the target user include: age characteristics, gender characteristics, user preference characteristics, etc. of the target user. Specifically, the user preference characteristics of the target user are obtained by processing historical behavior record data of the APP used by the target user. Wherein, the historical behavior record data of the target user comprises: the recorded historical behaviors such as resource objects and the like are received and approved by the target user in the historical time period, and the historical behaviors of coupons are received and approved by the target user in the historical time period. For example, the historical behavioral record data includes: the recorded target user browses the video type, browsing duration, etc. of the short video in the historical time period.
And age characteristics, gender, etc. are derived from user information stored on the server. The application scenarios are different, and the types of the corresponding user preference features are completely different or partially different. As far as the applied scenario is APP for an offline merchant recommendation type, the user preference information may be a consumption amount level (which of a plurality of preset consumption amount segments belongs to), a consumption type preference (one or more of consumption types of food, movie shows, medical health, leisure entertainment, shopping, sports fitness, life travel, etc.). The application scene is specific to map APP, and the user preference information is specifically life track preference (the life track preference is one or more of food, movie performance, medical health, leisure and entertainment, shopping, sports and fitness, life trip and other life track preference).
Under different application scenarios, the obtained resource object features of the candidate resource object may be completely different or partially different, and taking the case that the candidate resource object is a coupon, the resource object features include: industry attributes of the coupons, such as clothing, food, building materials, or digital products; the coupon's coupon amount, coupon value such as coupon discount, coupon threshold, coupon utilization, etc. It should be noted that, the discount amount and the discount represent the amount or the proportion of the discount that can be really reduced, and the discount threshold represents the difficulty of using the coupon, for example: the method can be determined by the threshold condition of the full coupon and the original price of the commodity of the redeemed coupon. For example, if a full coupon is 100-yuan less 10-yuan, the preference threshold is 100-yuan. For example, a certain redemption ticket is redeemed for goods below 100 yuan, and the preference threshold is 100 yuan. The coupon availability characterizes the applicable scope of the coupon, such as: the full-field coupon is applicable to all goods, but the egg-milk redemption coupon is applicable to only specified goods.
The above-described resource object features when candidate resource objects are coupons are for illustration only. When the candidate resource object is of other types, which resource object features include may be determined according to actual resource attributes, which are not described herein for brevity of description.
S202, CTR estimation is conducted on the M candidate resource objects respectively, and a CTR estimation result of each candidate resource object in the M candidate resource objects is obtained.
Specifically, the performing CTR estimation on the M candidate resource objects includes: for each candidate resource object, CTR prediction is performed through the following steps S2021 to S2022, so as to obtain M CTR prediction results corresponding to M candidate resource objects:
s2021, performing feature intersection on the resource object features of the candidate resource objects and the user features through a combined pre-estimation model to obtain combined features of the candidate resource objects, wherein the combined features comprise linear features, intra-group intersection features and inter-group intersection features which are obtained according to a multi-layer factorizer sub-model of the combined pre-estimation model, and high-order intersection features which are obtained according to a neural network sub-model of the combined pre-estimation model.
In the embodiment of the present specification, the improved combination model obtained by hierarchically improving the original combination model includes: a multi-layered factorization machine sub-model and a neural network sub-model. The raw combined model is a factorized machine sub-model that contains neural network sub-models and supports single layer feature crossing. Based on the training sample data, the improved combination model obtained through the layering improvement is jointly trained, so that each weight matrix of the improved combination model is adjusted, and a combination estimated model is obtained. The multi-layered factorer sub-model includes at least two layers of feature crossings. In a specific implementation process, the combined pre-estimation model may be: the combined estimated model is obtained by training a hierarchical modified deep FM (Deep Factorization Machines, depth factor decomposition) model or a modified combined model obtained by hierarchical modified Xdeep FM. The pre-estimated model obtained through training combines the advantages of the breadth model and the depth model.
For the combined pre-estimated model obtained by hierarchically improving deep FM, the combined pre-estimated model is a parallel structure formed by an FM (Factorization Machines, factorizer) sub-model and a DNN (Deep Neural Networks, deep neural network) sub-model, and the FM sub-model and the DNN sub-model share the same input: the grouping and embedding process is carried out on the user characteristics and the resource object characteristics of the M candidate resource objects by using the ebadd characteristic vector. The multi-layer FM sub-model forms first-order linear features and second, third, and fourth-order cross features by feature-interleaving (multiplying or cartesian-product) features in the N original feature groups. The DNN submodel forms higher order cross features (fifth, sixth order cross features, etc.) by feature-interleaving the N original feature groupings.
The process of constructing intra-group cross features and inter-group features based on the multi-layer factorizer submodel in the embodiments of the present specification is described in more detail below with reference to FIG. 3 (one circle in the figure represents a one-dimensional feature): specifically, the method comprises the following steps 1 to 4:
step 1, grouping and embedding the user characteristics and the resource object characteristics of the candidate resource object to obtain N original characteristic groups of the candidate resource object, wherein N is a positive integer.
Specifically, the obtained multi-dimensional user characteristics of the candidate resource object and the characteristics of the multi-dimensional resource object are divided into the same original characteristic groups, so that N original characteristic groups are obtained, and the number of the original characteristic groups is determined by the number of the characteristic types. The following m-dimensional raw features field_0, field_1, …, field_m are grouped according to the feature types of the m-dimensional raw features (i.e., various user features and various resource object features). The following N feature packets are formed: group_0, group_1, …, group_n, and performing ebeddings (i.e., embedding processing) on each dimension of the original feature in each feature group, to obtain N original feature groups correspondingly. And carrying out dimension reduction on m-dimensional original features which are sparse features through processing of an emmbeddings algorithm, specifically reducing the dimension through weight matrix calculation of an emmbedding layer, and obtaining emmbedding feature vectors in each of N original feature groups.
And step 2, performing intra-group feature intersection on the N original feature groups through the multi-layer factorizer sub-model to obtain N groups of intra-group intersection results of the candidate resource objects.
Specifically, intra-group feature intersection is performed on the ebedding feature vectors in the N original feature groups, so as to obtain N groups of intra-group intersection results, namely: crossing the ebedding feature vector in group_0 to obtain a first group of intra-group crossing results; crossing the ebedding feature vector in gropu_1 to obtain a second group of crossing results; …, interleaving the interleaving feature vectors in group_m to obtain a third intra-group interleaving result. It should be noted that, feature intersection is performed on the ebadd feature vector in each of the N original feature packets, so as to obtain second-order intersection features in the N sets of intra-set intersection results.
And 3, carrying out feature addition, inter-group feature intersection and integration of original features in the N original feature groups on N groups of intra-group intersection results and N original feature groups of the candidate resource objects through a multi-layer factorization machine sub-model to generate linear features, intra-group intersection features and inter-group intersection features of the candidate resource objects.
Specifically, the above step 3 is described in more detail with reference to fig. 3:
and 3A, embedding the N groups of intra-group crossing results of the candidate resource object, and embedding the original features of the N original feature groups to form N intermediate feature groups of the candidate resource object.
Specifically, the intermediate embedding layer implements the embedding of the intra-group cross results of each group, and the obtained embeddings feature vector after the intra-group cross results are subjected to the embedding and the embedding feature vector in the original feature packet for generating the intra-group cross results form an intermediate feature packet, so that N intermediate feature packets are formed corresponding to N original feature packets and N original feature packets. It can be seen that the same intermediate feature packet contains both the intra-group cross feature (second order) and the original feature (first order) of the ebadd feature vector. The resulting emmbedding feature vector within the N intermediate feature packets, which contains the second order cross features, is the feature used to perform the inter-group cross.
And 3B, performing inter-group feature intersection on N intermediate feature groups of the candidate resource objects to form inter-group intersection features of the candidate resource objects. Specifically, the inter-group feature cross is performed on the EMBedding feature vectors of the N intermediate feature groups, and the inter-group cross feature of the third order and the inter-group cross feature of the fourth order are produced,
step 3C: and carrying out feature addition on the N middle feature groups of the candidate resource object to obtain intra-group cross features of the candidate resource object.
Then, the second-order intra-group cross feature, the third-order inter-group cross feature and the fourth-order inter-group cross feature are integrated with the first-order ebadd feature vector. The first-order linear characteristics (including original characteristics and linear combination characteristics), the second-order intra-group cross characteristics and the third-order and fourth-order inter-group cross characteristics produced by the final multi-layer factorization machine submodel are enabled.
And 4, performing feature intersection on the N original feature groups through the neural network sub-model to construct high-order intersection features of the candidate resource objects.
In the embodiment of the present disclosure, for each candidate resource object, the high-order cross feature of the candidate resource object is obtained through a neural network sub-model, and the specific process includes the following steps:
and inputting the ebedding feature vectors in the N original feature groups into a neural network sub-model (such as a DNN model), and obtaining the high-order cross features of the candidate resource object through layer-by-layer downward transfer of the features. In the embodiment of the present disclosure, the high-order cross features include five-order cross features and above, and specifically, the structure of the neural network submodel may be set according to actual requirements by using high-order cross features of several orders.
S2022, performing CTR (Click Through Rate, click rate) estimation on the candidate resource objects according to the combined characteristics to obtain a CTR estimation result of the candidate resource objects.
Outputting the multi-order combined characteristics obtained through the two sub-models through an Output unit (Output Units) to obtain a CTR estimated result of the obtained candidate resource object, namely: and inputting the multi-order combined features into an output unit, and obtaining a CTR estimated result based on the sidoid activation function.
S203, determining a target resource object from the M candidate resource objects according to the CTR estimated result of each candidate resource object in the M candidate resource objects, and recommending the target resource object to the target user.
It should be noted that, the estimated CTR result of each candidate resource object in the M candidate resource objects includes the estimated click rate result of each candidate resource object. In an alternative embodiment, the M candidate resource objects are resource ordered according to the CTR estimate for each of the M candidate resource objects. And determining a plurality of candidate resource objects ranked in front according to the resource ranking result and the number of the display pits, and recommending the determined plurality of candidate resource objects to a target user.
In another alternative embodiment, the method specifically comprises the following steps: acquiring priority intervention information of the M candidate resource objects; integrating priority intervention information of the M candidate resource objects and CTR estimated results of each candidate resource object in the M candidate resource objects to obtain final recommendation ordering of the M candidate resource objects; and determining a target resource object from the M candidate resource objects according to the number of the display pits and the final recommendation sequence.
In particular, the priority intervention information is to increase the recommendation priority of a specific resource object, and in a specific implementation process, the specific resource object may be a coupon of a merchant who has put an advertisement, namely an advertisement coupon. In the case of a map scene, the particular resource object may be a user-specified type of function point, such as a gas station. For example, consider an advertisement coupon: if the advertisement coupon is already in the recommendation sequence before the recommendation priority of the advertisement coupon is not improved, the recommendation priority of the advertisement coupon is improved to the first sequence, and the final recommendation sequence is obtained; if the advertisement coupon is not within the recommendation ordering before the recommendation priority of the advertisement coupon is not increased, the recommendation priority of the advertisement coupon is increased to the end of the recommendation ordering to obtain a final recommendation ordering.
An interactive embodiment is given below taking coupons as an example: in the process of using a certain offline merchant APP, when reaching a certain mall B, in response to the search operation of the user A or no user operation trigger, recall all coupons provided by all merchants within 1000m of the current real-time position (B mall) of the user A as candidate coupons, and if the 30 candidate coupons provided by 20 offline merchants are recalled. Extracting the coupon characteristics of all the recalled candidate coupons and the user characteristics of the user A, carrying out characteristic intersection on the coupon characteristics of each candidate coupon and the user characteristics of the user A through the combined estimation model to obtain the combined characteristics of the candidate coupons, wherein the combined characteristics comprise first-order linear characteristics and multi-order intersection characteristics, and carrying out click rate estimation on the candidate coupons according to the combined characteristics of each candidate coupon to obtain click rate estimation results of each candidate coupon in 30 candidate coupons. And finally, sorting the 30 candidate coupons according to the click rate, and recommending 3 coupons with the last recommended sorting to the user A if 5 pits are displayed.
Through the technical scheme, CTR prediction of the candidate resource object is simultaneously influenced by the first-order linear characteristic, the second-order intra-group cross characteristic, the third-order inter-group cross characteristic and the fourth-order inter-group cross characteristic and the higher-order combination characteristic. The feature information is enriched, so that the characteristic of the resource can be reflected better, the click rate of each candidate coupon recalled by the target user can be predicted more accurately, a plurality of coupons ranked in front are recommended to the user based on the click rate ranking result, the accuracy and rationality of recommending the resource to the user can be improved, the success of recommending the resource to the user at one time is ensured, and repeated and invalid recommendation is avoided.
In a second aspect, based on the same inventive concept as the resource recommendation method in the foregoing embodiment, an embodiment of the present disclosure provides a resource recommendation device, referring to fig. 4, including:
a feature extraction unit 401, configured to obtain a user feature of a target user and resource object features of M candidate resource objects, where M is a positive integer;
a CTR estimation unit 402, configured to perform CTR estimation on the M candidate resource objects, to obtain a CTR estimation result of each candidate resource object in the M candidate resource objects, respectively:
wherein the CTR estimation is performed for the M candidate resource objects, respectively, including: for each candidate resource object, performing feature intersection on the resource object features of the candidate resource object and the user features through a combined prediction model to obtain combined features of the candidate resource object, and performing CTR prediction on the candidate resource object according to the combined features to obtain a CTR prediction result of the candidate resource object, wherein the combined features comprise linear features, intra-group intersection features and inter-group intersection features which are obtained according to a multi-layer factorizer sub-model of the combined prediction model, and higher-order intersection features which are obtained according to a neural network sub-model of the combined prediction model;
and the resource recommending unit 403 is configured to determine a target resource object from the M candidate resource objects according to a CTR estimation result of each candidate resource object in the M candidate resource objects, and recommend the target resource object to the target user.
In an alternative embodiment, the CTR estimating unit includes:
the characteristic preprocessing subunit is used for grouping and embedding the user characteristics and the resource object characteristics of the candidate resource object to obtain N original characteristic groups of the candidate resource object, wherein N is a positive integer;
an intra-group crossing subunit, configured to perform intra-group feature crossing on the N original feature groups through the multi-layer factorizer sub-model, to obtain N intra-group crossing results of the candidate resource object;
the inter-group crossing subunit is configured to perform feature addition, inter-group feature crossing, and integration of original features in the N original feature groups on the N intra-group crossing results and the N original feature groups of the candidate resource object through the multi-layer factorizer submodel, so as to generate a linear feature, an intra-group crossing feature, and an inter-group crossing feature of the candidate resource object;
and the high-order crossing subunit is used for carrying out feature crossing on the N original feature groups through the neural network sub-model to construct high-order crossing features of the candidate resource object.
In an alternative embodiment, the inter-group crossing sub-unit is specifically configured to:
embedding the N groups of intra-group crossing results of the candidate resource object, and embedding the original features of the N original feature groups to form N intermediate feature groups of the candidate resource object;
performing inter-group feature intersection on N intermediate feature groups of the candidate resource object to form inter-group intersection features of the candidate resource object;
and carrying out feature addition on the N intermediate feature groups of the candidate resource object to obtain intra-group cross features of the candidate resource object.
In an alternative embodiment, the apparatus further includes a resource recall unit, configured to obtain the M candidate resource objects by:
and recalling candidate resource objects provided by merchant equipment within a preset distance range of the real-time position of the target user according to a location-based service LBS recall strategy to obtain the M candidate resource objects.
In an alternative embodiment, the resource recommendation unit 403 is specifically configured to:
acquiring priority intervention information of the M candidate resource objects;
integrating priority intervention information of the M candidate resource objects and CTR estimated results of each candidate resource object in the M candidate resource objects to obtain final recommendation ordering of the M candidate resource objects;
and determining the target resource object from the M candidate resource objects according to the number of the display pits and the final recommendation sequence.
In a third aspect, based on the same inventive concept as the resource recommendation method in the foregoing embodiment, the embodiment of the present disclosure further provides a server, as shown in fig. 5, including a memory 504, a processor 502, and a computer program stored in the memory 504 and capable of running on the processor 502, where the processor 502 implements the steps of any embodiment of the resource recommendation method described in the foregoing when executing the program.
Where in FIG. 5 a bus architecture (represented by bus 500), bus 500 may include any number of interconnected buses and bridges, with bus 500 linking together various circuits, including one or more processors, represented by processor 502, and memory, represented by memory 504. Bus 500 may also link together various other circuits such as peripheral devices, voltage regulators, power management circuits, etc., as are well known in the art and, therefore, will not be described further herein. Bus interface 506 provides an interface between bus 500 and receiver 501 and transmitter 503. The receiver 501 and the transmitter 503 may be the same element, i.e. a transceiver, providing a means for communicating with various other apparatus over a transmission medium. The processor 502 is responsible for managing the bus 500 and general processing, while the memory 504 may be used to store data used by the processor 502 in performing operations.
In a fourth aspect, based on the inventive concept of resource recommendation as in the previous embodiments, the present embodiments further provide a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of any of the embodiments of the resource recommendation method described above.
The present description is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the specification. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
While preferred embodiments of the present description have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. It is therefore intended that the following claims be interpreted as including the preferred embodiments and all such alterations and modifications as fall within the scope of the disclosure.
It will be apparent to those skilled in the art that various modifications and variations can be made in the present specification without departing from the spirit or scope of the specification. Thus, if such modifications and variations of the present specification fall within the scope of the claims and the equivalents thereof, the present specification is also intended to include such modifications and variations.
Claims (10)
1. A resource recommendation method, comprising:
acquiring user characteristics of a target user and resource object characteristics of M candidate resource objects, wherein M is a positive integer;
CTR estimation is respectively carried out on the M candidate resource objects to obtain a CTR estimation result of each candidate resource object in the M candidate resource objects, and the CTR estimation result comprises the following steps: for each candidate resource object, carrying out feature intersection on the resource object features of the candidate resource object and the user features through a combined prediction model to obtain combined features of the candidate resource object, and carrying out CTR prediction on the candidate resource object according to the combined features to obtain a CTR prediction result of the candidate resource object; the method comprises the steps of grouping and embedding the user characteristics and the resource object characteristics of the candidate resource object to obtain N original characteristic groups of the candidate resource object, wherein N is a positive integer; performing intra-group feature intersection on the N original feature groups through a multi-layer factorizer sub-model to obtain N groups of intra-group intersection results of the candidate resource objects; performing feature addition, inter-group feature intersection and integration of original features in the N original feature groups on N groups of intra-group intersection results of the candidate resource object and the N original feature groups through the multi-layer factorization machine sub-model to generate linear features, intra-group intersection features and inter-group intersection features of the candidate resource object; performing feature intersection on the N original feature groups through a neural network sub-model to construct high-order intersection features of the candidate resource objects;
and determining a target resource object from the M candidate resource objects according to the CTR estimated result of each candidate resource object in the M candidate resource objects, and recommending the target resource object to the target user.
2. The method of claim 1, wherein the feature-crossing the resource object feature of the candidate resource object with the user feature by combining a pre-estimated model to obtain the combined feature of the candidate resource object, comprises:
grouping and embedding the user characteristics and the resource object characteristics of the candidate resource object to obtain N original characteristic groups of the candidate resource object, wherein N is a positive integer;
performing intra-group feature intersection on the N original feature groups through the multi-layer factorizer sub-model to obtain N groups of intra-group intersection results of the candidate resource objects;
performing feature addition, inter-group feature intersection and integration of original features in the N original feature groups on N groups of intra-group intersection results of the candidate resource object and the N original feature groups through the multi-layer factorization machine sub-model to generate linear features, intra-group intersection features and inter-group intersection features of the candidate resource object;
and performing feature intersection on the N original feature groups through the neural network sub-model to construct high-order intersection features of the candidate resource objects.
3. The method of claim 2, wherein said performing feature addition and inter-group feature intersection on the N sets of intra-group intersection results and the N original feature groupings of the candidate resource objects by the multi-layer factorizer submodel comprises:
embedding the N groups of intra-group crossing results of the candidate resource object, and embedding the original features of the N original feature groups to form N intermediate feature groups of the candidate resource object;
performing inter-group feature intersection on N intermediate feature groups of the candidate resource object to form inter-group intersection features of the candidate resource object;
and carrying out feature addition on the N intermediate feature groups of the candidate resource object to obtain intra-group cross features of the candidate resource object.
4. The method of claim 1, the step of obtaining the M candidate resource objects comprising:
and recalling candidate resource objects provided by merchant equipment within a preset distance range of the real-time position of the target user according to a location-based service LBS recall strategy to obtain the M candidate resource objects.
5. A resource recommendation device, comprising:
the feature extraction unit is used for acquiring user features of a target user and resource object features of M candidate resource objects, wherein M is a positive integer;
the CTR estimation unit is used for respectively carrying out CTR estimation on the M candidate resource objects to obtain a CTR estimation result of each candidate resource object in the M candidate resource objects:
the performing CTR estimation on the M candidate resource objects respectively includes: for each candidate resource object, the resource object features of the candidate resource object and the user features are subjected to feature intersection through a combined prediction model to obtain combined features of the candidate resource object, CTR prediction is carried out on the candidate resource object according to the combined features to obtain a CTR prediction result of the candidate resource object,
for each candidate resource object, performing feature intersection on the resource object features of the candidate resource object and the user features through a combined pre-estimation model to obtain combined features of the candidate resource object, including: grouping and embedding the user characteristics and the resource object characteristics of the candidate resource object to obtain N original characteristic groups of the candidate resource object, wherein N is a positive integer; performing intra-group feature intersection on the N original feature groups through a multi-layer factorizer sub-model to obtain N groups of intra-group intersection results of the candidate resource objects; performing feature addition, inter-group feature intersection and integration of original features in the N original feature groups on N groups of intra-group intersection results of the candidate resource object and the N original feature groups through the multi-layer factorization machine sub-model to generate linear features, intra-group intersection features and inter-group intersection features of the candidate resource object; performing feature intersection on the N original feature groups through a neural network sub-model to construct high-order intersection features of the candidate resource objects;
and the resource recommending unit is used for determining a target resource object from the M candidate resource objects according to the CTR estimated result of each candidate resource object in the M candidate resource objects and recommending the target resource object to the target user.
6. The apparatus of claim 5, the inter-group crossing sub-unit being specifically configured to:
embedding the N groups of intra-group crossing results of the candidate resource object, and embedding the original features of the N original feature groups to form N intermediate feature groups of the candidate resource object;
performing inter-group feature intersection on N intermediate feature groups of the candidate resource object to form inter-group intersection features of the candidate resource object;
and carrying out feature addition on the N intermediate feature groups of the candidate resource object to obtain intra-group cross features of the candidate resource object.
7. The apparatus of claim 5, further comprising a resource recall unit to obtain the M candidate resource objects by:
and recalling candidate resource objects provided by merchant equipment within a preset distance range of the real-time position of the target user according to a location-based service LBS recall strategy to obtain the M candidate resource objects.
8. The apparatus according to any of claims 5-7, the resource recommendation unit being specifically configured to:
acquiring priority intervention information of the M candidate resource objects;
integrating priority intervention information of the M candidate resource objects and CTR estimated results of each candidate resource object in the M candidate resource objects to obtain final recommended ordering of the M candidate resource objects;
and determining the target resource object from the M candidate resource objects according to the number of the display pits and the final recommendation sequence.
9. A server comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the method of any one of claims 1-4 when the program is executed.
10. A computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the method of any of claims 1-4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911398736.8A CN111340522B (en) | 2019-12-30 | 2019-12-30 | Resource recommendation method, device, server and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911398736.8A CN111340522B (en) | 2019-12-30 | 2019-12-30 | Resource recommendation method, device, server and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111340522A CN111340522A (en) | 2020-06-26 |
| CN111340522B true CN111340522B (en) | 2024-03-08 |
Family
ID=71179609
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911398736.8A Active CN111340522B (en) | 2019-12-30 | 2019-12-30 | Resource recommendation method, device, server and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111340522B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113157951A (en) * | 2021-03-26 | 2021-07-23 | 北京达佳互联信息技术有限公司 | Multimedia resource processing method, device, server and storage medium |
| CN113204697A (en) * | 2021-04-29 | 2021-08-03 | 五八有限公司 | Searching method, searching device, electronic equipment and storage medium |
| CN113327131A (en) * | 2021-06-03 | 2021-08-31 | 太原理工大学 | Click rate estimation model for feature interactive selection based on three-branch decision theory |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106446005A (en) * | 2015-08-07 | 2017-02-22 | 谷歌公司 | Factorized models |
| WO2017084344A1 (en) * | 2015-11-16 | 2017-05-26 | 北京百度网讯科技有限公司 | Information recommendation method and apparatus |
| WO2018090793A1 (en) * | 2016-11-18 | 2018-05-24 | 腾讯科技（深圳）有限公司 | Multimedia recommendation method and device |
| CN108132963A (en) * | 2017-11-23 | 2018-06-08 | 广州优视网络科技有限公司 | Resource recommendation method and device, computing device and storage medium |
| WO2018212711A1 (en) * | 2017-05-19 | 2018-11-22 | National University Of Singapore | Predictive analysis methods and systems |
| WO2018212710A1 (en) * | 2017-05-19 | 2018-11-22 | National University Of Singapore | Predictive analysis methods and systems |
| CN109408724A (en) * | 2018-11-06 | 2019-03-01 | 北京达佳互联信息技术有限公司 | Multimedia resource estimates the determination method, apparatus and server of clicking rate |
| CN110008399A (en) * | 2019-01-30 | 2019-07-12 | 阿里巴巴集团控股有限公司 | A kind of training method and device, a kind of recommended method and device of recommended models |
| CN110020156A (en) * | 2017-12-07 | 2019-07-16 | 深圳市腾讯计算机系统有限公司 | Information recommendation method, front end implementation method, device, equipment and storage medium |
| CN110046952A (en) * | 2019-01-30 | 2019-07-23 | 阿里巴巴集团控股有限公司 | A kind of training method and device, a kind of recommended method and device of recommended models |
| WO2019141072A1 (en) * | 2018-01-22 | 2019-07-25 | 阿里巴巴集团控股有限公司 | Method, device, and client for recommending store information |
| CN110162693A (en) * | 2019-03-04 | 2019-08-23 | 腾讯科技（深圳）有限公司 | A kind of method and server of information recommendation |
| CN110263243A (en) * | 2019-01-23 | 2019-09-20 | 腾讯科技（深圳）有限公司 | Media information recommending method, apparatus, storage medium and computer equipment |
-
2019
- 2019-12-30 CN CN201911398736.8A patent/CN111340522B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106446005A (en) * | 2015-08-07 | 2017-02-22 | 谷歌公司 | Factorized models |
| WO2017084344A1 (en) * | 2015-11-16 | 2017-05-26 | 北京百度网讯科技有限公司 | Information recommendation method and apparatus |
| WO2018090793A1 (en) * | 2016-11-18 | 2018-05-24 | 腾讯科技（深圳）有限公司 | Multimedia recommendation method and device |
| WO2018212711A1 (en) * | 2017-05-19 | 2018-11-22 | National University Of Singapore | Predictive analysis methods and systems |
| WO2018212710A1 (en) * | 2017-05-19 | 2018-11-22 | National University Of Singapore | Predictive analysis methods and systems |
| CN108132963A (en) * | 2017-11-23 | 2018-06-08 | 广州优视网络科技有限公司 | Resource recommendation method and device, computing device and storage medium |
| CN110020156A (en) * | 2017-12-07 | 2019-07-16 | 深圳市腾讯计算机系统有限公司 | Information recommendation method, front end implementation method, device, equipment and storage medium |
| WO2019141072A1 (en) * | 2018-01-22 | 2019-07-25 | 阿里巴巴集团控股有限公司 | Method, device, and client for recommending store information |
| CN109408724A (en) * | 2018-11-06 | 2019-03-01 | 北京达佳互联信息技术有限公司 | Multimedia resource estimates the determination method, apparatus and server of clicking rate |
| CN110263243A (en) * | 2019-01-23 | 2019-09-20 | 腾讯科技（深圳）有限公司 | Media information recommending method, apparatus, storage medium and computer equipment |
| CN110046952A (en) * | 2019-01-30 | 2019-07-23 | 阿里巴巴集团控股有限公司 | A kind of training method and device, a kind of recommended method and device of recommended models |
| CN110008399A (en) * | 2019-01-30 | 2019-07-12 | 阿里巴巴集团控股有限公司 | A kind of training method and device, a kind of recommended method and device of recommended models |
| CN110162693A (en) * | 2019-03-04 | 2019-08-23 | 腾讯科技（深圳）有限公司 | A kind of method and server of information recommendation |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111340522A (en) | 2020-06-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP5881153B2 (en) | Recommendation of correlated information | |
| JP6511455B2 (en) | Method and system for extracting features of user behavior and personalizing recommendations | |
| CN111340522B (en) | Resource recommendation method, device, server and storage medium | |
| CN111242748B (en) | Method, apparatus, and storage medium for recommending items to a user | |
| EP4181026A1 (en) | Recommendation model training method and apparatus, recommendation method and apparatus, and computer-readable medium | |
| US8732015B1 (en) | Social media pricing engine | |
| CN104951465A (en) | Application recommendation method and device | |
| CN106952166A (en) | The user force evaluation method and device of a kind of social platform | |
| CN102208086A (en) | Field-oriented personalized intelligent recommendation system and implementation method | |
| CN111798280B (en) | Multimedia information recommendation method, device and equipment and storage medium | |
| EP4242955A1 (en) | User profile-based object recommendation method and device | |
| CN110134783B (en) | Personalized recommendation method, device, equipment and medium | |
| CN112989169B (en) | Target object identification method, information recommendation method, device, equipment and medium | |
| CN108446297A (en) | A kind of recommendation method and device, electronic equipment | |
| CN103365842B (en) | A kind of page browsing recommends method and device | |
| KR20150113644A (en) | Method for selecting recommendation type in product recommeding system | |
| CN111489196B (en) | Prediction method and device based on deep learning network, electronic equipment and medium | |
| CN115358807A (en) | Article recommendation method and device, storage medium and electronic equipment | |
| JP6810184B2 (en) | Calculation device, calculation method and calculation program | |
| CN111967924A (en) | Commodity recommendation method, commodity recommendation device, computer device, and medium | |
| CN112561000B (en) | Classification method, device, equipment and storage medium based on combined model | |
| CN115238188A (en) | Object recommendation method and system and object recommendation model system | |
| Liu et al. | A network quotation framework for customised parts through rough requests | |
| CN113469786A (en) | Method and device for recommending articles, computer equipment and storage medium | |
| CN114493674A (en) | Advertisement click rate prediction model and method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |