CN112632405B - Recommendation method, recommendation device, recommendation equipment and storage medium - Google Patents

Recommendation method, recommendation device, recommendation equipment and storage medium Download PDF

Info

Publication number
CN112632405B
CN112632405B CN202011633927.0A CN202011633927A CN112632405B CN 112632405 B CN112632405 B CN 112632405B CN 202011633927 A CN202011633927 A CN 202011633927A CN 112632405 B CN112632405 B CN 112632405B
Authority
CN
China
Prior art keywords
user
tag
target
item
matching degree
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
Application number
CN202011633927.0A
Other languages
Chinese (zh)
Other versions
CN112632405A (en
Inventor
胡舒
何晨
徐彦廷
罗子科
李树山
胡磊
黄惠文
周耿辉
蔡晓铃
林淡红
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Digital Guangdong Network Construction Co Ltd
Original Assignee
Digital Guangdong Network Construction Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Digital Guangdong Network Construction Co Ltd filed Critical Digital Guangdong Network Construction Co Ltd
Priority to CN202011633927.0A priority Critical patent/CN112632405B/en
Publication of CN112632405A publication Critical patent/CN112632405A/en
Application granted granted Critical
Publication of CN112632405B publication Critical patent/CN112632405B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/90Details of database functions independent of the retrieved data types
    • G06F16/95Retrieval from the web
    • G06F16/953Querying, e.g. by the use of web search engines
    • G06F16/9536Search customisation based on social or collaborative filtering

Landscapes

  • Engineering & Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The embodiment of the invention discloses a recommendation method, a recommendation device, recommendation equipment and a storage medium. The method comprises the following steps: acquiring a user tag of a target user, and determining each reference item corresponding to the tag attribute of the user tag; calculating a first matching degree of the target user and each reference item according to the weight corresponding to the user tag and the similarity between the item to which the user tag belongs and each reference item; calculating a second matching degree between the target user and each reference item through a collaborative filtering method based on the corresponding relation between the similar user of the target user and each reference item; and calculating the weighted sum of each first matching degree and each second matching degree according to the preset matching degree weight, sequencing the reference items corresponding to each weighted sum, determining the target items according to the sequencing result, and recommending the target items to the target user. According to the embodiment of the invention, the matching degree of the target user and the possibly interested reference matters is calculated from different angles, the target matters to be recommended are determined, and the accuracy of recommending the handling matters to the user is improved.

Description

Recommendation method, recommendation device, recommendation equipment and storage medium
Technical Field
The embodiment of the invention relates to computer technology, in particular to a recommendation method, a recommendation device, recommendation equipment and a storage medium.
Background
At present, a user can transact business through a government service website only by actively searching or classifying and searching, so that the transacted business is found out in a single mode. Because of the similarity of item names and the difference of the item names with the daily words of users, such as the spoken birth control points, the administrative names are called as' the business insurance participants check the business, and the users can hardly find the required service through active searching. In addition, with the continuous entry of the item guide, taking Guangdong as an example, 180 tens of thousands of guide information exists at present, and the user needs to find out the office guide which accords with the user from mass information, so that the searching difficulty is further increased.
Although intelligent recommendation plates are partially offered in the market at present, the intelligent recommendation plates are basically random recommendation, and deep exploration is not performed on accuracy and thousands of people. Therefore, how to improve the accuracy of recommending transactions to users is a technical problem to be solved.
Disclosure of Invention
The embodiment of the invention provides a recommendation method, a recommendation device, recommendation equipment and a storage medium, which can improve the accuracy of recommending transacting items to a user.
In a first aspect, an embodiment of the present invention provides a recommendation method, including:
acquiring a user tag of a target user, and determining each reference item corresponding to the tag attribute of the user tag;
Calculating a first matching degree of the target user and each reference item according to the weight corresponding to the user tag and the similarity between the item to which the user tag belongs and each reference item;
Calculating a second matching degree between the target user and each reference item through a collaborative filtering method based on the corresponding relation between the similar user of the target user and each reference item;
And calculating weighted sums of the first matching degree and the second matching degree according to preset matching degree weights, sequencing reference matters corresponding to the weighted sums, determining target matters according to sequencing results, and recommending the target matters to a target user.
In a second aspect, an embodiment of the present invention further provides a recommendation apparatus, including:
the tag acquisition module is used for acquiring a user tag of a target user and determining each reference item corresponding to the tag attribute of the user tag;
the first matching degree calculation module is used for calculating the first matching degree of the target user and each reference item according to the weight corresponding to the user tag and the similarity of the item to which the user tag belongs and each reference item;
the second matching degree calculation module is used for calculating the second matching degree of the target user and each reference item through a collaborative filtering method based on the corresponding relation between the similar users of the target user and each reference item;
And the result recommending module is used for calculating the weighted sum of the first matching degree and the second matching degree according to the preset matching degree weight, sequencing the reference items corresponding to the weighted sums, determining the target items according to the sequencing result and recommending the target items to the target user.
In a third aspect, an embodiment of the present invention further provides a recommendation device, where the recommendation device includes:
One or more processors;
A memory for storing one or more programs,
The one or more programs, when executed by the one or more processors, cause the one or more processors to implement the recommended methods as provided by any embodiment of the invention.
In a fourth aspect, embodiments of the present invention also provide a storage medium containing computer-executable instructions which, when executed by a computer processor, are used to perform a recommendation method as provided by any of the embodiments of the present invention.
According to the embodiment of the invention, all the reference items corresponding to the label attributes of the user labels are determined based on the user labels of the target users, the first matching degree and the second matching degree of the target users and all the reference items are calculated from two angles of the relation between the user labels of the target users and all the reference items and the relation between the similar users of the target users and all the reference items, the matching degree of the target users and all the reference items is comprehensively judged in a weighted summation mode, all the matching degrees are ordered, and then the target items are determined according to the ordering result and recommended to the target users. According to the embodiment of the invention, all the reference matters which are possibly interested by the target user are determined based on the user tag of the target user, the matching degree of the target user and all the reference matters is calculated from different angles, so that the target matters which can be recommended to the target user are determined, the accuracy of calculating the matching degree between the user and the matters to be recommended is improved, and the accuracy of recommending the matters to be processed to the user is further improved.
Drawings
FIG. 1 is a flowchart of a recommendation method according to a first embodiment of the present invention;
FIG. 2 is a schematic diagram of a collaborative filtering method based on a user according to a first embodiment of the present invention;
FIG. 3 is a schematic diagram of a collaborative filtering method based on matters according to a first embodiment of the present invention;
FIG. 4 is a flowchart of another recommendation method according to the second embodiment of the present invention;
FIG. 5 is a flowchart of a recommendation method according to a second embodiment of the present invention;
FIG. 6 is a block diagram of an intelligent recommendation system according to a third embodiment of the present invention;
FIG. 6a is a block diagram of a basic data warehouse according to a third embodiment of the present invention;
FIG. 6b is a schematic diagram of a transaction modeling system according to a third embodiment of the present invention;
FIG. 6c is a block diagram of a user portrayal system according to a third embodiment of the present invention;
FIG. 6d is a schematic diagram of an intelligent recommendation center according to a third embodiment of the present invention;
FIG. 6e is a block diagram of an application management system according to a third embodiment of the present invention;
FIG. 6f is a schematic diagram of an optimization model according to a third embodiment of the present invention;
fig. 7 is a schematic structural diagram of a recommending apparatus according to a fourth embodiment of the present invention;
Fig. 8 is a schematic structural diagram of a recommendation device according to a fifth embodiment of the present invention.
Detailed Description
The invention is described in further detail below with reference to the drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting thereof. It should be further noted that, for convenience of description, only some, but not all of the structures related to the present invention are shown in the drawings.
Example 1
Fig. 1 is a flowchart of a recommendation method provided in an embodiment of the present invention, where the method may be performed by a recommendation device, and the device may be implemented in software and/or hardware, where the recommendation method may be applicable to a case where a user may need to perform a transaction. The apparatus may be configured in a recommendation device. As shown in fig. 1, the method includes:
Step S110, obtaining a user tag of a target user, and determining each reference item corresponding to the tag attribute of the user tag.
It should be noted that the target user may be a user who operates a government service website.
User tags may be used to characterize a user's interests, preferences, needs, etc. for a matter. Wherein, the item may be an item that the user can transact or browse. For example, the transaction may be a government service web page, a transaction guide page, a comment area for a office page or a office page, etc. Each user may correspond to a user identifier that is used to uniquely identify the user to distinguish and single point locate the user. The user tag can be obtained by extracting attribute information and/or behavior data of the user, analyzing the user characteristics and storing the user characteristics in a tag form. Wherein the attribute information may be used to characterize the static attributes of the user. For example, the attribute information may include the age of the user, identification card information, work information, and the like. The behavior data may be used to characterize user behavior-related data of the user. For example, browsing a government service web page, browsing a transaction guide page, searching for "certificates", commenting on a transaction guide for handling a port-australia pass, and approving a next port-australia pass, etc. can be regarded as user behaviors of the user, and the behavior data can be data generated when the user performs these user behaviors.
For example, a tag may be established for a user based on attribute information of the user. For example, it may be determined whether the address in the user's identification card information is located in the same province as the present residence in the work information, and if the address is located in a different province, it is indicated that the user may work in a different place, and a user tag such as a residence card tag or a social security card tag may be established for the user. Tags may also be established for users based on their behavioral data. For example, each item handled by the user may be acquired, item tags corresponding to each item may be determined, and for target item tags whose number exceeds a set threshold, a user tag corresponding to the target item tag may be established for the user. For example, a user transacts a plurality of items on the web, each of which corresponds to an online transacted item tag, so that a web-sponsor's user tag can be established for the user. Labels can also be established for users in combination with their attribute information and behavioral data. For example, a user tag for a startup is created for a user based on startup attributes entered by the user during the life stage and a historical search record of the user in relation to the startup.
The user tag may be stored in association with the weight. Each user tag may correspond to a weight that may be used to quantify a user's interest in an item, preferences, requirements, and the like. For example, the weight may be an interest index, a preference index, a demand level, or the like of the item to which the tag belongs, or may be understood as a credibility of the tag or a probability of the item to which the tag belongs when analyzing a preference of the user. The weights corresponding to the user tags may be initialized, updated in real-time, and updated periodically. Specifically, the initialization of the weights can process the data in the basic data warehouse through a big data technology, and corresponding initial weights are configured for preset user tags. The real-time updating of the weights may be to update the initial weights of the user tags in real-time by analyzing the user behavior when the user behavior related to the user tags is detected. For example, the behavior type of the user behavior can be determined, the behavior type can reflect the interest degree of the user on the matters to which the user tag belongs, and then the weight of the user tag is updated in real time according to the interest degree of the user. The timed updating of the weight may be updating the weight of the user tag at a preset period. For example, a change record of the weight in the preset period is obtained, and if the weight remains unchanged in the preset period, the weight can be reduced according to a preset reducing rule. That is, if the weight of the user tag remains unchanged in the preset period, it is indicated that the user does not transact or browse the item to which the user tag belongs in the preset period, so that it is indicated that the user is not interested in the item, and the weight of the user tag can be reduced.
The tag attributes of the user tags may be used to determine other user tags associated with the user tag. For example, the tag attributes may include a general tag, an integrated tag, or an associated tag, among others. The same user tag may correspond to at least one tag attribute. For example, for a user tag: the residence permit may correspond to a general label: social documents, comprehensive labels: and (5) transacting other user labels such as social security cards with association relation with the residence permit on the network. The same tag attribute may correspond to at least one user tag. Each user tag may belong to at least one item. For example, the residence may pertain to handling residence events, browsing residence transaction guidelines, and the like.
The reference item may be an item to which all user tags corresponding to tag attributes of the user tags belong, for determining an item related to the user tag to recommend to the target user.
Specifically, a user tag of a target user is obtained, the tag attribute of the user tag is determined, all the user tags under the tag attribute are determined according to the tag attribute, and then all the reference items to which all the user tags belong are determined.
Illustratively, a user tag of a target user is obtained: and if the social security card determines that the tag attribute of the social security card is social credentials and online processing, determining each reference item related to the social security card according to the tag attribute, wherein the reference item can comprise a harbor and australian pass with the tag attribute of the same social credentials and a resident certificate with the tag attribute of the same online processing.
Step S120, calculating a first matching degree between the target user and each reference item according to the weight corresponding to the user tag and the similarity between the item to which the user tag belongs and each reference item.
The similarity can be used for representing the similarity degree of the items to which the user tag belongs and each reference item. For example, the similarity can be obtained by calculating the similarity between the item feature vector of the item to which the user tag belongs and the item feature vector of each reference item.
Specifically, a weight corresponding to the user tag is obtained, the similarity between the item to which the user tag belongs and each reference item is calculated, and the first matching degree between the target user and each reference item is calculated according to the weight and the similarity. For example, the social security card tag of the target user has a weight of 0.6, the similarity between feature vectors of things such as social security cards and residence certificates is calculated, and the first matching degree between the target user and things such as residence certificates is calculated based on 0.6 and the similarity. The first matching degree can be calculated through a preset formula based on the weight and the similarity. For example, a weighted sum of the weight and the similarity is calculated, the weighted sum is taken as the first matching degree, or a product of the weight and the similarity is calculated, and the product result is taken as the first matching degree. Or calculating the first matching degree through other preset formulas.
Step S130, calculating a second matching degree between the target user and each reference item through a collaborative filtering method based on the corresponding relation between the similar user of the target user and each reference item.
The similar users of the target user may be users having similar preferences as the target user. For example, the similar user may be determined by calculating the similarity between the target user and other users, and the user whose similarity is not lower than the preset threshold is taken as the similar user. The similar users also have corresponding user tags, each user tag belonging item can comprise a reference item, and the interest degree of the similar users for each reference item can be determined based on the corresponding relation between the similar users and each reference item and the weight of the similar users corresponding to the reference item.
Specifically, each reference user tag corresponding to each reference item is determined, the weight of a similar user of the target user for each reference user tag is obtained, the weight of the target user and each reference user tag is calculated through a collaborative filtering method according to the weight of the similar user for each reference user tag, and then the second matching degree of the target user and each reference item is determined according to the corresponding relation between each reference user tag and each reference item.
Illustratively, collaborative filtering methods may include user-based collaborative filtering methods, transaction-based collaborative filtering methods, and model-based collaborative filtering methods.
Wherein the algorithm logic of the collaborative filtering method based on the user can be that similar people have the same preference. For example, by analyzing the weights of the similar users and the labels of the reference users, the user similarity of the target user and the similar users is calculated, and the weights of the target user and the labels of the reference users are determined according to the weights and the user similarity. Fig. 2 is a schematic diagram of a collaborative filtering method based on a user according to a first embodiment of the present invention. As shown in fig. 2, the target user a has similar users b and c, items A, B, C and D correspond to different reference user tags a, b, c and D, respectively, and weights of the second and third items A, B, C and D are determined based on weights corresponding to the reference user tags a, b, c and D of the second and third items, and further weights of the first and third items A, B, C and D are determined based on user similarity of the first and second items, user similarity of the first and third items, and weights of the second and third items A, B, C and D.
The algorithmic logic of the collaborative filtering approach based on items may be "users who like an item would like similar items". For example, the item similarity of each reference user tag is calculated by analyzing the weights of the similar users and each reference user tag, and the weights of the target users and each reference user tag are determined according to the weights and the item similarity. Fig. 3 is a schematic diagram of a collaborative filtering method based on matters according to a first embodiment of the present invention. As shown in fig. 3, the target user c has similar user a and b, items A, B, C and D correspond to different reference user tags a, b, c and D, weights of the first and second items A, B, C and D are determined based on weights corresponding to the reference user tags a, b, c and D of the first and second items, item similarities of two-by-two combinations in A, B, C and D are calculated, and then weights of the third item A, B, C and D are determined according to the item similarities and weights.
The algorithm logic of the collaborative filtering method based on the model can be to train a recommendation model based on the weights of the sample users and all matters, so that after the target users, the similar users of the target users and the labels of the reference users are input into the recommendation model, the weights of the target users and the labels of the reference users can be output through the recommendation model. The weights of the sample user and all matters can be obtained by determining all reference user labels corresponding to all matters based on the weights of the sample user for all reference user labels.
It should be noted that, the step S120 and the step S130 are not limited to the above-described execution sequence, and the step S130 may be executed first and then the step S120 may be executed, or the step S120 and the step S130 may be executed in parallel.
And step 140, calculating weighted sums of the first matching degree and the second matching degree according to preset matching degree weights, sequencing the reference items corresponding to the weighted sums, determining target items according to sequencing results, and recommending the target items to a target user.
The preset matching degree weight may be weights respectively assigned to the first matching degree and the second matching degree when a weighted sum is calculated on the first matching degree and the second matching degree, and is used for comprehensively determining the matching degree of the target user for each reference item. The preset matching degree weight may be predetermined, and the preset matching degree weight of the first matching degree and the preset matching degree weight of the second matching degree are directly obtained when the weighted sum of the first matching degree and the second matching degree is calculated. Or the preset matching degree weight can be determined based on the historical operation condition of the target user at the government service website. The first matching degree is calculated based on the weight of the user label existing in the user and the similarity between matters, and has smaller user behavior relation with the target user, so that the method is more suitable for a cold start semantic environment. The second matching degree is calculated based on the corresponding relation between the similar users of the target user and each reference item, and when the similar users of the target user are determined, the similar users similar to the target user are required to be determined according to the user attributes and the user behaviors of the target user, so that the user behaviors of the target user have a larger relation, and the hot start semantic environment can be adapted. In summary, the preset matching degree weight can judge whether the semantic environment more suitable for cold start or hot start is judged according to the historical operation condition of the target user in the government service website when the target user is recommended, and the higher preset matching degree weight is distributed to the corresponding matching degree according to the judging result. For example, for a target user whose historical number of operations is below a preset threshold or newly registered, a higher preset matching degree weight may be assigned to the first matching degree, and a lower preset matching degree weight may be assigned to the second matching degree. For target users whose historical operation times are not lower than a preset threshold or can be judged to be old users, a lower preset matching degree weight can be allocated to the first matching degree, and a higher preset matching degree weight can be allocated to the second matching degree.
Specifically, a weighted sum of each first matching degree and each second matching degree is calculated according to a preset matching degree weight, each weighted sum is determined to be a reference matching degree of a target user and each reference item, each reference matching degree is ordered, target items corresponding to the target matching degree are determined according to an ordering result, and the target items are recommended to the target user. The sorting result may be a result of sorting the weighted sums in ascending or descending order. If the sorting result is a descending sorting result, a preset number of reference items which are sorted in front can be selected as target items.
According to the embodiment of the invention, all the reference items corresponding to the label attributes of the user labels are determined based on the user labels of the target users, the first matching degree and the second matching degree of the target users and all the reference items are calculated from two angles of the relation between the user labels of the target users and all the reference items and the relation between the similar users of the target users and all the reference items, the matching degree of the target users and all the reference items is comprehensively judged in a weighted summation mode, all the matching degrees are ordered, and then the target items are determined according to the ordering result and recommended to the target users. According to the embodiment of the invention, all the reference matters which are possibly interested by the target user are determined based on the user tag of the target user, the matching degree of the target user and all the reference matters is calculated from different angles, so that the target matters which can be recommended to the target user are determined, the accuracy of calculating the matching degree between the user and the matters to be recommended is improved, and the accuracy of recommending the matters to be processed to the user is further improved.
Example two
Fig. 4 is a flowchart of another recommended method provided in the second embodiment of the present invention, where the optimization is performed based on the foregoing embodiment, and as shown in fig. 4, the method includes:
step S210, obtaining user labels of target users, and determining various references corresponding to other user labels with the same attribute as the user labels.
Fig. 5 is a flowchart illustrating a recommendation method according to a second embodiment of the present invention. As shown in fig. 5, the content image may be understood as a description of the item itself. For example, whether the transaction is an online transaction, whether express delivery is possible, applicable crowd, etc. The content image includes information related to the event. The user portrayal may be understood as a description of the user. The user portrait includes information related to the user, such as user labels, weights, and user attributes. By matching the content representation with the user representation, a preliminary match can be made to the target user with the item.
Optionally, before acquiring the user tag of the target user, the method may further include:
When the user behavior of a target user is detected, determining a user tag corresponding to the user behavior;
and acquiring a behavior coefficient corresponding to the behavior type of the user behavior, and updating the weight corresponding to the user tag according to the behavior coefficient.
The behavior type of the user behavior may be an operation type of the user on the item. For example, the behavior types may include browsing, transacting, searching, commenting, collecting, or praying, etc. When the user behavior of the target user is browsing the social security card interface, the corresponding behavior type may be browsing. When the user behavior of the target user is praise a comment on the social security card communication interface, the corresponding behavior type may be praise.
The behavior coefficient can preset the interest degree of the item to which the user tag belongs according to the behavior type of the user behavior, and is used for determining the influence degree of the user behavior on the user tag. For example, the corresponding behavior coefficient of browsing may be 1.1, and since the transaction is more interested in the event than browsing, the corresponding behavior coefficient of the transaction may be set to 1.6.
Illustratively, when the target user browses the social security card interface, the user tag is a social security card and the behavior type is browsing. The behavior coefficient corresponding to the behavior type is 1.1, and the weight corresponding to the user tag is updated according to the behavior coefficient, so that the effect that the more browsing is, the larger the weight is. When the target user transacts the social security card, the user label is the social security card, and the behavior type is transacting. The behavior type may correspond to a higher behavior factor of 1.6, and the weights are updated according to the behavior factor to recommend matters related to the social security card to the target user later.
According to the method and the device, the real-time user behaviors of the target user are detected in real time, the corresponding behavior coefficients are determined according to the behavior types of the user behaviors, and then the weights of the user tags corresponding to the user behaviors are updated according to the behavior coefficients, so that the weights of the user tags of the target user are updated in real time, the requirements of the target user can be followed in real time when the matters are recommended to the target user, and the accuracy of the matters recommendation is improved.
Further, determining the user tag corresponding to the user behavior may include:
Acquiring item content of the item corresponding to the user behavior, and judging whether a user tag corresponding to the item content exists or not;
If yes, executing the step of acquiring the behavior coefficient corresponding to the behavior type of the user behavior;
If not, establishing an association relation between the item content and the target user tag, acquiring an item website of the item corresponding to the user behavior, and determining that the weight of the target user tag is the weight corresponding to the item website according to the preset association relation between the item website and the weight.
The content of the item can be the content in the website, page or screen where the item is located, and is used for determining the user tag corresponding to the item. For example, the item content may include information of item type, item description, item attributes, and the like.
The target user tag may be a user tag newly set for the target user when the user tag corresponding to the item content does not exist in all the user tags of the target user, and the user tag is used for adding the user tag of the target user.
Specifically, item content of the user behavior corresponding item is obtained, and whether a corresponding user tag exists or not is judged according to the item content. If yes, continuing to execute the step of acquiring the behavior coefficient corresponding to the behavior type of the user behavior. If the user identification information does not exist, corresponding target user labels are determined in all preset user labels according to the item content, the association relation between the item content and the target user labels is established, item websites of items corresponding to user behaviors are obtained, weights corresponding to the item websites are determined according to the corresponding relation between the preset item websites and the weights, and the weights corresponding to the item websites are used as the weights of the target user labels. Wherein the transaction website determines the weight and the transaction content determines the target user tag. The transaction web address may be an internet page address where the transaction is located. For example, the transaction web address may be a transaction guide page URL (Uniform Resource Locator ) link for a transaction on a government service network or a function page link on a mobile government service application, etc.
According to the embodiment, whether the corresponding user tag exists in the item content of the item corresponding to the user behavior is judged, the target user tag is newly established for the target user under the condition that the corresponding user tag does not exist, and the weight of the target user tag is determined according to the item website of the item, so that the user tag of the user is updated in real time, a more complete user portrait is constructed, the characteristics of the user are more obvious, and the item more suitable for the user can be recommended for the user.
Optionally, determining each reference item corresponding to the tag attribute of the user tag includes:
If the user tag is a comprehensive tag, determining each reference item corresponding to other user tags with the same attribute as the user tag;
If the user tag is an associated tag, determining each reference item associated with the user tag.
The integrated tag may be determined based on the integrated attribute of the plurality of user tags, and may be understood as a user tag of a higher layer than a general tag. If the user tag is a comprehensive tag, it may be determined that there is another user tag having the same attribute as the user tag, and each item corresponding to the other user tag is acquired as a reference item. For example, the "network sponsor" of the user tag belongs to a comprehensive tag of a higher layer, and since the social security card tag and the resident certificate tag are handled on the network and have the same attribute as the "network sponsor", the social security card-related item corresponding to the social security card tag and the resident certificate-related item corresponding to the resident certificate tag can be used as reference items.
The association tag may be determined based on an association relationship with the user tag and other user tags. If the user tag is associated with the tag, it may be determined that other user tags having an association relationship with the user tag exist, and each item corresponding to the other user tags may be acquired as a reference item. For example, the user tag is a social security card, and the social security card has an association relationship with the resident certificate, and the preset association relationship is that the resident certificate is a subsequent association item for the social security card, so that the resident certificate item can be used as a reference item. The preset association relationship can be determined by storing a piece of knowledge in advance, wherein the knowledge is used for determining the association relationship between the user labels.
According to the method and the device for determining the reference items, the label attributes of the user labels are analyzed, different modes for determining the reference items are selected according to different label attributes, accuracy of determining the reference items can be improved, and a stable foundation is provided for recommending items possibly of interest to the user from the reference items. And all the reference items possibly interested by the target user are determined based on the user tag of the target user, so that the recall rate is higher, and the method can adapt to the cold start semantic environment.
Step S220, extracting item features of items to which the user labels belong, generating target feature vectors, extracting item features of each reference item, generating reference feature vectors, and calculating similarity between the target feature vectors and the reference feature vectors.
Wherein the transaction characteristics may be determined based on the transaction tag system. The transaction tag system may include transaction type, transaction area, transaction condition, usage scope, and transaction mode, among others.
Specifically, item features of items to which the user tag belongs and item features of reference items are extracted respectively, corresponding target feature vectors and reference feature vectors are generated, and similarity between the target feature vectors and the reference feature vectors is calculated to determine similarity between the items to which the user tag belongs and the reference items.
Step S230, obtaining a weight corresponding to the user tag, and calculating a first matching degree between the target user and each reference item according to the weight and the similarity.
Illustratively, as shown in FIG. 5, steps S220 through S230 may be understood as content-based recalls in FIG. 5.
Step S240, obtaining a user attribute of the target user, determining a similar user with similarity to the target user not lower than a preset threshold based on the user attribute, and obtaining a user tag of the similar user.
Wherein the user attribute may be determined based on a user tagging hierarchy. The user tagging system may include user tags, user groups, user portraits, user groups, portrayal information, and the like. The user attributes may include age, network office frequency, whether to have local household alike.
Step S250, determining target labels corresponding to the reference items in all user labels of the similar users according to the corresponding relation between the reference items and the user labels of the similar users, acquiring weights of the target labels, and generating a behavior matrix based on the weights of the target users and the similar users corresponding to the target labels respectively.
Specifically, since the similar users may have user tags corresponding to each reference item, target tags corresponding to each reference item may be determined from all user tags of the similar users according to the correspondence between each reference item and the user tag of the similar user, weights of the target users and the similar users for each target tag may be obtained, and a behavior matrix may be generated.
Optionally, generating the behavior matrix based on the weights of the target users and the similar users corresponding to the target tags respectively may include:
taking the target user, the similar user and each reference item as matrix dimensions, taking the weights of the target labels determined by the target user and the similar user for each reference item as matrix elements, and generating a behavior matrix based on the weights of the target user and the similar user corresponding to each target label respectively.
For example, when the behavior matrix of the target user, the similar user and each reference item is established, the target user and the similar user may be taken as row dimensions, each reference item is taken as column dimensions, and weights of the target labels determined by the target user and the similar user for each reference item are respectively filled into corresponding positions of the matrix to form the behavior matrix. There is some sparsity in the behavior matrix because the target user and similar users have no corresponding weight on the individual references, thus resulting in the absence of corresponding matrix elements in the behavior matrix. And the corresponding missing matrix elements can be supplemented by a collaborative filtering method, so that the corresponding matrix elements exist for all the reference matters of the target user.
Step S260, calculating a second matching degree between the target user and each reference item based on the behavior matrix and the collaborative filtering method.
Illustratively, as shown in fig. 5, steps S240 to S260 may be understood as recalling through the collaborative filtering method in fig. 5.
Optionally, calculating a second matching degree between the target user and each reference item based on the behavior matrix and a collaborative filtering method includes:
calculating the user side matching degree, the item side matching degree and the model side matching degree of the target user and each reference item based on the behavior matrix and a collaborative filtering method;
And for each reference item, calculating a weighted sum of the user side matching degree, the item side matching degree and the model side matching degree according to a preset collaborative filtering matching degree weight, and taking the weighted sum as a second matching degree of the target user and the reference item.
The user-side matching degree can be obtained through a collaborative filtering method based on users. The item-side matching degree can be obtained by a collaborative filtering method based on items. The model-side matching degree can be obtained by a collaborative filtering method based on a model.
The collaborative filtering matching degree weight may be weights preset for three collaborative filtering methods, respectively, and is used for comprehensively determining the second matching degree of the target user and the reference item.
According to the embodiment, the matching degree of the target user and each reference item is calculated Through three collaborative filtering methods, the weighted sum of the matching degrees obtained Through the three collaborative filtering methods is calculated based on the preset collaborative filtering matching degree weight, and then the second matching degree of the target user and each reference item is determined. And the collaborative filtering method can adapt to a hot start semantic environment, so that the accuracy of recommending the transaction to the user is further improved.
It should be noted that, the steps S220 to S260 are not limited to the above-mentioned execution sequence, and the steps S240 to S260 may be executed first, and then the steps S220 to S230 may be executed. Or steps S220 to S230 and steps S240 to S260 may be performed in parallel. Or other order of execution, as the invention is not limited in detail.
Step S270 calculates the weighted sum of the first matching degree and the second matching degree according to the preset matching degree weight, sorts the reference items corresponding to the weighted sums, determines the target items according to the sorting result and recommends the target items to the target user.
Alternatively, in addition to determining that the target item is recommended to the target user according to the sorting result, the user attribute and the trending information of the target user can be matched, and the trending information matched with the user attribute can be recommended to the target user. For example, local hot transactions are pushed based on where the user is logged in to the IP (Internet Protocol Address ). The recommendation of the trending information can be that a specific trending information recommendation edition block on a recommendation page is displayed.
Illustratively, as shown in FIG. 5, the content representation and the user representation are determined by an optimization model. By matching the content representation with the user representation, a preliminary match can be made to the target user with the item. And then carrying out content-based, collaborative filtering method and/or hot-based recall on the preliminarily matched reference items through an optimization model, sequencing all the recalled items, and carrying out content display. Recall, among other things, is understood to trigger as many correct results as possible from the full set of information and return the results to the "ranking". According to the embodiment, the CTR can be predicted by executing the recommendation method provided by the embodiment of the invention to construct an optimization model. And the weight of the user tag is updated in real time based on the use time of the user tag in the optimization model and the influence of the user behavior on the corresponding weight of the user tag, so that the positive feedback and the negative feedback of the overall recommendation method are realized.
According to the embodiment of the invention, the recommendation method adapting to the cold start semantic environment and the recommendation method adapting to the hot start semantic environment are fused, the matching degree between the target user and the possibly interested reference item is calculated based on the user tag of the target user and the corresponding weight, and then the target item with higher matching degree is recommended to the target user, so that the handling item is accurately recommended to the user, and the prediction accuracy of the CTR is improved.
Example III
Fig. 6 is a schematic diagram of an intelligent recommendation system according to a third embodiment of the present invention, where the present embodiment is optimized based on the foregoing embodiments. As shown in FIG. 6, the architecture comprises a basic data warehouse, a matter modeling system, a user portrayal system, an intelligent recommendation center, an application management system and a recommendation rule management center. The basic data warehouse comprises a business system, a data source, data acquisition and ETL (Extract-Transform-Load). The application management system comprises application scenes, applications, enterprises and common people. The recommendation rule management center includes an optimization model.
Fig. 6a is a schematic diagram of a basic data warehouse according to a third embodiment of the present invention. As shown in fig. 6a, a large number of data sources are acquired through business systems such as a transaction system, a unified application system, a good and bad evaluation system, a consultation complaint system and the like based on digital informatization construction. The data sources may include event data sources such as event libraries, user ratings, etc., and may also include user data sources such as user base data, enterprise base data, and electronic certificates, etc. Two major categories of event data and user data can be obtained from a data source through data acquisition technology and/or ETL. And extracting and converting unified user basic data, government service item data, personal and enterprise office work data, electronic license data, logistics payment data, consultation complaint data and user evaluation data through an ETL tool to form a data warehouse, and establishing a user and item fact table and a maintenance table centering on the user and the item. Based on the fact table and the dimension table, various multidimensional models (Cube) can be constructed, including star models, snowflake models and constellation models. The transaction data and user data may be stored through a data repository. The data warehouse model may be built based on a star model. The fact table mainly contains two information: and a metric. The specific description information of the dimension is recorded in the dimension table, and the dimension attribute in the fact table is only one key associated to the dimension table, and no specific information is recorded. The metrics typically record corresponding values for the office, such as the number of offices, etc. The information in the dimension table can be layered, such as the year, month, day, city, county, etc. in the time dimension and the province, city, county, etc., and the layered information can meet the requirement that the metrics in the fact table can be aggregated at different granularity, such as the handling amount of 2019 matters, the handling amount from Guangzhou markets, etc.
It should also be noted that the information update frequency of the dimension table is not high or remains relatively stable, for example, a ten year time dimension that has been established does not need to be updated in a short period, and a region dimension; the data in the fact table is updated or increased continuously because the event is constantly occurring and the user is constantly generating pieces.
Fig. 6b is a schematic diagram of a transaction modeling system according to a third embodiment of the present invention. As shown in fig. 6b, for the event modeling system, the event data is processed by a modeling algorithm to construct features of the event. Fig. 6c is a diagram of a user portrait system according to a third embodiment of the present invention. As shown in FIG. 6c, for the user portrayal system, user data is processed through a behavior modeling algorithm to construct a user portrayal. User portraits, namely user information tagging, abstract the business full view of a user after collecting and analyzing the data of main information such as social attributes, living habits, transacting behaviors and the like of 'transacting people', quickly and accurately find user groups and find user demands, and find interesting information from a large amount of information to provide government service information recommendation. The user portrayal consists of a number of user labels, each defining an angle at which the user is observed, perceived or described. The user labels are more or less according to the development condition of the service, but the set of all the user labels is a whole, and the whole is called user portrait. The user portrayal is constructed to restore the user information so that the data originates from all user-related data. For classification of user related data, an important classification idea is introduced: classification of the closure. For example, the world is divided into two people, one is a person who can not be handled by the internet, and the other is a person who can not be handled by the internet; users are classified into three categories, young, middle-aged, and elderly; the life cycle of the office is divided into a throw-in period, a growth period, a maturation period, a decay period and the like, and all sub-classifications form the whole collection of category space. Such a classification approach facilitates subsequent continual enumeration and iterative replenishment of missing information dimensions. There is no need to worry that the architecture does not consider the integrity of each layer of classification, and the dimension omission leaves the hidden danger of expansibility. In addition, different classification modes may be reasonable according to application scenes and different service requirements, and the classification modes may be divided according to requirements. The core work of the user portrayal is to set user tags for users, one of the important purposes of setting user tags is to make it understandable and convenient for computer processing. Classification statistics can be made as follows: the number of users who like to do the port Australian pass is the number of men and women in the crowd who like to do the port Australian pass. Data mining work can also be done, such as calculating what people like to do a port-Australian pass typically go to port-Australian. And analyzing the distribution situation of age segments of people who like to carry out the port and Australian passes by using a clustering algorithm. The big data processing is not separated from the operation of the computer, and the user tag provides a convenient way, so that the computer can process the information related to the person in a programmed way, and even can 'understand' the person through an algorithm and a model. When the computer has such capability, the accuracy and the information acquisition efficiency can be further improved no matter in various application fields such as search engines, recommendation engines, advertisement delivery and the like.
Fig. 6d is a schematic diagram of an intelligent recommendation center according to a third embodiment of the present invention. As shown in fig. 6d, the initial recommendation results are filtered by determining the initial recommendation results through a manual configuration or recommendation algorithm, and ranking is performed on the filtered recommendation results. Or carrying out initial matching, recalling the initial matching, and sequencing the recalled results until a recommended result is obtained. The intelligent recommendation center also comprises an evaluation system which can evaluate the recommendation result from different angles.
Fig. 6e is a schematic diagram of an application management system according to a third embodiment of the present invention. As shown in fig. 6e, for example, a local person handles a transaction, first, the local person is judged whether or not to be the local person based on the presence or absence of the registered user by means of the license and the family directory information. And judging whether the user is related to residence according to behavior data such as search words, browse matters and the like in three months of the user, and pushing matters such as residence registration, residence permit and the like if the user is related to residence. Moreover, the usual things of the foreign are recommended, such as talent introduction, foreign complement of ID card, and transfer of accumulation of money from different places. And finally, pushing local hot business matters according to the IP location where the user logs in. In conclusion, the personalized recommended items of the user are formed after comprehensive ordering and duplication removal are performed according to the portrait activities, group handling and region handling.
Fig. 6f is a schematic diagram of an optimization model according to a third embodiment of the present invention. Fig. 6f is similar to fig. 5 and will not be described in detail herein.
The embodiment of the invention provides an intelligent recommendation system, which provides huge data support for an overall architecture through a basic data warehouse, respectively portrays the items and the users through an item modeling system and a user portrayal system so as to realize accurate positioning and analysis in the follow-up recommendation process, provides a recommendation method with higher accuracy for the users through an intelligent recommendation center and a recommendation rule management center, and realizes interaction to a user side through an application management system. The intelligent recommendation system provided by the embodiment of the invention can realize the omnibearing analysis and intelligent recommendation of the user, and improves the accuracy of recommending the transaction matters to the user.
Example IV
Fig. 7 is a schematic structural diagram of a recommending apparatus according to a fourth embodiment of the present invention. The device can be realized by software and/or hardware, can be generally integrated in recommendation equipment, and can be used for improving the accuracy of recommending the transacting items to the user by executing a recommendation method. As shown in fig. 7, the apparatus includes:
a tag obtaining module 310, configured to obtain a user tag of a target user, and determine each reference item corresponding to a tag attribute of the user tag;
A first matching degree calculating module 320, configured to calculate a first matching degree between the target user and each reference item according to the weight corresponding to the user tag and the similarity between the item to which the user tag belongs and each reference item;
A second matching degree calculating module 330, configured to calculate a second matching degree between the target user and each reference item by using a collaborative filtering method based on a correspondence between similar users of the target user and each reference item;
The result recommending module 340 is configured to calculate a weighted sum of the first matching degree and the second matching degree according to a preset matching degree weight, sort the reference items corresponding to the weighted sums, determine the target item according to the sorting result, and recommend the target item to the target user.
Optionally, the apparatus further comprises:
The tag determining module is used for determining the user tag corresponding to the user behavior when the user behavior of the target user is detected before the user tag of the target user is acquired;
And the weight updating module is used for acquiring the behavior coefficient corresponding to the behavior type of the user behavior and updating the weight corresponding to the user tag according to the behavior coefficient.
Optionally, the tag determining module is specifically configured to:
Acquiring item content of the item corresponding to the user behavior, and judging whether a user tag corresponding to the item content exists or not;
If yes, executing the step of acquiring the behavior coefficient corresponding to the behavior type of the user behavior;
If not, establishing an association relation between the item content and the target user tag, acquiring an item website of the item corresponding to the user behavior, and determining that the weight of the target user tag is the weight corresponding to the item website according to the preset association relation between the item website and the weight.
Optionally, the tag obtaining module 310 is specifically configured to:
If the user tag is a comprehensive tag, determining each reference item corresponding to other user tags with the same attribute as the user tag;
If the user tag is an associated tag, determining each reference item associated with the user tag.
Optionally, the first matching degree calculating module 320 is specifically configured to:
Extracting item features of items to which the user tag belongs to generate a target feature vector;
extracting item features of each reference item to generate a reference feature vector;
calculating the similarity between the target feature vector and the reference feature vector;
and acquiring the weight corresponding to the user tag, and calculating the first matching degree of the target user and each reference item according to the weight and the similarity.
Optionally, the second matching degree calculating module 330 is specifically configured to:
Acquiring user attributes of the target user, determining similar users with similarity not lower than a preset threshold value based on the user attributes, and acquiring user tags of the similar users;
Determining target labels corresponding to all the reference items in all the user labels of the similar users according to the corresponding relation between the reference items and the user labels of the similar users, acquiring the weight of the target labels, and generating a behavior matrix based on the weights of the target users and the similar users corresponding to the target labels respectively;
and calculating a second matching degree between the target user and each reference item based on the behavior matrix and a collaborative filtering method.
Optionally, the second matching degree calculating module 330 is specifically configured to:
taking the target user, the similar user and each reference item as matrix dimensions, taking the weights of the target labels determined by the target user and the similar user for each reference item as matrix elements, and generating a behavior matrix based on the weights of the target user and the similar user corresponding to each target label respectively.
Optionally, the second matching degree calculating module 330 is specifically configured to:
calculating the user side matching degree, the item side matching degree and the model side matching degree of the target user and each reference item based on the behavior matrix and a collaborative filtering method;
And for each reference item, calculating a weighted sum of the user side matching degree, the item side matching degree and the model side matching degree according to a preset collaborative filtering matching degree weight, and taking the weighted sum as a second matching degree of the target user and the reference item.
The recommending device provided by the embodiment of the invention can execute the recommending method provided by any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the executing method.
Example five
Fig. 8 is a schematic structural diagram of a recommendation device according to a fifth embodiment of the present invention, and as shown in fig. 8, the recommendation device includes a processor 400, a memory 410, an input device 420 and an output device 430; the number of processors 400 in the recommendation device may be one or more, one processor 400 being taken as an example in fig. 8; the processor 400, memory 410, input means 420 and output means 430 in the recommendation device may be connected by a bus or other means, in fig. 8 by way of example.
The memory 410 is used as a computer readable storage medium, and may be used to store a software program, a computer executable program, and modules, such as program instructions and/or modules corresponding to the recommendation method in the embodiment of the present invention (for example, the tag obtaining module 310, the first matching degree calculating module 320, the second matching degree calculating module 330, and the result recommending module 340 in the recommending device). The processor 400 executes various functional applications of the recommendation device and data processing, i.e., implements the recommendation method described above, by running software programs, instructions, and modules stored in the memory 410.
Memory 410 may include primarily a program storage area and a data storage area, wherein the program storage area may store an operating system, at least one application program required for functionality; the storage data area may store data created according to the use of the terminal, etc. In addition, memory 410 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid-state storage device. In some examples, memory 410 may further include memory remotely located with respect to processor 400, which may be connected to the recommendation device via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The input means 420 may be used to receive entered numeric or character information and to generate key signal inputs related to user settings and function control of the recommendation device. The output 430 may include a display device such as a display screen.
Example six
A sixth embodiment of the present invention also provides a storage medium containing computer-executable instructions, which when executed by a computer processor, are for performing a recommendation method, the method comprising:
acquiring a user tag of a target user, and determining each reference item corresponding to the tag attribute of the user tag;
Calculating a first matching degree of the target user and each reference item according to the weight corresponding to the user tag and the similarity between the item to which the user tag belongs and each reference item;
Calculating a second matching degree between the target user and each reference item through a collaborative filtering method based on the corresponding relation between the similar user of the target user and each reference item;
And calculating weighted sums of the first matching degree and the second matching degree according to preset matching degree weights, sequencing reference matters corresponding to the weighted sums, determining target matters according to sequencing results, and recommending the target matters to a target user.
Of course, the storage medium containing the computer executable instructions provided in the embodiments of the present invention is not limited to the method operations described above, and may also perform the related operations in the recommended method provided in any embodiment of the present invention.
From the above description of embodiments, it will be clear to a person skilled in the art that the present invention may be implemented by means of software and necessary general purpose hardware, but of course also by means of hardware, although in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as a floppy disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a FLASH Memory (FLASH), a hard disk, or an optical disk of a computer, etc., and include several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method according to the embodiments of the present invention.
It should be noted that, in the embodiment of the recommendation device, each unit and module included are only divided according to the functional logic, but not limited to the above division, so long as the corresponding functions can be implemented; in addition, the specific names of the functional units are also only for distinguishing from each other, and are not used to limit the protection scope of the present invention.
Note that the above is only a preferred embodiment of the present invention and the technical principle applied. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, while the invention has been described in connection with the above embodiments, the invention is not limited to the embodiments, but may be embodied in many other equivalent forms without departing from the spirit or scope of the invention, which is set forth in the following claims.

Claims (9)

1. A recommendation method, comprising:
acquiring a user tag of a target user, and determining each reference item corresponding to the tag attribute of the user tag;
Calculating a first matching degree of the target user and each reference item according to the weight corresponding to the user tag and the similarity between the item to which the user tag belongs and each reference item;
Calculating a second matching degree between the target user and each reference item through a collaborative filtering method based on the corresponding relation between the similar user of the target user and each reference item;
Calculating weighted sums of the first matching degree and the second matching degree according to preset matching degree weights, sequencing reference matters corresponding to the weighted sums, determining target matters according to sequencing results and recommending the target matters to a target user;
Wherein the same user tag corresponds to at least one tag attribute; the same tag attribute corresponds to at least one user tag; each user tag belonging to at least one item;
The determining each reference item corresponding to the label attribute of the user label comprises the following steps: and acquiring the user tag of the target user, determining the tag attribute of the user tag, determining all the user tags under the tag attribute according to the tag attribute, and further determining each reference item to which all the user tags belong.
2. The method of claim 1, further comprising, prior to obtaining the user tag of the target user:
When the user behavior of a target user is detected, determining a user tag corresponding to the user behavior;
and acquiring a behavior coefficient corresponding to the behavior type of the user behavior, and updating the weight corresponding to the user tag according to the behavior coefficient.
3. The method of claim 2, wherein determining the user tag corresponding to the user behavior comprises:
Acquiring item content of the item corresponding to the user behavior, and judging whether a user tag corresponding to the item content exists or not;
If yes, executing the step of acquiring the behavior coefficient corresponding to the behavior type of the user behavior;
If not, establishing an association relation between the item content and the target user tag, acquiring an item website of the item corresponding to the user behavior, and determining that the weight of the target user tag is the weight corresponding to the item website according to the preset association relation between the item website and the weight.
4. The method of claim 1, wherein determining the respective references to the tag attributes of the user tag comprises:
If the user tag is a comprehensive tag, determining each reference item corresponding to other user tags with the same attribute as the user tag;
If the user tag is an associated tag, determining each reference item associated with the user tag.
5. The method according to claim 1, wherein calculating the first matching degree between the target user and each reference item according to the weight corresponding to the user tag and the similarity between the item to which the user tag belongs and each reference item comprises:
Extracting item features of items to which the user tag belongs to generate a target feature vector;
extracting item features of each reference item to generate a reference feature vector;
calculating the similarity between the target feature vector and the reference feature vector;
and acquiring the weight corresponding to the user tag, and calculating the first matching degree of the target user and each reference item according to the weight and the similarity.
6. The method of claim 1, wherein calculating a second degree of matching of the target user with each of the references by a collaborative filtering method based on correspondence between similar users of the target user and each of the references, comprises:
Acquiring user attributes of the target user, determining similar users with similarity not lower than a preset threshold value based on the user attributes, and acquiring user tags of the similar users;
Determining target labels corresponding to all the reference items in all the user labels of the similar users according to the corresponding relation between the reference items and the user labels of the similar users, acquiring the weight of the target labels, and generating a behavior matrix based on the weights of the target users and the similar users corresponding to the target labels respectively;
Calculating a second matching degree between the target user and each reference item based on the behavior matrix and a collaborative filtering method;
The generating a behavior matrix based on weights of the target users and the similar users corresponding to the target labels respectively includes:
Taking the target user, the similar user and each reference item as matrix dimensions, taking the weights of the target labels determined by the target user and the similar user aiming at each reference item as matrix elements, and generating a behavior matrix based on the weights of the target user and the similar user corresponding to each target label respectively;
The calculating, based on the behavior matrix and the collaborative filtering method, a second matching degree between the target user and each reference item includes:
calculating the user side matching degree, the item side matching degree and the model side matching degree of the target user and each reference item based on the behavior matrix and a collaborative filtering method;
And for each reference item, calculating a weighted sum of the user side matching degree, the item side matching degree and the model side matching degree according to a preset collaborative filtering matching degree weight, and taking the weighted sum as a second matching degree of the target user and the reference item.
7. A recommendation device, comprising:
the tag acquisition module is used for acquiring a user tag of a target user and determining each reference item corresponding to the tag attribute of the user tag;
the first matching degree calculation module is used for calculating the first matching degree of the target user and each reference item according to the weight corresponding to the user tag and the similarity of the item to which the user tag belongs and each reference item;
the second matching degree calculation module is used for calculating the second matching degree of the target user and each reference item through a collaborative filtering method based on the corresponding relation between the similar users of the target user and each reference item;
The result recommending module is used for calculating the weighted sum of the first matching degree and the second matching degree according to the preset matching degree weight, sequencing the reference items corresponding to the weighted sums, determining target items according to the sequencing result and recommending the target items to a target user;
Wherein the same user tag corresponds to at least one tag attribute; the same tag attribute corresponds to at least one user tag; each user tag belonging to at least one item;
the tag acquisition module is specifically configured to: and acquiring the user tag of the target user, determining the tag attribute of the user tag, determining all the user tags under the tag attribute according to the tag attribute, and further determining each reference item to which all the user tags belong.
8. A recommendation device, characterized in that the recommendation device comprises:
One or more processors;
A memory for storing one or more programs,
When executed by the one or more processors, causes the one or more processors to implement the recommendation method of any one of claims 1-6.
9. A storage medium containing computer executable instructions which, when executed by a computer processor, are for performing the recommendation method according to any one of claims 1-6.
CN202011633927.0A 2020-12-31 2020-12-31 Recommendation method, recommendation device, recommendation equipment and storage medium Active CN112632405B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011633927.0A CN112632405B (en) 2020-12-31 2020-12-31 Recommendation method, recommendation device, recommendation equipment and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011633927.0A CN112632405B (en) 2020-12-31 2020-12-31 Recommendation method, recommendation device, recommendation equipment and storage medium

Publications (2)

Publication Number Publication Date
CN112632405A CN112632405A (en) 2021-04-09
CN112632405B true CN112632405B (en) 2024-05-10

Family

ID=75289896

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011633927.0A Active CN112632405B (en) 2020-12-31 2020-12-31 Recommendation method, recommendation device, recommendation equipment and storage medium

Country Status (1)

Country Link
CN (1) CN112632405B (en)

Families Citing this family (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113158056A (en) * 2021-04-27 2021-07-23 口碑(上海)信息技术有限公司 Recommendation language generation method and device
CN113326203B (en) * 2021-06-22 2022-08-12 深圳前海微众银行股份有限公司 Information recommendation method, equipment and storage medium
CN113672800B (en) * 2021-07-12 2024-09-06 浪潮软件股份有限公司 Item recommendation method and storage medium for real-name authentication of natural person user
CN114357184A (en) * 2021-12-21 2022-04-15 讯飞智元信息科技有限公司 Item recommendation method and related device, electronic equipment and storage medium
CN114239984A (en) * 2021-12-22 2022-03-25 中国建设银行股份有限公司 Service item optimization method, service item optimization device, electronic device, service item optimization medium, and program product
CN114969566B (en) * 2022-06-27 2023-03-24 中国测绘科学研究院 Distance-measuring government affair service item collaborative filtering recommendation method
CN116484109B (en) * 2023-06-21 2023-09-01 九一金融信息服务(北京)有限公司 Customer portrait analysis system and method based on artificial intelligence
CN116975396A (en) * 2023-09-25 2023-10-31 北京市大数据中心 Intelligent recommendation method, system, equipment and storage medium for government service

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106952130A (en) * 2017-02-27 2017-07-14 华南理工大学 Common user item based on collaborative filtering recommends method
CN108121737A (en) * 2016-11-29 2018-06-05 阿里巴巴集团控股有限公司 A kind of generation method, the device and system of business object attribute-bit
CN109033101A (en) * 2017-06-08 2018-12-18 华为软件技术有限公司 Label recommendation method and device
CN109543111A (en) * 2018-11-28 2019-03-29 广州虎牙信息科技有限公司 Recommendation information screening technique, device, storage medium and server
CN110223186A (en) * 2018-05-09 2019-09-10 腾讯科技(深圳)有限公司 User's similarity determines method and information recommendation method
WO2020081594A1 (en) * 2018-10-15 2020-04-23 Ask Sydney, Llc Iterative, multi-user selection and weighting recommendation engine
CN111143443A (en) * 2019-11-29 2020-05-12 数字广东网络建设有限公司 Method, device, system, terminal and storage medium for displaying government affair information
CN111859116A (en) * 2020-06-19 2020-10-30 湖南师范大学 Personalized course recommendation method and device based on big data and lifetime planning

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8019766B2 (en) * 2007-03-30 2011-09-13 Amazon Technologies, Inc. Processes for calculating item distances and performing item clustering
US10762136B2 (en) * 2017-09-15 2020-09-01 Telefonaktiebolaget Lm Ericsson (Publ) Tag-based, user directed media recommendations

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108121737A (en) * 2016-11-29 2018-06-05 阿里巴巴集团控股有限公司 A kind of generation method, the device and system of business object attribute-bit
CN106952130A (en) * 2017-02-27 2017-07-14 华南理工大学 Common user item based on collaborative filtering recommends method
CN109033101A (en) * 2017-06-08 2018-12-18 华为软件技术有限公司 Label recommendation method and device
CN110223186A (en) * 2018-05-09 2019-09-10 腾讯科技(深圳)有限公司 User's similarity determines method and information recommendation method
WO2020081594A1 (en) * 2018-10-15 2020-04-23 Ask Sydney, Llc Iterative, multi-user selection and weighting recommendation engine
CN109543111A (en) * 2018-11-28 2019-03-29 广州虎牙信息科技有限公司 Recommendation information screening technique, device, storage medium and server
CN111143443A (en) * 2019-11-29 2020-05-12 数字广东网络建设有限公司 Method, device, system, terminal and storage medium for displaying government affair information
CN111859116A (en) * 2020-06-19 2020-10-30 湖南师范大学 Personalized course recommendation method and device based on big data and lifetime planning

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
综合社区与关联序列挖掘的电子政务推荐算法;黄亚坤 等;计算机应用;正文1-7页 *
跨源多维政务数据共享与服务创新研究;肖炯恩;社会科学Ⅰ辑;全文 *

Also Published As

Publication number Publication date
CN112632405A (en) 2021-04-09

Similar Documents

Publication Publication Date Title
CN112632405B (en) Recommendation method, recommendation device, recommendation equipment and storage medium
CN110704572B (en) Suspected illegal fundraising risk early warning method, device, equipment and storage medium
CN111831636B (en) Data processing method, device, computer system and readable storage medium
CN106156127B (en) Method and device for selecting data content to push to terminal
CN104361111B (en) A kind of archives are compiled and grind method automatically
CN105447184B (en) Information extraction method and device
CN109918563B (en) Book recommendation method based on public data
CN107908606A (en) Method and system based on different aforementioned sources automatic report generation
CN106844407B (en) Tag network generation method and system based on data set correlation
CN103310003A (en) Method and system for predicting click rate of new advertisement based on click log
Irudeen et al. Big data solution for Sri Lankan development: A case study from travel and tourism
CN112182246A (en) Method, system, medium, and application for creating an enterprise representation through big data analysis
CN112269805A (en) Data processing method, device, equipment and medium
CN112631889B (en) Portrayal method, device, equipment and readable storage medium for application system
CN114371946B (en) Information push method and information push server based on cloud computing and big data
CN111191111A (en) Content recommendation method, device and storage medium
CN110880124A (en) Conversion rate evaluation method and device
CN113420018A (en) User behavior data analysis method, device, equipment and storage medium
CN116384889A (en) Intelligent analysis method for information big data based on natural language processing technology
CN111310032A (en) Resource recommendation method and device, computer equipment and readable storage medium
CN105786810B (en) The method for building up and device of classification mapping relations
US20150142782A1 (en) Method for associating metadata with images
CN115982429B (en) Knowledge management method and system based on flow control
TWM573493U (en) System for predicting conversion probability by visitors' browsing paths
CN111460300B (en) Network content pushing method, device and storage medium

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