CN116843394B - AI-based advertisement pushing method, device, equipment and storage medium - Google Patents
AI-based advertisement pushing method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN116843394B CN116843394B CN202311124241.2A CN202311124241A CN116843394B CN 116843394 B CN116843394 B CN 116843394B CN 202311124241 A CN202311124241 A CN 202311124241A CN 116843394 B CN116843394 B CN 116843394B
- Authority
- CN
- China
- Prior art keywords
- user
- advertisement
- purchased
- product
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 42
- 238000004458 analytical method Methods 0.000 claims abstract description 12
- 238000002372 labelling Methods 0.000 claims abstract description 4
- 238000012549 training Methods 0.000 claims description 12
- 239000013598 vector Substances 0.000 claims description 12
- 238000005065 mining Methods 0.000 claims description 8
- 238000005259 measurement Methods 0.000 claims description 4
- 238000012098 association analyses Methods 0.000 claims description 3
- 230000010354 integration Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000012163 sequencing technique Methods 0.000 claims 1
- 238000011156 evaluation Methods 0.000 abstract 1
- 241000220225 Malus Species 0.000 description 4
- 238000013459 approach Methods 0.000 description 2
- 230000006399 behavior Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 241000234295 Musa Species 0.000 description 1
- 240000008790 Musa x paradisiaca Species 0.000 description 1
- 235000018290 Musa x paradisiaca Nutrition 0.000 description 1
- 235000021016 apples Nutrition 0.000 description 1
- 235000021015 bananas Nutrition 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003252 repetitive effect Effects 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
-
- 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
- G06Q30/0253—During e-commerce, i.e. online transactions
-
- 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
- G06Q30/0254—Targeted advertisements based on statistics
-
- 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
Abstract
The invention discloses an AI-based advertisement pushing method, an AI-based advertisement pushing device, AI-based advertisement pushing equipment and an AI-based advertisement storage medium, and relates to the field of advertisement pushing, wherein the AI-based advertisement pushing method comprises the steps of collecting user purchase records and labeling field information; performing inter-domain relevance analysis by using relevance rule learning to generate a relevance ranking list; when the product advertisement in the non-purchased domain is required to be pushed for a specific user, calling a relevancy list, and identifying the reference domain of the purchase record; each product is provided with a label, such as brands, price interval properties, user evaluation and the like, so that user preferences can be captured accurately; according to the purchase history of the user in the reference field, the system gathers tag preferences, searches for products similar to the preferences and pushes advertisements. And setting scores for association rules by combining the support degree, the confidence degree and the promotion degree, and optimizing the pushing effect. The invention realizes intelligent and accurate advertisement pushing and improves advertisement effectiveness and user satisfaction.
Description
Technical Field
The present invention relates to the field of advertisement pushing, and in particular, to an AI-based advertisement pushing method, apparatus, device, and storage medium.
Background
With the rapid development of electronic commerce and network advertising, providing personalized advertising content for users has become a core target for advertisement pushing. Currently, most advertisement pushing systems rely primarily on the user's purchase history or browsing records to push. This approach does to some extent meet the needs of the user, as it pushes products that are similar to their historical behavior according to the user's past behavior habits.
This approach has significant limitations. Existing advertisement delivery systems tend to face dilemma when users wish to enter a new domain, purchasing products that they have never contacted before. Because these systems rely too much on the user's historical data, they lack enough data to make accurate pushes for the user when they are involved in unknown or new areas. As a result, the user may receive advertisements that do not match their actual needs or that are not related to their buying intent, resulting in a significant discount on the effectiveness of the advertisement.
Depending solely on the browsing and purchase history of the user, the push content may also be too single and repetitive to provide the user with a wider and varied selection of products. Therefore, how to perform more accurate and effective advertisement pushing when facing the new field or purchase intention of the user has become a problem to be solved in the development of advertisement technology.
Disclosure of Invention
The invention provides an AI-based advertisement pushing method, an AI-based advertisement pushing device, AI-based advertisement pushing equipment and an AI-based advertisement pushing storage medium, so as to solve the problems in the background art.
In order to achieve the above purpose, the invention adopts the following technical scheme:
an AI-based advertisement pushing method comprises the following steps:
s1, acquiring purchase records of a user used as training data, wherein each record marks corresponding field information;
s2, integrating purchase records of all users in the S1, carrying out relevance analysis among different fields by using a relevance rule learning algorithm, and generating a relevance ranking list for each field, wherein other fields are ranked in the relevance ranking list according to the relevance of the relevance ranking list to the field;
s3, when the product advertisement of the non-purchased domain is required to be pushed for a specific user, calling out a relevancy sorting list of the non-purchased domain, and identifying one or more domains with purchase records in the relevancy sorting list as reference domains;
s4: for each reference field, pushing product advertisements of the non-purchased field to a specific user based on the label specifically comprises the following steps:
s4.1, distributing labels for each product;
s4.2, summarizing the product label preference of the specific user according to the products purchased by the specific user in the reference field;
s4.3, searching products similar to the tag preference in the non-purchased field to push advertisements, wherein the method specifically comprises the following steps:
converting the label of each product in the reference field into a label vector by using a coding algorithm and calculating the average value of the label vectors to obtain an overall label preference vector;
and calculating the similarity between each product in the non-purchased field and the overall tag preference vector by using a similarity measurement algorithm, and selecting the priority of advertisement pushing according to the similarity sorting.
In some embodiments, the pushed advertisement is a video advertisement;
in the S1, the training data also comprises the stay time of the user on the historical video advertisement; if the stay time of the user on the historical video advertisement exceeds a preset threshold, the product corresponding to the video advertisement is used as a purchase record of the user to be added into the relevance analysis in S2. The stay of the historical video advertisement can be judged by whether the user skips the advertisement or not, and can also be judged by eye tracking of the user, which is gradually realized in the VR field, such as apple vision pro recently released by apple company;
and S4, adding the products corresponding to the video advertisements with the stay time exceeding the preset threshold value into the products purchased by the specific user in the reference field according to the stay time of the specific user on the historical video advertisements.
The two preset thresholds may be set separately or may be set the same. For example, each may be set to 10 seconds.
In some embodiments, S1 further comprises: data cleansing is performed on the collected plurality of user purchase records, including removing duplicate purchase records, processing missing data, and converting unstructured data.
In some embodiments, the tag in S4 includes at least any one or more of: brands, price interval properties, user rating keywords.
In some embodiments, S2 specifically includes:
s2.1, converting the purchase record of each user into a transaction data set, wherein the transaction data set comprises a plurality of transactions, and each transaction comprises all fields purchased by the user in one shopping;
s2.2, a frequent item set is found out from the transaction data set by adopting a frequent item set mining method;
s2.3, for each frequent item set, calculating the support, confidence and lifting degree of the frequent item set;
s2.4, generating association rules from one domain to another domain according to the frequent item set; each association rule has a corresponding support degree, confidence degree and promotion degree;
s2.5, setting an association rule score for each association rule, wherein the association rule score is a weighted sum of support degree, confidence degree and lifting degree;
and S2.6, calculating the association rule score from one domain to other domains, and sorting from high to low according to the association rule score to obtain an association degree sorting list.
In some embodiments, the frequent item set mining method is an Apriori algorithm.
In some embodiments, advertisement pushing includes a passive pushing scenario and an automatic pushing scenario:
the passive pushing scene is as follows: when a particular user searches for products in an un-purchased area;
the automatic pushing scene is as follows: sorting the purchased product areas according to the product purchase quantity of each of the product areas purchased by the specific user; and calling a relevance ranking list for ranking the previous one or more purchased product fields, and querying the previous one or more non-purchased product fields in the relevance ranking list as advertisement pushing fields.
The invention also discloses an AI-based advertisement pushing device, which comprises:
and a data acquisition module: the method comprises the steps of acquiring purchase records of a user used as training data, and labeling corresponding field information for each record;
and the data integration and association analysis module: the method comprises the steps of integrating purchase records of all users acquired by a data acquisition module, carrying out relevance analysis among different fields by using relevance rule learning, and generating a relevance ranking list;
an advertisement push selection module: when a product advertisement of an un-purchased domain is required to be pushed for a specific user, calling a relevancy ranking list, and identifying one or more previous domains with purchase records in the relevancy ranking list as reference domains;
and the pushing module is used for: user preferences are summarized based on product tags for each reference domain, and products similar to the tag preferences are searched in the non-purchased domain for advertisement pushing.
The invention also discloses a computer readable storage medium, which stores thereon:
executing an instruction: when the execution instruction is executed by a computer, the computer is caused to implement the advertisement pushing method;
domain information database: storing user purchase records of various fields;
association ranking list library: storing a relevance ranking list of each field;
product label library: storing the related label information for each product;
user tag preference database: the tag preferences of the user are stored according to their purchase records in different fields.
The invention also discloses an AI-based advertisement pushing device, which comprises:
a processor;
the storage medium: the method comprises an execution instruction, a domain information database, a relevancy ranking list database, a product tag database and a user tag preference database;
the apparatus implements the advertisement pushing method when the processor executes the execution instructions on the storage medium.
Compared with the prior art, the method has the advantages that the association degree between different fields can be identified and calculated by deeply analyzing the purchase records of a large number of users and the field association thereof and combining the association rule learning algorithm, so that the associated purchase fields are confirmed according to the existing purchase records of the users, and when the advertisement pushing is required to be carried out on the products in the non-purchased fields, the label property of the purchased products in the most associated fields can be searched according to the association degree sorting list, and the products which are required to be searched for pushing in the non-purchased fields are confirmed, so that the pushing accuracy in the non-purchased product fields is greatly improved.
Drawings
FIG. 1 is a schematic flow chart of the method of the present invention.
Detailed Description
The following describes specific embodiments of the present invention with reference to the drawings.
A general flow chart of the invention is shown in fig. 1.
The invention discloses an AI-based advertisement pushing method, which comprises the following steps:
s1, acquiring purchase records of a user used as training data, wherein each record marks corresponding field information; the user or purchase records herein may be all user or purchase records on a certain e-commerce platform;
s2, integrating purchase records of all users, carrying out relevance analysis among different fields by using an relevance rule learning algorithm, and generating a relevance ranking list for each field, wherein other fields are ranked in the relevance ranking list according to the relevance of the relevance ranking list to the field;
s3, when the product advertisement of the non-purchased domain is required to be pushed for a specific user, calling out a relevancy sorting list of the non-purchased domain, and identifying one or more domains with purchase records in the relevancy sorting list as reference domains;
s4: for each reference field, pushing product advertisements of the non-purchased field to the user based on the label specifically comprises the following steps:
s4.1, distributing labels for each product;
s4.2, summarizing label preference of products purchased by a user in the reference field;
s4.3, searching products similar to the tag preference in the non-purchased field to push advertisements, wherein the method specifically comprises the following steps:
converting the label of each product in the reference field into a label vector using a coding algorithm (such as One-hot coding or TF-IDF coding) and calculating the average value of the label vectors to obtain an overall label preference vector;
and calculating the similarity between each product in the non-purchased field and the overall tag preference vector by using a similarity measurement algorithm (such as cosine similarity measurement), and selecting the priority of advertisement pushing according to the similarity ordering.
In some embodiments, S1 further comprises: data cleansing is performed on the collected plurality of user purchase records, including removing duplicate purchase records, processing missing data, and converting unstructured data.
In some embodiments, the labels in S4 include, but are not limited to, brands, price range properties, and user rating keywords.
In some embodiments, S2 specifically includes:
s2.1, converting the purchase record of each user into a transaction data set, wherein the transaction data set comprises a plurality of transactions, and each transaction comprises all fields purchased by the user in one shopping;
s2.2, adopting a frequent item set mining method to find out frequent item sets from the transaction data sets (the frequent item sets represent field combinations which frequently occur together in the data sets, and the support degree of the frequent item sets is greater than or equal to a preset minimum support degree threshold value);
s2.3, for each frequent item set, calculating the support, confidence and lifting degree of the frequent item set;
s2.4, generating association rules from one domain to another domain according to the frequent item set; each association rule has a corresponding support degree, confidence degree and promotion degree;
s2.5, setting an association rule score for each association rule, wherein the association rule score is a weighted sum of support degree, confidence degree and lifting degree;
and S2.6, calculating the association rule score from one domain to other domains, and sorting from high to low according to the association rule score to obtain an association degree sorting list.
A transaction is typically a record in a database. In the context of shopping basket analysis, a transaction may be considered a shopping basket or a customer's record of purchases. For example, assuming a customer purchases apples, bananas, and oranges in a single purchase, the set of three items is a matter of business. A transaction is an instance of a collection of items. A term set is a collection of one or more terms. The term set does not refer specifically to any particular transaction, but rather is a generic set. For example, { apple, banana } is a set of items that can appear in multiple transactions. In association rule mining, item sets that frequently occur in transactions, i.e., frequent item sets, are typically sought.
The support represents the frequency with which a certain set of items (item combination) appears in all transactions. It is defined as:
support (X) =transaction number contains item set (X)/total transaction number;
the confidence level indicates the probability of occurrence of the item set Y in the case where one item set X occurs. Confidence is defined as:
confidence (x→y) =support (xuy)/support (X);
here, x→y denotes an association rule, and x→y denotes the union of the term set X and the term set Y.
The degree of promotion is used to evaluate whether two items in an association rule are truly related or whether they are only commonly occurring together due to their respective high degree of support.
The calculation formula of the lifting degree is:
degree of lift (x→y) =degree of confidence (x→y)/degree of support (Y);
if the degree of promotion is >1, then there is a positive correlation between item sets X and Y.
If lift = 1, it means that X and Y are independent without any association.
If the degree of lift is <1, it indicates that there is a negative correlation between X and Y.
The following embodiments detail how the relevancy ranking is calculated by a shopping mall's purchase record.
1. First assume that the following simplified version of the user purchase record:
transaction 1: user a purchased { book, notebook, bookmark };
transaction 2: user B purchased { book, bookmark };
transaction 3: user C purchased { book, notebook };
transaction 4: user D purchased { notebook, rubber };
2. find frequent item sets: setting the minimum support threshold to 50% (2/4), the frequent item sets are:
{ book };
{ notebook };
{ bookmark };
{ book, notebook };
{ book, bookmark };
3. calculating the support, the confidence and the promotion degree:
taking { book, notebook } as an example:
support degree: (number of times book and notebook occur together)/(total number of transactions) =2/4=50%;
confidence level: (number of occurrences of book and notebook)/(number of occurrences of book) =2/3≡66.67%;
degree of lifting: (support { book, notebook })/(support { book } ×support { notebook }) =0.5/(0.75×0.75) = 0.8888;
4. generating an association rule: the association rules from books to notebooks are as follows: book= > notebook. For this association rule, its support, confidence and promotion have been calculated in the previous step.
5. Association rule score: for simplicity, assuming that we give the same weights for support, confidence and promotion, all are 1, the association rule score is:
score = support + confidence + boost = 0.5 + 0.6667 + 0.8888 = 2.0555;
6. association ordered list: the score is calculated for all generated association rules, then for a particular domain (e.g., book), the association rule scores for the book and other domains may be calculated and ranked from high to low.
In some embodiments, the frequent item set mining method is an Apriori algorithm. The support, confidence and promotion are all parameters conventionally calculated in the Apriori algorithm.
The Apriori algorithm is a classical algorithm for association rule learning that exploits the nature of frequent item sets to reduce the amount of necessary computation.
In some embodiments, advertisement pushing includes a passive pushing scenario and an automatic pushing scenario:
the passive pushing scene is as follows: when a particular user searches for products in an un-purchased area;
the automatic pushing scene is as follows: sorting the purchased product areas according to the product purchase quantity of each of the product areas purchased by the specific user; and calling a relevance ranking list for ranking the previous one or more purchased product fields, and querying the previous one or more non-purchased product fields in the relevance ranking list as advertisement pushing fields.
In other embodiments, the pushed advertisement is a video advertisement; in S1, training data also comprises stay time of a user on a historical video advertisement; if the stay time of the user on the historical video advertisement exceeds a preset threshold, the product corresponding to the video advertisement is used as a purchase record of the user to be added into the relevance analysis in S2. The stay on the historical video advertisement can be judged by whether the user skips the advertisement or not, and can also be judged by eye tracking of the user, which is gradually realized in the VR field, such as the latest apple vision pro released by apple company.
In the embodiment, the stay time of the user on the historical video advertisement can be used as training data for training alone, and the data can be combined with actual purchase data for training; when recommending products or video advertisements for a specific user, products with stay time exceeding a threshold value can be added into the reference field of the specific user as purchase records (can be independently used as the purchase records or can be combined with actual purchase records) according to stay time on the historical video advertisements, and video advertisement pushing of corresponding products is carried out according to product labels in the reference field.
The invention also discloses an AI-based advertisement pushing device, which comprises:
and a data acquisition module: the method comprises the steps of acquiring purchase records of a large number of users, and labeling corresponding field information for each record;
and the data integration and association analysis module: the method comprises the steps of integrating purchase records of all users, carrying out relevance analysis among different fields by using relevance rule learning, and generating a relevance ranking list;
an advertisement push selection module: the method comprises the steps of calling a relevancy ranking list when a product advertisement of an un-purchased domain is required to be pushed for a specific user, and identifying one or more domains, which exist in the relevancy ranking list and are recorded by purchase, as reference domains;
and the pushing module is used for: for summarizing user preferences based on product tags for each reference domain and searching for products similar to the tag preferences in the non-purchased domain for advertisement pushing.
In some embodiments, the following specific hardware and software may be employed:
a Web API is built using the Python's flash or Django framework for receiving a user's purchase records and storing in a database such as PostgreSQL.
The data record may be attached with a domain label, for example: books, household articles, etc.
The user purchase records are analyzed to find frequent item sets and association rules using the Apriori or FP-growth algorithm library of Python, e.g., efficient-Apriori.
The results are stored in a database as a ranked list of relevancy.
When an advertisement is to be pushed for a particular user, the user's purchase record and relevancy ranking list are queried to determine the domain that should be referenced.
According to the preference label of the user in the reference field, a cosine similarity or other text matching algorithm is used for finding products similar to the label in the non-purchased field.
Advertisement pushing is performed using a push library of Python or other languages, such as Firebase Cloud Messaging.
All modules can be packaged as a Docker container and deployed and expanded using Kubernetes or Docker Swarm.
ELK Stack (elastic search, logstack, kibana) or Grafana+Prometheus was used to monitor system operation.
To increase security, it may be considered to connect to a server using a VPN or a private network and encrypt all transmitted data using SSL/TLS.
The invention also discloses a computer readable storage medium, which stores thereon:
executing an instruction: when the execution instruction is executed by the computer, the computer is caused to implement the advertisement pushing method;
domain information database: storing user purchase records of various fields;
association ranking list library: storing a relevance ranking list of each field;
product label library: storing the related label information for each product;
user tag preference database: the tag preferences of the user are stored according to their purchase records in different fields.
The invention also discloses an AI-based advertisement pushing device, which comprises:
a processor;
the storage medium described above: the method comprises an execution instruction, a domain information database, a relevancy ranking list database, a product tag database and a user tag preference database;
the apparatus implements the advertisement pushing method described above when the processor executes the execution instructions on the storage medium.
The foregoing is only a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art, who is within the scope of the present invention, should be covered by the protection scope of the present invention by making equivalents and modifications to the technical solution and the inventive concept thereof.
Claims (8)
1. An AI-based advertisement pushing method is characterized by comprising the following steps:
s1, acquiring purchase records of a user used as training data, wherein each record marks corresponding field information;
s2, integrating purchase records of all users in the S1, carrying out relevance analysis among different fields by using a relevance rule learning algorithm, and generating a relevance ranking list for each field, wherein other fields are ranked in the relevance ranking list according to the relevance of the relevance ranking list to the field;
s2 specifically comprises:
s2.1, converting the purchase record of each user into a transaction data set, wherein the transaction data set comprises a plurality of transactions, and each transaction comprises all fields purchased by the user in one shopping;
s2.2, a frequent item set is found out from the transaction data set by adopting a frequent item set mining method;
s2.3, for each frequent item set, calculating the support, confidence and lifting degree of the frequent item set;
s2.4, generating association rules from one domain to another domain according to the frequent item set; each association rule has a corresponding support degree, confidence degree and promotion degree;
s2.5, setting an association rule score for each association rule, wherein the association rule score is a weighted sum of support degree, confidence degree and lifting degree;
s2.6, for one field, calculating association rule scores from the field to other fields, and sorting from high to low according to the association rule scores to obtain an association degree sorting list;
s3, when the product advertisement of the non-purchased domain is required to be pushed for a specific user, calling out a relevancy sorting list of the non-purchased domain, and identifying one or more domains with purchase records in the relevancy sorting list as reference domains;
s4: for each reference field, pushing product advertisements of the non-purchased field to a specific user based on the label specifically comprises the following steps:
s4.1, distributing labels for each product;
s4.2, summarizing the product label preference of the specific user according to the products purchased by the specific user in the reference field;
s4.3, searching products similar to the tag preference in the non-purchased field to push advertisements, wherein the method specifically comprises the following steps:
converting the label of each product in the reference field into a label vector by using a coding algorithm and calculating the average value of the label vectors to obtain an overall label preference vector;
calculating the similarity between each product in the non-purchased field and the overall label preference vector by using a similarity measurement algorithm, and selecting the priority of advertisement pushing according to the similarity sorting;
the pushed advertisement is a video advertisement;
in the S1, the training data also comprises the stay time of the user on the historical video advertisement; if the stay time of the user on the historical video advertisement exceeds a preset threshold, taking a product corresponding to the video advertisement as a purchase record of the user and adding the record into the relevance analysis in S2;
and S4, adding the products corresponding to the video advertisements with the stay time exceeding the preset threshold value into the products purchased by the specific user in the reference field according to the stay time of the specific user on the historical video advertisements.
2. The AI-based advertisement pushing method of claim 1, wherein:
s1 further comprises: data cleansing is performed on the collected plurality of user purchase records, including removing duplicate purchase records, processing missing data, and converting unstructured data.
3. The AI-based advertisement pushing method of claim 1, wherein the tag in S4 includes at least any one or more of: brands, price interval properties, user rating keywords.
4. The AI-based advertisement pushing method of claim 1, wherein the frequent item set mining method is an Apriori algorithm.
5. The AI-based advertisement pushing method of claim 1, wherein advertisement pushing includes a passive pushing scenario and an automatic pushing scenario:
the passive pushing scene is as follows: when a particular user searches for products in an un-purchased area;
the automatic pushing scene is as follows: sorting the purchased product areas according to the product purchase quantity of each of the product areas purchased by the specific user; and calling a relevance ranking list for ranking the previous one or more purchased product fields, and querying the previous one or more non-purchased product fields in the relevance ranking list as advertisement pushing fields.
6. An AI-based advertisement pushing apparatus, comprising:
and a data acquisition module: the method comprises the steps of acquiring purchase records of a user used as training data, and labeling corresponding field information for each record;
and the data integration and association analysis module: the method comprises the steps of integrating purchase records of all users acquired by a data acquisition module, carrying out relevance analysis among different fields by using relevance rule learning, and generating a relevance ranking list;
an advertisement push selection module: when a product advertisement of an un-purchased domain is required to be pushed for a specific user, calling a relevancy ranking list, and identifying one or more previous domains with purchase records in the relevancy ranking list as reference domains;
and the pushing module is used for: summarizing user preferences based on the product tags in each reference field, and searching products similar to the tag preferences in the non-purchased fields for advertisement pushing;
the pushed advertisement is a video advertisement;
the training data also comprises the stay time of the user on the historical video advertisement; if the stay time of the user on the historical video advertisement exceeds a preset threshold, taking a product corresponding to the video advertisement as a purchase record of the user and adding the record into the relevance analysis in S2;
adding products corresponding to the video advertisements with the stay time exceeding a preset threshold value into products purchased by the specific user in the reference field according to the stay time of the specific user on the historical video advertisements;
the obtaining process of the association degree sequencing list specifically comprises the following steps:
s2.1, converting the purchase record of each user into a transaction data set, wherein the transaction data set comprises a plurality of transactions, and each transaction comprises all fields purchased by the user in one shopping;
s2.2, a frequent item set is found out from the transaction data set by adopting a frequent item set mining method;
s2.3, for each frequent item set, calculating the support, confidence and lifting degree of the frequent item set;
s2.4, generating association rules from one domain to another domain according to the frequent item set; each association rule has a corresponding support degree, confidence degree and promotion degree;
s2.5, setting an association rule score for each association rule, wherein the association rule score is a weighted sum of support degree, confidence degree and lifting degree;
and S2.6, calculating the association rule score from one domain to other domains, and sorting from high to low according to the association rule score to obtain an association degree sorting list.
7. A computer-readable storage medium, wherein the storage medium has stored thereon:
executing an instruction: when the execution instructions are executed by a computer, cause the computer to implement the advertisement pushing method of claim 1;
domain information database: storing user purchase records of various fields;
association ranking list library: storing a relevance ranking list of each field;
product label library: storing the related label information for each product;
user tag preference database: the tag preferences of the user are stored according to their purchase records in different fields.
8. An AI-based advertisement pushing apparatus, comprising:
a processor;
the storage medium of claim 7: the method comprises an execution instruction, a domain information database, a relevancy ranking list database, a product tag database and a user tag preference database;
the apparatus implements the advertisement pushing method of claim 1 when the processor executes the execution instructions on the storage medium.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311124241.2A CN116843394B (en) | 2023-09-01 | 2023-09-01 | AI-based advertisement pushing method, device, equipment and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311124241.2A CN116843394B (en) | 2023-09-01 | 2023-09-01 | AI-based advertisement pushing method, device, equipment and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN116843394A CN116843394A (en) | 2023-10-03 |
CN116843394B true CN116843394B (en) | 2023-11-21 |
Family
ID=88172927
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311124241.2A Active CN116843394B (en) | 2023-09-01 | 2023-09-01 | AI-based advertisement pushing method, device, equipment and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116843394B (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107798021A (en) * | 2016-09-07 | 2018-03-13 | 北京京东尚科信息技术有限公司 | Data correlation processing method, system and electronic equipment |
CN110990717A (en) * | 2019-11-22 | 2020-04-10 | 广西师范大学 | Interest point recommendation method based on cross-domain association |
CN111291261A (en) * | 2020-01-21 | 2020-06-16 | 江西财经大学 | Cross-domain recommendation method integrating label and attention mechanism and implementation system thereof |
CN116629983A (en) * | 2023-07-24 | 2023-08-22 | 成都晓多科技有限公司 | Cross-domain commodity recommendation method and system based on user preference |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8775230B2 (en) * | 2008-11-03 | 2014-07-08 | Oracle International Corporation | Hybrid prediction model for a sales prospector |
-
2023
- 2023-09-01 CN CN202311124241.2A patent/CN116843394B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107798021A (en) * | 2016-09-07 | 2018-03-13 | 北京京东尚科信息技术有限公司 | Data correlation processing method, system and electronic equipment |
CN110990717A (en) * | 2019-11-22 | 2020-04-10 | 广西师范大学 | Interest point recommendation method based on cross-domain association |
CN111291261A (en) * | 2020-01-21 | 2020-06-16 | 江西财经大学 | Cross-domain recommendation method integrating label and attention mechanism and implementation system thereof |
CN116629983A (en) * | 2023-07-24 | 2023-08-22 | 成都晓多科技有限公司 | Cross-domain commodity recommendation method and system based on user preference |
Also Published As
Publication number | Publication date |
---|---|
CN116843394A (en) | 2023-10-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Wu et al. | Turning clicks into purchases: Revenue optimization for product search in e-commerce | |
CN112035742B (en) | User portrait generation method, device, equipment and storage medium | |
US9767166B2 (en) | System and method for predicting user behaviors based on phrase connections | |
US10268670B2 (en) | System and method detecting hidden connections among phrases | |
US11257049B1 (en) | Item identification based on receipt image | |
CN108664513B (en) | Method, device and equipment for pushing keywords | |
CN107833082B (en) | Commodity picture recommendation method and device | |
US9846885B1 (en) | Method and system for comparing commercial entities based on purchase patterns | |
CN111199428A (en) | Commodity recommendation method and device, storage medium and computer equipment | |
CN109635198B (en) | Method, device, medium and electronic equipment for presenting user search results on commodity display platform | |
CN105653562B (en) | The calculation method and device of correlation between a kind of content of text and inquiry request | |
CN111161021B (en) | Quick secondary sorting method for recommended commodities based on real-time characteristics | |
CN111325609A (en) | Commodity recommendation list determining method and device, electronic equipment and storage medium | |
CN108090807B (en) | Information recommendation method and device | |
US20230089850A1 (en) | Real-time product environmental impact scoring | |
CN115544242B (en) | Big data-based similar commodity model selection recommendation method | |
CN110111167A (en) | A kind of method and apparatus of determining recommended | |
CN111914188A (en) | Method, system, device and storage medium for selecting recommendation target user | |
CN112488781A (en) | Search recommendation method and device, electronic equipment and readable storage medium | |
KR20160070282A (en) | Providing system and method for shopping mall web site, program and recording medium thereof | |
Prasetyo | Searching cheapest product on three different e-commerce using k-means algorithm | |
KR20190081671A (en) | Method and server for searching for similar items on online shoppingmall integrated management system | |
CN107153697A (en) | Product search method and device in a kind of commodity transaction website | |
CN112036987B (en) | Method and device for determining recommended commodity | |
CN116843394B (en) | AI-based advertisement pushing method, device, equipment 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 |