CN109474703B - Personalized product combination pushing method, device and system - Google Patents
Personalized product combination pushing method, device and system Download PDFInfo
- Publication number
- CN109474703B CN109474703B CN201811581395.3A CN201811581395A CN109474703B CN 109474703 B CN109474703 B CN 109474703B CN 201811581395 A CN201811581395 A CN 201811581395A CN 109474703 B CN109474703 B CN 109474703B
- Authority
- CN
- China
- Prior art keywords
- characteristic information
- product
- cluster
- terminal
- clusters
- 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
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/55—Push-based network services
-
- 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
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
Landscapes
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Engineering & Computer Science (AREA)
- Finance (AREA)
- Marketing (AREA)
- Signal Processing (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Strategic Management (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application provides a personalized product combination pushing method, a personalized product combination pushing device and a personalized product combination pushing system, wherein the method comprises the following steps: extracting a plurality of terminal characteristic information and a plurality of product characteristic information from historical transaction data, and constructing a characteristic information set comprising the plurality of terminal characteristic information and the plurality of product characteristic information; the characteristic dimension in the terminal characteristic information is the same as the characteristic dimension in the product characteristic information; performing clustering operation on the feature information set to obtain a plurality of clusters; wherein each cluster includes at least one product characteristic information, and, one or more terminal characteristic information; and pushing personalized product combinations to the terminals corresponding to the terminal characteristic information in each cluster, wherein the personalized product combinations comprise products corresponding to one or more product characteristic information in the cluster. According to the method and the device, the personalized product combination is pushed based on the transaction data, so that the accuracy of pushing the personalized product combination can be improved.
Description
Technical Field
The application relates to the technical field of internet, in particular to a personalized product combination pushing method, device and system.
Background
With the development of a fusion technology in the internet in recent years, a service (the personalized product combination comprises one or more products) for pushing the personalized product combination to the terminal is derived, so that the terminal can display the personalized product combination, and a user can conveniently perform operations such as viewing, paying attention to, subscribing and the like on the personalized product combination.
In order to implement a service of pushing personalized product combinations to a terminal, historical behavior characteristics (e.g., browsing product behavior characteristics, product behavior characteristics of interest, and product behavior characteristics of subscription) of a plurality of terminals are generally collected to build a preference model, and then the personalized product combinations of the terminal are estimated by using the preference model to push the personalized product combinations to the terminal.
However, in view of factors such as the specificity and legal compliance of the transaction type scene to which the product belongs, it is difficult to expand a social scene rich in personalized product combination services, so that there are an expansibility problem (Scalability) and a Sparsity problem (Sparsity) in the construction of the preference model.
Therefore, the method for obtaining the personalized product combination based on the preference model cannot accurately determine the personalized product combination of the terminal, so that the accurate personalized product combination cannot be pushed to the terminal.
Disclosure of Invention
In view of this, the present application provides a method, an apparatus, and a system for pushing personalized product combinations, which can push accurate personalized product combinations to a terminal.
In order to achieve the above object, the present application provides the following technical features:
a personalized product combination pushing method comprises the following steps:
extracting a plurality of terminal characteristic information and a plurality of product characteristic information from historical transaction data, and constructing a characteristic information set comprising the plurality of terminal characteristic information and the plurality of product characteristic information; the characteristic dimension in the terminal characteristic information is the same as the characteristic dimension in the product characteristic information;
performing clustering operation on the feature information set to obtain a plurality of clusters; wherein each cluster includes at least one product characteristic information, and, one or more terminal characteristic information;
and pushing an individualized product combination to a terminal corresponding to the terminal characteristic information in each cluster, wherein the individualized product combination comprises products corresponding to one or more product characteristic information in the cluster.
Optionally, the performing a clustering operation on the feature information set to obtain a plurality of clusters includes:
determining K pieces of product characteristic information from the characteristic information set, and respectively using the K pieces of product characteristic information as cluster centers of K clusters;
dividing the rest of feature information in the feature information set into clusters with the minimum distance from the cluster center to obtain K clusters;
re-determining the cluster centers of the K clusters; wherein, the cluster center of each cluster is product characteristic information in the cluster;
judging whether the cluster centers of the K clusters change or not; if yes, entering the step: dividing the rest of feature information in the feature information set into clusters with the minimum distance from the cluster center to obtain K clusters;
and if not, determining the K clusters as a plurality of clusters obtained by the clustering operation.
Optionally, the determining the cluster centers of the K clusters again includes:
the following process is performed for each of the K clusters:
calculating the real cluster center of the cluster, respectively calculating the distance between each product characteristic information in the cluster and the real cluster center, and determining the product characteristic information with the minimum distance between the cluster and the real cluster center as the cluster center of the cluster.
Optionally, the plurality of terminal feature information and the plurality of product feature information are extracted from the historical transaction data, and a feature information set including the plurality of terminal feature information and the plurality of product feature information is constructed; wherein, the characteristic dimension in the terminal characteristic information is the same as the characteristic dimension in the product characteristic information, and the method comprises the following steps:
extracting a plurality of terminal feature sets corresponding to a plurality of terminals one by one and extracting a plurality of product feature sets corresponding to a plurality of products one by one from historical transaction data; the feature dimension in the terminal feature set is the same as the feature dimension in the product feature set;
determining the weight corresponding to each characteristic dimension;
executing vectorization operation and normalization operation on a plurality of terminal feature sets and a plurality of product feature sets to obtain a plurality of terminal feature information and a plurality of product feature information;
and determining the plurality of terminal characteristic information and the plurality of product characteristic information as a characteristic information set.
Optionally, the feature dimensions in the terminal feature set and the feature dimensions in the product feature set include:
asset class, transaction class, profit-loss class, interest class, and capability class.
A personalized product combination pusher comprising:
the construction unit is used for extracting a plurality of terminal characteristic information and a plurality of product characteristic information from historical transaction data and constructing a characteristic information set containing the plurality of terminal characteristic information and the plurality of product characteristic information; the characteristic dimension in the terminal characteristic information is the same as the characteristic dimension in the product characteristic information;
the clustering unit is used for carrying out clustering operation on the characteristic information set to obtain a plurality of clusters; wherein each cluster includes at least one product characteristic information, and, one or more terminal characteristic information;
and the pushing unit is used for pushing personalized product combinations to the terminals corresponding to the terminal characteristic information in each cluster, and each personalized product combination comprises one or more products corresponding to the product characteristic information in the cluster.
Optionally, the clustering unit includes:
a first cluster center determining unit, configured to determine K pieces of product feature information from the feature information set, where the K pieces of product feature information are used as cluster centers of the K clusters, respectively;
the dividing unit is used for dividing the rest of feature information in the feature information set into clusters with the minimum distance from the center of each cluster to obtain K clusters;
a second cluster center determining unit for re-determining the cluster centers of the K clusters; wherein, the cluster center of each cluster is product characteristic information in the cluster;
the judging unit is used for judging whether the cluster centers of the K clusters change or not;
if the cluster centers of the K clusters change, entering a dividing unit;
and the determining unit is used for determining that the clustering operation is finished and determining the K clusters as a plurality of clusters obtained by the clustering operation if the cluster centers of the K clusters are not changed. A personalized product combination pushing system comprises:
the server is used for extracting a plurality of terminal characteristic information and a plurality of product characteristic information from historical transaction data and constructing a characteristic information set comprising the plurality of terminal characteristic information and the plurality of product characteristic information; the characteristic dimension in the terminal characteristic information is the same as the characteristic dimension in the product characteristic information; performing clustering operation on the feature information set to obtain a plurality of clusters; wherein each cluster includes at least one product characteristic information, and, one or more terminal characteristic information; pushing an individualized product combination to a terminal corresponding to the terminal characteristic information in each cluster, wherein the individualized product combination comprises products corresponding to one or more product characteristic information in the cluster;
and the terminal is used for receiving and displaying the personalized product combination.
An electronic device, comprising:
a processor; and
a memory for storing executable instructions of the processor;
wherein the processor is configured to execute the personalized product portfolio push method via execution of the executable instructions.
A storage medium for storing a software program, the software program being operable to implement the personalized product combination push method.
Through the technical means, the following beneficial effects can be realized:
compared to determining personalized product combinations using a preference model built from historical behavioral features (extrinsic features), the present application proposes that the personalized product combinations are determined in a clustered manner based on a feature information set determined from historical transaction data (intrinsic features).
In order to realize the clustering operation of two different types of objects of a terminal and a product, the terminal feature set and the product feature set have the same feature dimension, so that the terminal and the product can be regarded as the same type of object during the clustering operation, and the clustering operation of the terminal and the product can be realized.
Since the historical transaction data (internal characteristics) can reflect the essential relationship between the terminal and the product more than the historical behavior characteristics (external characteristics), the personalized product combination can be accurately determined based on the historical transaction data, and therefore the accuracy of pushing the personalized product combination to the terminal is improved.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
Fig. 1 is a schematic diagram of a personalized product combination pushing system disclosed in an embodiment of the present application;
fig. 2 is a flowchart of a method for pushing a personalized product combination disclosed in an embodiment of the present application;
fig. 3 is a flowchart of a personalized product combination pushing device disclosed in the embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
The applicant found that: the prior art usually collects historical behavior characteristics (for example, historical browsing product object behavior characteristics, concerned product object behavior characteristics and subscribed product object behavior characteristics) of a plurality of terminals to build a preference model.
The historical behavior characteristics are the behavior characteristics exposed from the external layer surface of the terminal user, so the historical behavior characteristics mainly reflect the relationship between the terminal and the product from the external layer surface, and the historical behavior characteristics can be called as external characteristics.
The applicant found in the course of research that: there is a large amount of historical transaction data between the terminal and the product object, and the historical transaction data can reflect the relationship between the terminal and the product from an internal level (particularly can reflect the financial characteristics between the terminal and the product, such as the characteristics of average daily property, stock holding days, short, medium and long line preference, and the like), so the characteristics extracted from the historical transaction data are called internal characteristics.
It can be understood that, because the intrinsic characteristics can more deeply represent the essential relationship between the terminal and the product object than the extrinsic characteristics, the personalized product combination can be more accurately determined based on the historical transaction data, so that the accuracy of pushing the personalized product combination to the terminal is improved.
In order to facilitate the understanding of application scenarios for those skilled in the art, the present application provides a push system for personalized product combinations. Referring to fig. 1, the push system of the personalized product portfolio includes:
a server 100, and a plurality of terminals 200 connected to the server 100.
The terminal may include electronic devices such as a mobile phone, a computer, and an ipad, which are not illustrated herein; the server 100 may be implemented by a single server or a server cluster, and the specific implementation of the server in real-world applications is not limited.
It is understood that during the product transaction between the server 100 and the terminal 200, the server stores the transaction data. The transaction data over a period of time is referred to as historical transaction data. The process of performing product transactions and storing historical transaction data by the server 100 is well-established and will not be described in detail herein.
The application provides a personalized product combination pushing method which is applied to a personalized product combination pushing system shown in figure 1.
Referring to fig. 2, the personalized product combination pushing method may include the following steps:
step S201: the server determines a plurality of feature dimensions.
In general, terminal features representing user characteristics and product features representing product characteristics belong to different feature types, so that the terminal features and the product features cannot be clustered.
In order to subsequently realize that the terminal characteristics and the product characteristics execute clustering operation, the server can determine a plurality of characteristic dimensions, and the terminal and the product can share the plurality of characteristic dimensions.
That is, on the premise that the terminal feature uses the feature dimension representing the user characteristic, the method and the device personify the product, so that the product feature is also suitable for the feature dimension representing the user characteristic, and the feature dimension in the terminal feature is the same as the feature dimension in the product feature.
Therefore, the terminal characteristics and the product characteristics can be regarded as the same characteristic types, and the clustering operation of the terminal characteristic information set and the product characteristic information set can be further realized.
For this purpose, a plurality of feature dimensions suitable for the terminal and the product, and calculation rules of the plurality of feature dimensions may be manually set, and the calculation rules of the plurality of feature dimensions and the plurality of feature dimensions may be stored in the server.
The following illustrates a plurality of characteristic dimensions stored by the server, and table 1 illustrates the plurality of characteristic dimensions:
TABLE 1
The multiple feature dimensions in the server may be dynamically determined according to an actual application scenario (the feature dimensions may be dynamically expanded or deleted), and of course, the calculation rule of the feature dimensions may also be determined according to the actual application scenario, and the specific calculation rule about the feature dimensions is not described in detail again.
Step S202: and the server gives corresponding weight to each characteristic dimension.
The server stores a plurality of feature dimensions and divides the plurality of features into a plurality of types. Taking table 1 as an example, a plurality of feature dimensions may be divided into: assets, transactions, gains and losses, interests and abilities.
The influence degree of different types of feature dimensions on the personalized product combination can be determined according to the actual application scene, the feature dimensions with large influence degree are endowed with large weight, and the feature dimensions with small influence degree are endowed with small weight. The specific implementation process is a mature technology and is not described in detail herein.
The foregoing is a preliminary execution process of the present application, and the execution process of the present application is described below.
Step S203: the server extracts a plurality of terminal characteristic information and a plurality of product characteristic information from historical transaction data, and constructs a characteristic information set comprising the plurality of terminal characteristic information and the plurality of product characteristic information; and the characteristic dimension in the terminal characteristic information is the same as the characteristic dimension in the product characteristic information.
According to an embodiment provided by the present application, step 203 may include steps S2031 to S2034:
step S2031: the server extracts a plurality of terminal feature sets corresponding to a plurality of terminals one by one and a plurality of product feature sets corresponding to a plurality of products one by one from historical transaction data; and the feature dimension in the terminal feature set is the same as the feature dimension in the product feature set.
In the embodiment, the plurality of terminals and the plurality of products can be all terminals and all products in the personalized product combination recommendation service which needs to be executed; of course, it may be a partial terminal and a partial product. The specific application scenario may depend on the application scenario, and is not limited herein.
The server stores historical transaction data, a plurality of characteristic dimensions and calculation rules of the characteristic dimensions. For a plurality of terminals, terminal features corresponding to each feature dimension can be extracted from historical transaction data according to a calculation rule of each feature dimension, so that a plurality of terminal feature sets corresponding to the plurality of terminals one by one are obtained.
The historical transaction data is actually user transaction data, so a terminal feature set corresponding to the terminal is extracted from the historical transaction data, which is equivalent to extracting a user feature set corresponding to the user, that is, feature dimensions in the terminal feature set are feature dimensions representing characteristics of the user.
In order to realize the clustering operation of the terminal feature set and the product feature set, the product is personified, namely for a plurality of products: and extracting product features corresponding to each feature dimension according to a calculation rule of each feature dimension (the feature dimension representing the characteristics of the user) from the historical transaction data, so as to obtain a plurality of product feature sets corresponding to a plurality of products one by one.
The feature dimensions in the terminal feature set are the same as those in the product feature set, so the terminal feature set and the product feature set can be regarded as the same feature type.
Step S2032: and the server acquires the weight corresponding to each characteristic dimension.
The data values of different feature dimensions may be greatly different, so that the data of each feature dimension needs to be normalized, the normalization process may be performed in a min-max normalization manner, and the specific implementation process is a mature technology and is not described in detail herein.
Step S2033: and the server executes vectorization operation and normalization operation on the plurality of terminal characteristic sets and the plurality of product characteristic sets to obtain a plurality of terminal characteristic information and a plurality of product characteristic information.
For convenience of subsequent processing, the terminal feature set and the product feature set may be vectorized. Taking the terminal feature set as an example, assuming that the terminal feature set includes T types of features, and each type includes N feature dimensions, the terminal feature set may be represented by T N-dimensional vectors.
Of course, the feature dimensions may be different in each type of feature. Taking table 1 as an example, the terminal feature set includes 5 types of features: if the asset class comprises 8 characteristic dimensions, an 8-dimensional vector representation can be adopted; the transaction class comprises 7 characteristic dimensions which can be represented by a 7-dimensional vector; the profit and loss class comprises 13 characteristic dimensions, and a 13-dimensional vector can be used for representation; the interest class comprises 4 characteristic dimensions, and a 4-dimensional vector can be used for representation; the capability class includes 4 characteristic dimensions; a 4-dimensional vector representation may be employed.
After the plurality of terminal feature sets are normalized and vectorized, a plurality of terminal feature sets are generated, and the terminal feature sets are called as a plurality of pieces of terminal feature information for being distinguished from the terminal feature sets.
After the plurality of product feature sets are normalized and vectorized, a plurality of product feature vector sets are generated, and the product feature vector sets are called as a plurality of product feature information for being distinguished from the plurality of product feature sets.
Step S2034: and constructing a feature information set containing a plurality of terminal feature information and a plurality of product feature information.
Step S203 then proceeds to step S204: the server carries out clustering operation on the characteristic information set to obtain a plurality of clusters; wherein each cluster includes at least one product characteristic information, and one or more terminal characteristic information.
As the name implies, the clustering operation is to cluster similar feature information together, so that the terminal feature information and the product feature information in one cluster after the clustering operation are relatively similar. That is, the terminal corresponding to the terminal characteristic information is similar to the product corresponding to the product characteristic information, i.e., the terminal is suitable for recommending the products in the cluster to the terminals in the cluster.
According to an embodiment provided by the present application, step S204 includes steps S2041 to S2044:
step S2041: and the server determines K pieces of product characteristic information from the characteristic information set, and the K pieces of product characteristic information are respectively used as cluster centers of the K clusters.
After determining the feature information set, the server randomly determines K pieces of product feature information from the feature information set, and the K pieces of product feature information serve as cluster centers of K clusters. The purpose of selecting K pieces of product characteristic information instead of the terminal characteristic information is to enable each cluster to at least comprise one piece of product characteristic information so as to recommend a product corresponding to the product characteristic information to a terminal corresponding to the terminal characteristic information in the cluster.
Step S2042: and the server divides the rest of the feature information in the feature information set into the cluster with the minimum distance from the center of the cluster to obtain K clusters.
The server executes the following operation aiming at each piece of feature information (including terminal feature information and remaining product feature information) in the rest feature information (except the K cluster centers) in the feature information set:
respectively calculating Euclidean distances between the feature information and K cluster centers, determining the cluster center with the minimum Euclidean distance with the feature information, and dividing the feature information into the clusters to which the cluster centers belong.
And after the server executes the upper-section description operation on each piece of feature information in the rest feature information in the feature information set, the server realizes the clustering operation on the feature information set once. That is, performing a clustering operation on the feature information set obtains K clusters.
Step S2043: the server re-determines the cluster centers of the K clusters; wherein the cluster center of each cluster is product characteristic information in the cluster.
The server performs the following process for each of the K cluster centers:
s1: the server computes the real cluster center of the cluster.
Assume that a cluster contains M pieces of feature information (terminal feature information and product feature information), each piece of feature information contains a feature type, and the type contains an N-dimensional vector.
The real cluster center is included as a feature information, the feature information includes a feature type, the feature type includes N-dimensional vectors, and the data value of each dimensional vector is an average value of the data values of the dimensional vectors in the M pieces of feature information.
S2: and the server respectively calculates the distance between each product characteristic information in the cluster and the center of the real cluster.
And calculating the Euclidean distance between each piece of characteristic information and the center of the real cluster according to each piece of characteristic information in the cluster.
S3: and the server determines the product characteristic information with the minimum distance from the center of the real cluster in the cluster as the cluster center of the cluster.
Since the real cluster center may not correspond to a certain product feature information, in order to ensure that each cluster has at least one product feature information, the embodiment biases to the product feature information when the cluster center is determined again.
That is, it is ensured that the cluster center is the product characteristic information each time, so that at least one product characteristic information in each cluster can be ensured, and the purpose is to recommend the product corresponding to the product characteristic information to the terminal corresponding to the terminal characteristic information in the cluster.
Step S2043 then proceeds to step S2044: the server judges whether the cluster centers of the K clusters change or not; if yes, the process proceeds to step S2042, otherwise, the process ends.
The server judges whether the original cluster centers of the K clusters are completely consistent with the cluster centers after the K clusters are re-determined, if the original cluster centers of the K clusters are completely consistent with the cluster centers after the K clusters are re-determined, the K cluster centers are not changed, the K clusters are not changed, the characteristic information set is clustered, and the clustering operation can be finished.
If the clusters are not completely consistent, that is, the cluster centers are changed, it indicates that the clustering of the feature information set is not completed, and step S2042 may be entered to start the next cyclic clustering operation until the K cluster centers are not changed any more, that is, the K clusters are not changed any more.
Step S204 proceeds to step S205: and the server pushes an individualized product combination to the terminal corresponding to the terminal characteristic information in each cluster, wherein the individualized product combination comprises products corresponding to one or more product characteristic information in the cluster.
It can be understood that after the clustering operation is finished, each cluster comprises at least one product characteristic information and a plurality of terminal characteristic information, and the product characteristic information corresponds to the product and the terminal characteristic information corresponds to the terminal; each cluster includes at least one product and several terminals.
The closest distance between the terminal in each cluster and the products in that cluster, the products in one cluster are more adapted to the terminals in that cluster than the products in the other clusters. Therefore, the product in each cluster is the personalized product of the terminal in the cluster.
The server performs the following process for each of the K clusters:
it can be understood that one product feature information corresponds to one product, so that a personalized product combination can be formed by products corresponding to one or more product feature information in the cluster.
The personalized product combination can be represented in a list form, and optionally, one or more products in the list can be randomly arranged, so that the recommended frequency of each product is equalized, the investment frequency of the products in each cluster, which are invested by a user, is also equalized, and the standardization of the industry is enhanced.
Step S206: and receiving and displaying the personalized product combination by a plurality of terminals.
After the server pushes the corresponding personalized product combinations to the terminals, the terminals can display the personalized product combinations for the user to perform operations such as viewing, paying attention to, subscribing and the like.
According to the embodiments, the following advantages are provided:
in contrast to using historical behavioral, i.e. extrinsic, characteristics to determine personalized product combinations, the present application proposes to determine personalized product combinations based on transaction data, i.e. intrinsic characteristics. Because the intrinsic characteristics can reflect the essential relationship between the terminal and the product more than the extrinsic characteristics, the personalized product combination is pushed based on the transaction data, and the accuracy of pushing the personalized product combination can be improved.
In addition, in order to realize the clustering operation of the terminal and the product, the method personifies the product object, so that the characteristic set of the terminal and the characteristic set of the product have the same characteristic dimension, the terminal object and the product object are both regarded as the same type of object, and the clustering operation of the terminal and the product can be further realized.
According to the method and the device, clustering operation is carried out on the user and the products in the subsequent process, so that the personalized product combination which is most suitable for the user is determined and then pushed to the terminal corresponding to the user.
Referring to fig. 3, the present application further provides a personalized product combination pushing device, including:
a construction unit 31, configured to extract a plurality of terminal feature information and a plurality of product feature information from historical transaction data, and construct a feature information set including the plurality of terminal feature information and the plurality of product feature information; the characteristic dimension in the terminal characteristic information is the same as the characteristic dimension in the product characteristic information;
a clustering unit 32, configured to perform a clustering operation on the feature information set to obtain a plurality of clusters; wherein each cluster includes at least one product characteristic information, and, one or more terminal characteristic information;
the pushing unit 33 is configured to push an individualized product combination to a terminal corresponding to the terminal feature information in each cluster, where the individualized product combination includes products corresponding to one or more product feature information in the cluster.
Wherein the clustering unit 32 comprises:
a first cluster center determining unit 321, configured to determine K pieces of product feature information from the feature information set, where the K pieces of product feature information are respectively used as cluster centers of the K clusters;
a dividing unit 322, configured to divide the remaining feature information in the feature information set into clusters with a minimum distance from a cluster center, to obtain K clusters;
a second cluster center determining unit 324 for re-determining the cluster centers of the K clusters; wherein, the cluster center of each cluster is product characteristic information in the cluster;
a judging unit 325, configured to judge whether cluster centers of the K clusters change;
if the cluster centers of the K clusters change, the K clusters enter a dividing unit 322;
a cluster determining unit 326, configured to determine that the clustering operation is ended if the cluster centers of the K clusters do not change, and determine the K clusters as multiple clusters obtained by the clustering operation.
Wherein the second determination cluster center unit 324 includes:
the following process is performed for each of the K clusters: calculating the real cluster center of the cluster, respectively calculating the distance between each product characteristic information in the cluster and the real cluster center, and determining the product characteristic information with the minimum distance between the cluster and the real cluster center as the cluster center of the cluster.
Wherein, the construction unit 31 includes:
an extracting unit 311, configured to extract, from historical transaction data, a plurality of terminal feature sets that are in one-to-one correspondence with a plurality of terminals and a plurality of product feature sets that are in one-to-one correspondence with a plurality of products; the feature dimension in the terminal feature set is the same as the feature dimension in the product feature set;
a weight determining unit 312, configured to determine weights corresponding to the feature dimensions;
a processing set unit 313, configured to perform vectorization operation and normalization operation on the plurality of terminal feature sets and the plurality of product feature sets, so as to obtain a plurality of terminal feature information and a plurality of product feature information;
a determining set unit 314, configured to determine a plurality of terminal feature information and a plurality of product feature information as a feature information set.
Wherein, the characteristic dimension in the terminal characteristic set and the characteristic dimension in the product characteristic set include:
asset class, transaction class, profit-loss class, interest class, and capability class.
For a specific implementation of the personalized product combination pushing device, reference may be made to the embodiment shown in fig. 2, which is not described herein again.
Referring to fig. 1, the present application provides a personalized product combination pushing system, including:
the server is used for extracting a plurality of terminal characteristic information and a plurality of product characteristic information from historical transaction data and constructing a characteristic information set comprising the plurality of terminal characteristic information and the plurality of product characteristic information; the characteristic dimension in the terminal characteristic information is the same as the characteristic dimension in the product characteristic information; performing clustering operation on the feature information set to obtain a plurality of clusters; wherein each cluster includes at least one product characteristic information, and, one or more terminal characteristic information; pushing an individualized product combination to a terminal corresponding to the terminal characteristic information in each cluster, wherein the individualized product combination comprises products corresponding to one or more product characteristic information in the cluster;
and the terminal is used for receiving and displaying the personalized product combination.
For a specific implementation of the personalized product combination pushing system, reference may be made to the embodiment shown in fig. 2, which is not described herein again.
The present application further provides an electronic device, comprising:
a processor; and
a memory for storing executable instructions of the processor;
wherein the processor is configured to execute the personalized product combination pushing method shown in fig. 2 via executing the executable instructions. For specific implementation of the personalized product combination push amplification, reference may be made to the embodiment shown in fig. 2, which is not described herein again.
The application also provides a storage medium, wherein the storage medium is used for storing a software program, and the software program can be used for realizing the personalized product combination pushing method. For specific implementation of the personalized product combination push amplification, reference may be made to the embodiment shown in fig. 2, which is not described herein again.
The functions described in the method of the present embodiment, if implemented in the form of software functional units and sold or used as independent products, may be stored in a storage medium readable by a computing device. Based on such understanding, part of the contribution to the prior art of the embodiments of the present application or part of the technical solution may be embodied in the form of a software product stored in a storage medium and including several instructions for causing a computing device (which may be a personal computer, a server, a mobile computing device or a network device) to execute all or part of the steps of the method described in the embodiments of the present application. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
The embodiments are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same or similar parts among the embodiments are referred to each other.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (8)
1. A personalized product combination pushing method is characterized by comprising the following steps:
extracting a plurality of terminal characteristic information and a plurality of product characteristic information from historical transaction data, and constructing a characteristic information set comprising the plurality of terminal characteristic information and the plurality of product characteristic information; the characteristic dimension in the terminal characteristic information is the same as the characteristic dimension in the product characteristic information;
performing clustering operation on the feature information set to obtain a plurality of clusters; wherein each cluster includes at least one product characteristic information, and, one or more terminal characteristic information;
pushing an individualized product combination to a terminal corresponding to the terminal characteristic information in each cluster, wherein the individualized product combination comprises products corresponding to one or more product characteristic information in the cluster;
the performing a clustering operation on the feature information set to obtain a plurality of clusters includes:
determining K pieces of product characteristic information from the characteristic information set, and respectively using the K pieces of product characteristic information as cluster centers of K clusters;
dividing the rest of feature information in the feature information set into clusters with the minimum distance from the cluster center to obtain K clusters;
re-determining the cluster centers of the K clusters; wherein, the cluster center of each cluster is product characteristic information in the cluster;
judging whether the cluster centers of the K clusters change or not; if yes, entering the step: dividing the rest of feature information in the feature information set into clusters with the minimum distance from the cluster center to obtain K clusters;
and if not, determining the K clusters as a plurality of clusters obtained by the clustering operation.
2. The method of claim 1, wherein said re-determining cluster centers for K clusters comprises:
the following process is performed for each of the K clusters:
calculating the real cluster center of the cluster, respectively calculating the distance between each product characteristic information in the cluster and the real cluster center, and determining the product characteristic information with the minimum distance between the cluster and the real cluster center as the cluster center of the cluster.
3. The method of claim 1, wherein the extracting of the plurality of terminal characteristic information and the plurality of product characteristic information from the historical transaction data constructs a characteristic information set comprising the plurality of terminal characteristic information and the plurality of product characteristic information; wherein, the characteristic dimension in the terminal characteristic information is the same as the characteristic dimension in the product characteristic information, and the method comprises the following steps:
extracting a plurality of terminal feature sets corresponding to a plurality of terminals one by one and extracting a plurality of product feature sets corresponding to a plurality of products one by one from historical transaction data; the feature dimension in the terminal feature set is the same as the feature dimension in the product feature set;
determining the weight corresponding to each characteristic dimension;
executing vectorization operation and normalization operation on a plurality of terminal feature sets and a plurality of product feature sets to obtain a plurality of terminal feature information and a plurality of product feature information;
and determining the plurality of terminal characteristic information and the plurality of product characteristic information as a characteristic information set.
4. The method of claim 3, wherein the feature dimensions in the terminal feature set and the feature dimensions in the product feature set comprise:
asset class, transaction class, profit-loss class, interest class, and capability class.
5. A personalized product combination pushing device is characterized by comprising:
the construction unit is used for extracting a plurality of terminal characteristic information and a plurality of product characteristic information from historical transaction data and constructing a characteristic information set containing the plurality of terminal characteristic information and the plurality of product characteristic information; the characteristic dimension in the terminal characteristic information is the same as the characteristic dimension in the product characteristic information;
the clustering unit is used for carrying out clustering operation on the characteristic information set to obtain a plurality of clusters; wherein each cluster includes at least one product characteristic information, and, one or more terminal characteristic information;
the pushing unit is used for pushing personalized product combinations to the terminals corresponding to the terminal characteristic information in each cluster, and the personalized product combinations comprise products corresponding to one or more product characteristic information in the cluster;
the clustering unit includes:
a first cluster center determining unit, configured to determine K pieces of product feature information from the feature information set, where the K pieces of product feature information are used as cluster centers of the K clusters, respectively;
the dividing unit is used for dividing the rest of feature information in the feature information set into clusters with the minimum distance from the center of each cluster to obtain K clusters;
a second cluster center determining unit for re-determining the cluster centers of the K clusters; wherein, the cluster center of each cluster is product characteristic information in the cluster;
the judging unit is used for judging whether the cluster centers of the K clusters change or not;
if the cluster centers of the K clusters change, entering a dividing unit;
and the determining unit is used for determining that the clustering operation is finished and determining the K clusters as a plurality of clusters obtained by the clustering operation if the cluster centers of the K clusters are not changed.
6. A personalized product combination pushing system is characterized by comprising:
the server is used for extracting a plurality of terminal characteristic information and a plurality of product characteristic information from historical transaction data and constructing a characteristic information set comprising the plurality of terminal characteristic information and the plurality of product characteristic information; the characteristic dimension in the terminal characteristic information is the same as the characteristic dimension in the product characteristic information; performing clustering operation on the feature information set to obtain a plurality of clusters; wherein each cluster includes at least one product characteristic information, and, one or more terminal characteristic information; pushing an individualized product combination to a terminal corresponding to the terminal characteristic information in each cluster, wherein the individualized product combination comprises products corresponding to one or more product characteristic information in the cluster; the performing a clustering operation on the feature information set to obtain a plurality of clusters includes: determining K pieces of product characteristic information from the characteristic information set, and respectively using the K pieces of product characteristic information as cluster centers of K clusters; dividing the rest of feature information in the feature information set into clusters with the minimum distance from the cluster center to obtain K clusters; re-determining the cluster centers of the K clusters; wherein, the cluster center of each cluster is product characteristic information in the cluster; judging whether the cluster centers of the K clusters change or not; if yes, entering the step: dividing the rest of feature information in the feature information set into clusters with the minimum distance from the cluster center to obtain K clusters; if not, determining the K clusters as a plurality of clusters obtained by clustering operation;
and the terminal is used for receiving and displaying the personalized product combination.
7. An electronic device, comprising:
a processor; and
a memory for storing executable instructions of the processor;
wherein the processor is configured to execute the personalized product combination pushing method according to any one of claims 1 to 4 via executing the executable instructions.
8. A storage medium for storing a software program, wherein the software program is used for implementing the personalized product combination pushing method according to any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811581395.3A CN109474703B (en) | 2018-12-24 | 2018-12-24 | Personalized product combination pushing method, device and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811581395.3A CN109474703B (en) | 2018-12-24 | 2018-12-24 | Personalized product combination pushing method, device and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109474703A CN109474703A (en) | 2019-03-15 |
| CN109474703B true CN109474703B (en) | 2021-08-31 |
Family
ID=65676783
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811581395.3A Active CN109474703B (en) | 2018-12-24 | 2018-12-24 | Personalized product combination pushing method, device and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109474703B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113112318B (en) * | 2020-01-10 | 2022-04-29 | 阿里巴巴集团控股有限公司 | Method, device, electronic equipment and computer storage medium for determining product combination |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106951489A (en) * | 2017-03-13 | 2017-07-14 | 杭州师范大学 | A kind of personalized recommendation method and device for sparse big data |
Family Cites Families (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7689457B2 (en) * | 2007-03-30 | 2010-03-30 | Amazon Technologies, Inc. | Cluster-based assessment of user interests |
| CN102254028A (en) * | 2011-07-22 | 2011-11-23 | 青岛理工大学 | Personalized commodity recommendation method and system integrating attributes and structural similarity |
| CN102609523B (en) * | 2012-02-10 | 2018-03-06 | 上海视畅信息科技有限公司 | The collaborative filtering recommending method classified based on taxonomy of goods and user |
| US8880204B2 (en) * | 2012-06-27 | 2014-11-04 | Ubiquiti Networks, Inc. | Method and apparatus for monitoring and processing sensor data in an interfacing-device network |
| CN102750647A (en) * | 2012-06-29 | 2012-10-24 | 南京大学 | Merchant recommendation method based on transaction network |
| CN103886003A (en) * | 2013-09-22 | 2014-06-25 | 天津思博科科技发展有限公司 | Collaborative filtering processor |
| CN105869001A (en) * | 2015-01-19 | 2016-08-17 | 苏宁云商集团股份有限公司 | Customized product recommendation guiding method and system |
| CN105931066A (en) * | 2015-09-24 | 2016-09-07 | 中国银联股份有限公司 | Transaction data processing method and device |
| CN107657500B (en) * | 2016-12-03 | 2022-04-29 | 平安证券股份有限公司 | Stock recommendation method and server |
-
2018
- 2018-12-24 CN CN201811581395.3A patent/CN109474703B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106951489A (en) * | 2017-03-13 | 2017-07-14 | 杭州师范大学 | A kind of personalized recommendation method and device for sparse big data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109474703A (en) | 2019-03-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US9348898B2 (en) | Recommendation system with dual collaborative filter usage matrix | |
| CN103678672B (en) | Method for recommending information | |
| WO2020048084A1 (en) | Resource recommendation method and apparatus, computer device, and computer-readable storage medium | |
| CN109242612B (en) | Product recommendation method and device | |
| CN107122369B (en) | Service data processing method, device and system | |
| CN110046929B (en) | Fraudulent party identification method and device, readable storage medium and terminal equipment | |
| CN112214499B (en) | Graph data processing method and device, computer equipment and storage medium | |
| CN110727857A (en) | Method and device for identifying key features of potential users aiming at business objects | |
| CN106933878B (en) | Information processing method and device | |
| CN110851675A (en) | Data extraction method, device and medium | |
| CN112861963A (en) | Method, device and storage medium for training entity feature extraction model | |
| CN112581226A (en) | Product recommendation method and system | |
| CN111538909A (en) | Information recommendation method and device | |
| CN113254788B (en) | Big data based recommendation method and system and readable storage medium | |
| CN109474703B (en) | Personalized product combination pushing method, device and system | |
| CN111062490A (en) | Method and device for processing network data containing private data | |
| CN108959289B (en) | Website category acquisition method and device | |
| CN113448876B (en) | Service testing method, device, computer equipment and storage medium | |
| CN110929285A (en) | Method and device for processing private data | |
| CN110765350A (en) | Data fusion method and device for member points | |
| CN110427558B (en) | Resource processing event pushing method and device | |
| CN113781117A (en) | Marketing scheme generation method and device, electronic equipment and computer storage medium | |
| CN113076450A (en) | Method and device for determining target recommendation list | |
| CN106330745B (en) | Traffic policy selection method and traffic policy selection device | |
| CN112328709B (en) | Entity labeling method and device, server 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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20210811 Address after: Room 323, 3 / F, Hang Seng building, 3588 Jiangnan Avenue, Binjiang District, Hangzhou City, Zhejiang Province 310053 Applicant after: ZHEJIANG JINGTENG NETWORK TECHNOLOGY Co.,Ltd. Address before: 310053 8th floor, building 2, No. 3588 Jiangnan Avenue, Binjiang District, Hangzhou, Zhejiang Applicant before: HANGZHOU YUNJI NETWORK TECHNOLOGY Co.,Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |