WO2023065691A1 - User information fusion method and system under multilayer association, and terminal and storage medium - Google Patents

Abstract

A user information fusion method and system under multilayer association, and a terminal and a storage medium. The method comprises: firstly, a data source of user information to be integrated is selected, the data source corresponds to a material table, channel type information is further obtained according to an association relation in the material table, a user information graph is constructed according to the channel type information, the user information graph is split into independent user connected graphs, a linear material table relation is expanded into a two-dimensional plane graph relation, multilayer complex association is conveniently supported, and in the complex association of a large amount of data, records belonging to a same natural user are found; on the basis of the user connected graphs, an entity unique ID connected graph is further constructed in combination with a historical material table, and two independent user entities in data of the historical material table are fused together with the help of new association provided by newly added user information data. The problem of user information fusion under multilayer association is solved.

Description

User information fusion method, terminal, storage medium and system under multi-layer association technical field

The present invention relates to the technical field of user information fusion, and more specifically, to a user information fusion method, terminal, storage medium and system under multi-layer association.

Background technique

The Customer Data Platform (CDP) includes relevant data about audiences, users or customers that are generated and accumulated during the operation of the enterprise. In CDP, member (user entity) data spreads across multiple information systems/data sources of the enterprise, such as customer relationship management system (Customer Relationship Management, CRM), POS transactions, behavior log collection data, etc. However, the same natural user member will have different unique identifiers in different information systems or channels. For example, JD.com’s user ID and mobile phone are included in the purchase records of JD.com, and there are corresponding user IDs in the enterprise’s own CRM. , the user ID of the same natural user member in JD.com and CRM is different, which brings difficulties to subsequent data processing, labeling, and promotion of marketing activities; moreover, a natural user may have multiple accounts, which are marked in the system For different users, therefore, a unique identifier is needed to identify a natural user, connect user data from different systems and different channels, identify the user data behind which is actually the same natural user, and give it a unique entity ID (after fusion Entity Unique Identifier), which is used for subsequent operations such as subsequent data processing, user labeling system construction, user operation statistical analysis, and marketing campaign push.

On March 14, 2015, a Chinese invention patent (publication number: CN104394118A) disclosed a user identification method and system, the basic information formed through user registration, including user ID, user name, Email, telephone, computer IP, etc. , to extract the user behavior data of the website, the comprehensive behavior data involves user ID, user name, Email, phone number, cookie, computer IP and other information, establish the relationship between the two user information and assign a unique identity, which can be used for the current B2B The users in the website can be identified uniformly, establish the relationship of identity characteristics, distinguish new and old users, and can effectively track user behavior, so that a series of applications can be established for users to improve user experience, but this patent cannot handle multi-layer associated user information In a fusion scenario, for example, if a user's mobile phone number A (corresponding to user ID1) purchases a coupon, and then uses the coupon to make a purchase through the spare mobile phone number B (corresponding to user ID2), the original records of mobile phone number A and mobile phone number B are two different user ID, but through this purchase behavior, it can be considered that the two user IDs belong to the same natural user; however, this method uses the user ID as the unique user identifier, which cannot be associated with this scenario and cannot be integrated. information.

Contents of the invention

In order to solve the problem that the current user information fusion method cannot realize the user information fusion under multi-layer association, the present invention proposes a user information fusion method, terminal, storage medium and system under multi-layer association, which associates and identifies different accounts of natural users come out, and then integrate and determine the unique identifier to facilitate the push processing of subsequent related marketing activities.

In order to achieve the above-mentioned technical effect, the technical scheme of the present invention is as follows:

A user information fusion method under multi-layer association, the method at least includes:

S1. Select the data source of the user information to be integrated, each data source corresponds to a material table, determine the field in the material table that can identify the user and its corresponding channel type, and obtain the channel type information;

S2. Determine the vertices and edges according to the channel type information, build a user information graph based on the vertices and edges, and then split the user information graph into independent user connectivity graphs;

S3. Use the channel type association table to query the historical data of the entity unique ID corresponding to each vertex of the user connectivity graph, and obtain all entity unique IDs corresponding to the user connectivity graph. The relationship of the unique ID of each entity is used as an edge to construct a connected graph of the unique ID of the entity, thereby using the historical data to associate the connected graph of the user;

S4. Determine whether the unique ID of each user entity only appears in one user connectivity graph, if so, fuse user information on demand for each user connectivity graph; otherwise, construct the entity unique ID connectivity graph, and execute step S5;

S5. Determine the entity unique IDs connected together in the entity unique ID connected graph, and associate the user connected graphs corresponding to the connected entity unique IDs;

S6. Read the entity unique IDs corresponding to all vertices of the user connectivity graph, and deduplicate the entity unique IDs;

S7. Utilize the unique ID of the entity after deduplication processing to update the channel type association table.

In this technical solution, by constructing and splitting the user connectivity graph, multi-layer complex association relationships are supported instead of fixed simple association rules. Users only need to pay attention to the association of two material tables, and do not need to pay attention to all material tables How are they related, and supports the operation of new user information, based on the user connection graph, combined with the channel type association table (a table that can reflect the mapping relationship between the channel type value and the unique ID of the entity under historical data) , to further construct the entity unique ID connectivity graph, so that with the help of the new association provided by the new incremental data, the two independent user entities considered in the historical data can be merged together, and the scope of application is not limited to very specific scenarios and businesses Logic, but can sort out the channel type relationship according to the actual business, adapt to different scenarios, expand from user information fusion to product and place information fusion.

Preferably, in the customer data platform CDP of the enterprise, user information spreads across multiple data sources of the enterprise, each data source corresponds to a material table, and the material table includes: enterprise member information table, mall user information table, coupon Use the record table; the channel types in the material table include: user mobile phone number, member ID, Email, WeChat unionID, and WeChat openID. Records with the same channel type in different material tables correspond to the same natural user; the channel type field is for each combination of channel types and material tables corresponding to each channel type.

Preferably, "channel type + channel type field value" is used as the vertex attribute, and the vertex attribute is hashed to obtain the vertex ID. The vertex ID records the attribute of a natural user on a certain channel type;

Multiple "channel type field values" appearing in a certain record in the material table are used as edges, and the edges are connections between vertices, recording the association of a natural user on different channel types;

Connect vertices and edges to build a user information graph; then use the connected component algorithm to split the user information graph to obtain the minimum vertex ID of the user information graph where each vertex is located, group and aggregate the minimum vertex ID to obtain each independent user All vertices of the connected graph, that is, each independent user connected graph is obtained.

Here, "channel type + channel type field value" is used as the vertex attribute, and there is no need to record the information of the material table where it is located. It is simple and easy to operate. From the linear material table relationship to the two-dimensional plane graph relationship, it is convenient Multi-layer complex associations are supported.

Preferably, if the unique ID of each user entity only appears in one user connectivity graph, multiple natural users generated by newly added user information are also independent natural users in the historical material table, and the connectivity graph of each user is fused as needed User Info.

Preferably, the entity unique ID connectivity graph is constructed with each entity unique ID as a vertex, and the relationship between multiple entity unique IDs appearing in the same user connectivity graph as an edge, which greatly reduces the amount of data that needs to be involved in the calculation , so as to effectively improve the computational efficiency.

Preferably, the method for determining the entity unique IDs connected together in the entity unique ID connected graph described in step S5 is a connected component algorithm; after associating the user connected graphs corresponding to the connected entity unique IDs, each user connected graph corresponds to a natural user.

Preferably, when performing deduplication processing on the unique ID of the entity described in step S6, it includes:

a. The user connectivity graph has no entity unique ID: the current natural user is a new user, an entity unique ID is generated based on the user connectivity graph, and UUID is used to ensure its uniqueness;

b. The user connectivity graph only has one entity unique ID: the old and new user information data of the current natural user belong to the same natural user, and the old entity unique ID will be used;

c. There are multiple entity unique IDs in the user connectivity graph: Adding user information generates user information fusion, and only retains any one of the entity unique IDs.

The present invention proposes a terminal, including a processor, a memory, and a computer program stored in the memory, and the processor executes the computer program stored in the memory to realize the steps of the user information fusion method under multi-layer association .

The present invention proposes a computer storage medium, where computer program instructions are stored on the computer-readable storage medium, and when the instructions are executed by a processor, the steps of the user information fusion method under multi-layer association are realized.

The present invention also proposes a user information fusion system under multi-layer association, the system is used to realize the user information fusion method under multi-layer association, including:

The channel type information acquisition module selects the data source of the user information to be integrated, each data source corresponds to a material table, determines the field that can identify the user in the material table and its corresponding channel type, and obtains the channel type information;

The user connectivity graph construction module determines the vertices and edges based on the channel type information, constructs the user information graph based on the vertices and edges, and then splits the user information graph into independent user connectivity graphs;

The historical data association module uses the channel type association table to query the historical data of the unique ID of the entity corresponding to each vertex of the user connectivity graph, and obtains the unique IDs of all entities corresponding to the user connectivity graph. The relationship between the unique IDs of multiple entities in the network is used as an edge to construct a connected graph of entity unique IDs, so as to use historical data to associate user connected graphs;

Judging module, used to judge whether the unique ID of each user entity only appears in one user connectivity graph, if so, fuse user information on demand for each user connectivity graph; otherwise, construct the entity unique ID connectivity graph;

The user connected graph association module determines the entity unique IDs connected together in the entity unique ID connected graph, and associates the user connected graphs corresponding to the connected entity unique IDs;

The deduplication processing module reads the entity unique IDs corresponding to all vertices of the user connectivity graph, and performs deduplication processing on the entity unique IDs;

The update module uses the unique ID of the entity after deduplication to update the channel type association table.

Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

The present invention proposes a user information fusion method, terminal, storage medium and system under multi-layer association. Firstly, the data source of the user information to be integrated is selected. The data source corresponds to the material table, and further according to the association relationship in the material table Obtain channel type information, build a user information graph according to the channel type information, and then split the user information graph into independent user connectivity graphs, expanding from a linear material table relationship to a two-dimensional plane graph relationship, conveniently supporting multiple Layer complex association, in the complex association of a large amount of data, find the records belonging to the same natural user, which can be used for subsequent data processing, user label system construction, user operation statistical analysis, marketing activity push and other operations; based on the user connectivity graph In the above, combined with the historical material table, the unique ID connection graph of the entity is further constructed. With the help of the new association provided by the new user information data, the two independent user entities in the historical material table data are fused together, which greatly reduces the need to participate in the calculation. The amount of data can effectively improve the calculation efficiency.

Description of drawings

FIG. 1 shows a schematic flow diagram of a user information fusion method under multi-layer association proposed in Embodiment 1 of the present invention;

Fig. 2 shows the independent user connectivity graph that divides the user information graph into proposed in Embodiment 1 of the present invention;

Fig. 3 shows the user connectivity diagram corresponding to the historical data in the historical material table proposed in Embodiment 1 of the present invention;

Fig. 4 represents the schematic diagram of the unique ID of the user entity in the corresponding channel type association table for each vertex of the current user connectivity graph proposed in Embodiment 1 of the present invention;

Fig. 5 shows a schematic diagram of associating user connectivity graphs corresponding to the entity unique IDs proposed in Embodiment 1 of the present invention;

Fig. 6 shows the final user connectivity graph proposed in Embodiment 1 of the present invention;

FIG. 7 shows a schematic structural diagram of a user information fusion system under multi-layer association proposed in Embodiment 3 of the present invention.

Detailed ways

The accompanying drawings are for illustrative purposes only and cannot be construed as limiting the patent;

For those skilled in the art, it is understandable that some well-known content descriptions in the drawings may be omitted.

The technical solutions of the present invention will be further described below in conjunction with the accompanying drawings and embodiments.

The positional relationship described in the drawings is only for illustrative purposes, and cannot be construed as a limitation to this patent;

Example 1

As shown in Figure 1, in the present embodiment, at first a kind of user information fusion method under multi-layer association is proposed, and described method comprises:

S1. Select the data source of the user information to be integrated, each data source corresponds to a material table, determine the field in the material table that can identify the user and its corresponding channel type, and obtain the channel type information;

Taking the enterprise's customer data platform CDP as the background, user information spreads across multiple data sources of the enterprise, and each data source corresponds to a material table. The material table includes: enterprise member information table, mall user information table, and coupon usage records Table; the channel types in the material table include: user mobile phone number, member ID, Email, WeChat unionID and WeChat openID. Records with the same channel type in different material tables correspond to the same natural user; the channel type field is for each channel Combination of type and material table corresponding to each channel type. That is, first find out which data sources in the system need to integrate information, analyze the material tables corresponding to these data sources, and find the fields in which users can be identified; a certain value of this field corresponds to a user, but a user in this material This field of the table can have multiple values, such as coupon ID, a user can purchase and use multiple coupons, but a coupon can only be purchased and used by one user; such a field is a channel type field , such as coupon ID, email, mobile phone number, channel type indicates the scene where the user entity has a relationship with the outside world, and determines the channel type field that can be associated between the material table, and each such field set is recorded as a "channel type". For example, the "mobile phone number" field in the coupon use record table, the "mobile phone number" in the user information table, and other fields in the material table that represent the mobile phone number of the consumer user constitute a channel type. Similarly, there may also be channel types such as ID card and Email.

S2. Determine the vertices and edges according to the channel type information, build a user information graph based on the vertices and edges, and then split the user information graph into independent user connectivity graphs;

Like a graph in the sense of data structure, a user information graph is also composed of vertices and connections between vertices. According to the definition of channel type in step S1, records with the same channel type in different material tables should correspond to the same natural Users, such as the record with mobile phone number 13344445555 in the coupon use record table and the record with mobile phone number 13344445555 in the mall user information table belong to the same natural user, so the "channel type + channel type field value" is used as the Vertex attribute, hash the vertex attribute to get the vertex ID, which records the attribute of a natural user on a certain channel type; the hash calculation in this process is realized by the currently public MurMusHash algorithm, which is the Conventional technologies in the field will not be repeated here.

Multiple "channel type field values" appearing in a record in the material table are used as edges, and the edges are connections between vertices, which record the association of a natural user on different channel types; in addition, the edited The weighted value of can be the update time of this record, which can be used in subsequent user connectivity graphs to resolve conflicts.

Connect vertices and edges to build a user information graph. Ideally, an independent user connectivity graph corresponds to a unique natural user, and there are no connected edges with other user connectivity graphs. Use the connected component algorithm to split the user information graph , the connected component algorithm currently has a relatively mature programming language that can be directly applied, such as Spark GraphX, etc. The direct result of the calculation is the minimum vertex ID of the user information graph where each vertex is located, and the minimum vertex ID is grouped and aggregated, using The GROUP function can be implemented to obtain all the vertices of each independent user connectivity graph, that is, to obtain the information (vertices and corresponding edges) of each independent user connectivity graph.

S3. Use the channel type association table to query the historical data of the entity unique ID corresponding to each vertex of the user connectivity graph, and obtain all entity unique IDs corresponding to the user connectivity graph. The relationship of the unique ID of each entity is used as an edge to construct a connected graph of the unique ID of the entity, thereby using the historical data to associate the connected graph of the user;

This step is the process of correlating historical data. If the method proposed by the present invention is executed for the first time, there is no need to correlate history, and this step can be skipped. For the data source to be integrated with user information is a new computing task, it can be assumed that the mapping already exists. The user connection graph formed by it does not represent comprehensive user channel type information, and historical data needs to be associated to perform accurate user information fusion calculations;

S4. Determine whether the unique ID of each user entity only appears in one user connection graph, if so, fuse user information on demand for each user connection graph; otherwise, construct the entity unique ID connection graph, and execute step S5; if each user The unique ID of the entity only appears in one user connectivity graph, and the multiple natural users generated by the newly added user information are also independent natural users in the historical material table. Therefore, in the future, only the user information needs to be fused on demand for each user connectivity graph ; If multiple user connectivity graphs have the same user entity unique ID, the multiple natural users generated by the new user information data actually belong to the same natural user in the old data and can be merged. Multiple connected graphs of the unique ID of the entity are connected together, which can be realized by constructing a connected graph of the unique ID of the entity.

Among them, the entity unique ID connectivity graph is constructed with the unique ID of each entity as the vertex, and the relationship between multiple entity unique IDs in the same user connectivity graph as the edge, which greatly reduces the amount of data that needs to be involved in the calculation, thereby effectively improving Computational efficiency.

S5. Determine the entity unique IDs connected together in the entity unique ID connected graph, and associate the user connected graphs corresponding to the connected entity unique IDs;

S6. Read the entity unique IDs corresponding to all vertices of the user connected graph, and deduplicate the entity unique IDs; the method of determining the entity unique IDs connected together in the entity unique ID connected graph described in step S5 is also a connected component algorithm , belongs to a kind of graph algorithm, connectivity means that in an undirected graph, if there is a path from vertex v to vertex w, then v and w are said to be connected. If any two vertices in graph G are connected, then graph G is called connected graph, otherwise it is called disconnected graph. After associating the user connectivity graphs corresponding to the unique IDs of the connected entities, each user connectivity graph corresponds to a natural user;

When deduplicating the unique ID of an entity, it includes:

a. The user connectivity graph does not have an entity unique ID: the current natural user is a new user, an entity unique ID is generated based on the user connectivity graph, and UUID is used to ensure its uniqueness; UUID is the abbreviation of Universal Unique Identification Code, which is a relatively conventional Technology.

b. The user connectivity graph only has one entity unique ID: the old and new user information data of the current natural user belong to the same natural user, and the old entity unique ID will be used;

c. There are multiple entity unique IDs in the user connectivity graph: Adding user information generates user information fusion, and only retaining any one of the entity unique IDs, because the specific saving of which one has less impact on the calculation results, so the first one is generally saved.

S7. Utilize the unique ID of the entity after deduplication processing to update the channel type association table;

Specifically, traverse all vertices. The attributes of the vertices contain information about the channel type and its value, and store its relationship with the unique ID of the entity in the channel type association table. The table name of the association table is channel_entity_relation_<channel type>, such as channel_entity_relation _Mobile phone number, the structure is:

Channel type field value	Entity Unique ID
13344445555	33f5ec7daf694f9ca09c66767272d415
...	...

In summary, by constructing and splitting user connectivity graphs, multi-layer complex association relationships are supported instead of fixed simple association rules. Users only need to pay attention to the association between two material tables, and do not need to pay attention to the relationship between all material tables. How to associate, and support the operation of new user information, based on the user connectivity graph, combined with the historical material table (historical data), to further construct the entity unique ID connectivity graph, so that the new association provided by the new incremental data With the help of the historical data, the two independent user entities are considered to be fused together. The scope of application is not limited to very specific scenarios and business logic, but can sort out channel type relationships and adapt to different scenarios according to the actual business. Information fusion extends to information fusion of commodities and places.

The method proposed in this embodiment will be further described below in combination with specific implementation scenarios. Taking the CDP system as an example, the scenario is simplified as follows: the user receives coupons in the membership system, and then uses the coupons to shop in JD.com.

Assume that the data source only contains 3 material tables: coupon usage record table, JD user information table and internal member information table.

Analyze the association relationship of the above material table, and express it in the form of channel type and channel type field, as shown in Table 1.

Table 1

Assume that the coupon use record table has the records shown in Table 2.

Table 2

Phone number	Coupon number	Member ID	...
13344445555	T00003	C001	the
13455556666	T00004	C001	the
13344445555	T00005	C001	the
13566667777	T00006	C002	the

The Jingdong user information table has the records shown in Table 3.

table 3

JD User ID	Phone number	...
JD001	13344445555	the
JD002	13455556666	the
JD003	13566667777	the

The internal member information table has the records shown in Table 4.

Table 4

Member ID

Email

...

C001	c001@test.com	the
C002	c002@test.com	the

All channel types are: "Mobile Phone Number", "Member ID" and "JD User ID".

According to the collated channel type information, a user information graph is constructed and split into independent user connectivity graphs. The vertices of the user information graph are calculated according to the information in the material table, as shown in Table 5. Here, for the convenience of illustration and to improve the recognition of the examples, a relatively simple value is used to refer to the actual vertex attribute hash value.

table 5

Vertex ID	vertex attributes
1111	[mobile phone, 13344445555]
2222	[mobile phone, 13455556666]
3333	[Mobile phone_13566667777]
4444	[Member ID_, C001]
5555	[Member ID_, C002]
6666	[JD user ID, JD001]
7777	[JD user ID, JD002]
8888	[JD user ID, JD003]

Use the connected component algorithm to calculate the minimum vertex ID of the connected graph where each vertex is located. The calculation results are shown in Table 6:

Table 6

Vertex ID	The minimum vertex ID of the connected graph	vertex attributes
1111	1111	[mobile phone, 13344445555]
2222	1111	[mobile phone, 13455556666]
3333	3333	[mobile phone, 13566667777]
4444	1111	[Member ID, C001]
5555	3333	[Member ID, C002]
6666	1111	[JD user ID, JD001]
7777	1111	[JD user ID, JD002]
8888	3333	[JD user ID, JD003]

According to the above results, the user connectivity graph is shown in Figure 2. It can be seen from Figure 2 that, according to the "minimum vertex ID of the connected graph" column, the entire user information graph is split into two independent user connectivity graphs, corresponding to Two natural users, including user connectivity graph 1 and user connectivity graph 2.

Assuming that this calculation is not the first calculation, the historical data in the historical material table has been calculated, and the channel type association table is generated. The historical data is the coupon use record table with the records shown in Table 7:

Table 7

Phone number	Coupon number	Member ID	...
13566667777	T00001	C001	the
13344445555	T00002	C002	the

The calculation result of the previous execution is shown in Figure 3, and two unique IDs for natural use corresponding to the records shown in Table 7 are obtained, so the contents of the "mobile phone number" channel type association table are shown in Table 8.

Table 8

Phone number	Entity Unique ID
13566667777	5109b4d4e0e7411c992da7d5ea112538
13344445555	73c823904daa4c20a4c2035132d0e7c2

The content of the channel type association table of "member ID" is shown in Table 9.

Table 9

Member ID	Entity Unique ID
C001	5109b4d4e0e7411c992da7d5ea112538
C002	73c823904daa4c20a4c2035132d0e7c2

At this time, all the vertex attributes of each user's connected graph are queried in the corresponding channel type association table to find the entity's unique ID and recorded as the attribute of the user's connected graph, as shown in Figure 4. According to the entity's unique ID, the entity's unique ID connected graph is constructed, and the corresponding user The connectivity graph generates connections, and a new user connectivity graph is obtained, as shown in Figure 5.

After associating the historical data in the historical material table, since some vertices of the two new user connectivity graphs belong to the same user connectivity graph in the historical data, the two can be associated together, as shown in Figure 5, all current data It constitutes the same natural user, and the entity unique ID owned by it contains two after deduplication:

5109b4d4e0e7411c992da7d5ea112538

73c823904daa4c20a4c2035132d0e7c2

It is considered that in this calculation, due to the new user coupon use record, the old user information is related to two independent users, and they are the same user, and user information fusion has occurred.

Only one unique user ID is reserved at this time, such as reserved: 5109b4d4e0e7411c992da7d5ea112538, as shown in Figure 6.

Finally, update the channel type association table to facilitate subsequent calculations. The updated channel type association table includes the "mobile phone number" channel type association table shown in Table 10, the "member ID" channel type association table shown in Table 11, and the "JD user ID" channel type association table shown in Table 12.

Thereafter, the use of CDP data is based on the unique ID of the entity. For example, to calculate the user portrait, calculate the unique ID of the entity, and get the user (entity) portrait that is the unique ID of each entity; if you want to do marketing activities in the membership system, you can filter the crowd according to the user portrait, and get the corresponding user portrait group A series of entity unique IDs, and then according to the JD user ID channel association table, the corresponding JD user ID can be found, so as to perform marketing campaign push and other touch operations.

Example 2

The present invention proposes a terminal, including a processor, a memory, and a computer program stored on the memory, and the processor executes the computer program stored on the memory to realize the user information fusion under the multi-layer association described in Embodiment 1 The steps of the method, wherein the memory can be a magnetic disk, flash memory or any other non-volatile storage medium, and the processor is connected to the memory, and can be implemented as one or more integrated circuits, specifically, it can be a microprocessor or a microcontroller , when executing the computer program stored on the memory, the fusion of user information under multi-layer association is realized.

The present invention proposes a computer storage medium. The computer readable storage medium stores computer program instructions. When the instructions are executed by a processor, the steps of the user information fusion method under multi-layer association described in Embodiment 1 are realized.

Example 2

Referring to Fig. 7, the present invention also proposes a user information fusion system under multi-layer association, the system is used to implement the user information fusion method under multi-layer association described in Embodiment 1, including:

The channel type information acquisition module selects the data source of the user information to be integrated, each data source corresponds to a material table, determines the channel type and the channel type field identifying the association relationship in the material table, and obtains the channel type information;

The user connectivity graph construction module determines the vertices and edges based on the channel type information, constructs the user information graph based on the vertices and edges, and then splits the user information graph into independent user connectivity graphs;

The historical data association module is used to collect the historical material table, determine the mapping between each channel type in the historical material table and the unique ID of the user entity, form an initial channel type association table, and further search for the channel type corresponding to each vertex of the user connectivity graph The unique ID of the user entity in the association table;

Judging module, used to judge whether the unique ID of each user entity only appears in one user connectivity graph, if so, fuse user information on demand for each user connectivity graph; otherwise, construct the entity unique ID connectivity graph;

The user connected graph association module determines the entity unique IDs connected together in the entity unique ID connected graph, and associates the user connected graphs corresponding to the connected entity unique IDs;

The deduplication processing module reads the entity unique IDs corresponding to all vertices of the user connectivity graph, and performs deduplication processing on the entity unique IDs;

The update module uses the unique ID of the entity after deduplication to update the channel type association table.

Apparently, the above-mentioned embodiments of the present invention are only examples for clearly illustrating the present invention, rather than limiting the implementation of the present invention. For those of ordinary skill in the art, other changes or changes in different forms can be made on the basis of the above description. It is not necessary and impossible to exhaustively list all the implementation manners here. All modifications, equivalent replacements and improvements made within the spirit and principles of the present invention shall be included within the protection scope of the claims of the present invention.

Claims (10)

1. A user information fusion method under multi-layer association, characterized in that the method at least includes:

   S1. Select the data source of the user information to be integrated, each data source corresponds to a material table, determine the field in the material table that can identify the user and its corresponding channel type, and obtain the channel type information;

   S2. Determine the vertices and edges according to the channel type information, build a user information graph based on the vertices and edges, and then split the user information graph into independent user connectivity graphs;

   S3. Use the channel type association table to query the historical data of the entity unique ID corresponding to each vertex of the user connectivity graph, and obtain all entity unique IDs corresponding to the user connectivity graph. The relationship of the unique ID of each entity is used as an edge to construct a connected graph of the unique ID of the entity, so as to use the historical data to associate the connected graph of the user;

   S4. Determine whether the unique ID of each user entity only appears in one user connectivity graph, if so, fuse user information on demand for each user connectivity graph; otherwise, construct the entity unique ID connectivity graph, and execute step S5;

   S5. Determine the entity unique IDs connected together in the entity unique ID connected graph, and associate the user connected graphs corresponding to the connected entity unique IDs;

   S6. Read the entity unique IDs corresponding to all vertices of the user connectivity graph, and deduplicate the entity unique IDs;

   S7. Utilize the unique ID of the entity after deduplication processing to update the channel type association table.
2. The user information fusion method under multi-layer association according to claim 1 is characterized in that, in the customer data platform CDP of the enterprise, user information spreads over a plurality of data sources of the enterprise, and each data source corresponds to a material list, so The material table described includes: enterprise membership information table, mall user information table, coupon use record table; the channel types in the material table include: user mobile phone number, member ID, Email, WeChat unionID and WeChat openID, and the channel types are in different material tables Records with the same value correspond to the same natural user; the channel type field is a combination of each channel type and the material table corresponding to each channel type.
3. The user information fusion method under multi-layer association according to claim 2, characterized in that in step S2, "channel type + channel type field value" is used as the vertex attribute, and the vertex attribute is hashed to obtain the vertex ID, the vertex ID records the attributes of a natural user on a certain channel type;

   Multiple "channel type field values" appearing in a certain record in the material table are used as edges, and the edges are connections between vertices, recording the association of a natural user on different channel types;

   Connect vertices and edges to build a user information graph; then use the connected component algorithm to split the user information graph to obtain the minimum vertex ID of the user information graph where each vertex is located, group and aggregate the minimum vertex ID to obtain each independent user All vertices of the connected graph, that is, each independent user connected graph is obtained.
4. The user information fusion method under multi-layer association according to claim 3, characterized in that, in step S4, if the unique ID of each user entity only appears in one user connection graph, then the multiple generated by the newly added user information A natural user is also an independent natural user in the historical material table, and user information is fused on-demand for each user connectivity graph.
5. The user information fusion method under multi-layer association according to claim 4, wherein the entity unique ID connectivity graph uses the unique ID of each entity as a vertex, and multiple entities appear in the same user connectivity graph Unique ID relationships are constructed as edges.
6. The user information fusion method under multi-layer association according to claim 5, wherein the method for determining the entity unique IDs connected together in the entity unique ID connected graph in step S5 is a connected component algorithm; After the user connectivity graph corresponding to the unique ID of the entity is associated, each user connectivity graph corresponds to a natural user.
7. The user information fusion method under multi-layer association according to claim 6, characterized in that, when performing deduplication processing on the unique ID of the entity described in step S6, it includes:

   a. The user connectivity graph does not have an entity unique ID: the current natural user is a new user, an entity unique ID is generated based on the user connectivity graph, and UUID is used to ensure its uniqueness;

   b. The user connectivity graph only has one entity unique ID: the old and new user information data of the current natural user belong to the same natural user, and the old entity unique ID will be used;

   c. There are multiple entity unique IDs in the user connectivity graph: Adding user information generates user information fusion, and only retains any one of the entity unique IDs.
8. A terminal, characterized in that it includes a processor, a memory, and a computer program stored in the memory, and the processor executes the computer program stored in the memory to realize the multi-layered system described in any one of claims 1-7. The steps of the user information fusion method under association.
9. A computer storage medium, characterized in that computer program instructions are stored on the computer-readable storage medium, and when the instructions are executed by a processor, the user under the multi-layer association described in any one of claims 1 to 7 is realized. Steps of the information fusion method.
10. A user information fusion system under multi-layer association, the system is used to realize the user information fusion method under multi-layer association according to claim 1, characterized in that it includes:

    The channel type information acquisition module selects the data source of the user information to be integrated, each data source corresponds to a material table, determines the field that can identify the user in the material table and its corresponding channel type, and obtains the channel type information;

    The user connectivity graph construction module determines the vertices and edges based on the channel type information, constructs the user information graph based on the vertices and edges, and then splits the user information graph into independent user connectivity graphs;

    The historical data association module uses the channel type association table to query the historical data of the unique ID of the entity corresponding to each vertex of the user connectivity graph, and obtains the unique IDs of all entities corresponding to the user connectivity graph. The relationship between the unique IDs of multiple entities in the network is used as an edge to construct a connected graph of entity unique IDs, so as to use historical data to associate user connected graphs;

    Judging module, used to judge whether the unique ID of each user entity only appears in one user connectivity graph, if so, fuse user information on demand for each user connectivity graph; otherwise, construct the entity unique ID connectivity graph;

    The user connected graph association module determines the entity unique IDs connected together in the entity unique ID connected graph, and associates the user connected graphs corresponding to the connected entity unique IDs;

    The deduplication processing module reads the entity unique IDs corresponding to all vertices of the user connectivity graph, and performs deduplication processing on the entity unique IDs;

    The update module uses the unique ID of the entity after deduplication to update the channel type association table.

Images