Edge data increasing method and device, medium and equipment

Abstract

Embodiments of the present specification provide a method and device for increasing data, a medium and equipment, the method comprising: splitting each customer address in the full amount of customers of a graph to obtain regional information and address information of the customer address; generating a first matrix according to each address information corresponding to the same regional information; reducing the dimension of the first matrix to obtain a second matrix, and classifying each address information corresponding to the same regional information according to the second matrix, and the address information in the same class has similarity; calculating the similarity of each two address information in the same class, and selecting an address information pair with a similarity greater than a preset threshold; for each address information pair selected, determining two nodes corresponding to the two customers of the address information pair in the graph, and adding a relationship edge between the two determined nodes. The embodiments of the present application can enrich the graph data.

Description

Technical Field

[0001] This specification relates to one or more embodiments in the field of network communication, and more particularly to a method, apparatus, medium, and device for side data augmentation. Background Technology

[0002] Graph computing is widely used in the auditing of suspicious transactions. It can also be used to uncover suspicious groups. The graph used in graph computing includes two types of data: nodes and edges. Nodes typically include types such as customers and bank cards, while edges typically contain records of transactions between customers, multiple customers using the same account, etc. The richer the data in the graph, the more complex the group relationships can be uncovered. Summary of the Invention

[0003] This specification describes one or more embodiments of a method, apparatus, medium, and device for enriching edge data, which can enrich spectral data.

[0004] According to a first aspect, the edge data augmentation method provided in the embodiments of the present invention includes:

[0005] Each customer address in the full customer database is broken down to obtain the geographic information and address information of that customer address; wherein, the address information corresponds one-to-one with the customer.

[0006] A first matrix is ​​generated based on the address information corresponding to the same geographical information; wherein, the element in the i-th row and j-th column of the first matrix indicates whether the j-th address information corresponding to the same geographical information contains the i-th string in the string set corresponding to that geographical information; the number of columns in the first matrix is ​​the number of address information corresponding to the same geographical information, and the number of rows in the first matrix is ​​the number of strings in the string set;

[0007] The first matrix is ​​reduced in dimensionality to obtain a second matrix. Based on the second matrix, the address information corresponding to the same regional information is classified, and the address information of the same category has similarity.

[0008] Calculate the similarity between any two address information in the same category, and select address information pairs with a similarity greater than a preset threshold.

[0009] For each selected pair of address information, determine the two nodes in the graph corresponding to the two customers of the pair of address information, and add a relationship edge between the two nodes.

[0010] According to a second aspect, the edge data augmentation apparatus provided in the embodiments of the present invention includes:

[0011] The address splitting module is used to split the address of each customer in the full customer map to obtain the geographic information and address information of that customer address; wherein, the address information corresponds one-to-one with the customer;

[0012] A matrix generation module is used to generate a first matrix based on the address information corresponding to the same geographical information; wherein, the element in the i-th row and j-th column of the first matrix indicates whether the j-th address information corresponding to the same geographical information contains the i-th string in the string set corresponding to that geographical information; the number of columns in the first matrix is ​​the number of address information corresponding to the same geographical information, and the number of rows in the first matrix is ​​the number of strings in the string set;

[0013] The dimensionality reduction and classification module is used to reduce the dimensionality of the first matrix to obtain a second matrix, and classify the address information corresponding to the same regional information according to the second matrix, where address information of the same category has similarity.

[0014] The similarity calculation module is used to calculate the similarity between any two address information in the same category and select address information pairs with a similarity greater than a preset threshold.

[0015] The edge-adding module is used to determine, for each selected pair of address information, the two nodes corresponding to the two customers in the graph, and add a relationship edge between the two determined nodes.

[0016] According to a third aspect, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the method provided in the first aspect.

[0017] According to a fourth aspect, a computing device is provided, including a memory and a processor, wherein the memory stores executable code, and the processor, when executing the executable code, implements the method provided in the first aspect.

[0018] The edge data augmentation methods, apparatuses, media, and devices provided in the embodiments of this specification, individually or in combination, have the following beneficial effects:

[0019] (1) Customer addresses are split to obtain regional information and address information. A first matrix is ​​generated based on the address information corresponding to the same regional information. The first matrix is ​​then dimensionality-reduced to obtain a second matrix. The second matrix is ​​then used for classification, thereby reducing the problem of calculating the number of customers from the entire set to the problem of calculating customers of the same class within the same regional information. This greatly reduces the computation time and computational load, and improves computational efficiency. Furthermore, the similarity between every two address information within the same class is calculated. An edge is added between the two nodes corresponding to two customers with high similarity, thus increasing the edge data and enabling the calculation of address similarity for massive numbers of customers to be used in actual production environments. It can be seen that this embodiment of the invention determines the necessary relationship edges to be added based on address similarity, thereby enriching the edge data used in graph computation.

[0020] (2) In one embodiment, the first matrix is ​​reduced in dimensionality using the minimum hash algorithm to obtain the second matrix. The minimum hash algorithm is a technique for quickly determining whether two objects are similar. It can map data objects in a high-dimensional set to a smaller set of hash values ​​while preserving the similarity between objects. It can be seen that using the minimum hash algorithm to reduce dimensionality will not affect the subsequent similarity calculation.

[0021] (3) In one embodiment, a locality-sensitive hash function is used for bucketing, and the bucketing results are then used to classify the address information. Based on the characteristics of locality-sensitive hashing, if multiple address information is assigned to the same bucket number at any one time, then these multiple address information are highly similar in probability. Therefore, based on the bucketing results, the problem of calculating all customers can be reduced to the problem of calculating customers of the same type in the same region, thereby further improving the computational efficiency. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating a method for adding edge data in one embodiment of this specification;

[0024] Figure 2 This is a structural block diagram of a side data addition device in one embodiment of this specification. Detailed Implementation

[0025] The solution provided in this specification will now be described with reference to the accompanying drawings.

[0026] As mentioned in the background section, the richer the data in the graph, the more complex the group relationships can be uncovered; in other words, the richer the data in the graph, the more beneficial it is for group discovery. The richness of edge data in the graph is particularly important. However, identifying customer pairs with highly similar address information from the entire customer base is a technically challenging project. Performing pairwise comparisons across the entire customer base is extremely time-consuming, and the computational complexity makes it impossible to complete in scenarios with hundreds of millions of customers.

[0027] Therefore, according to a first aspect, embodiments of this specification provide a method for adding edge data, the method comprising: splitting each customer address in the full customer set of the graph to obtain the geographic information and address information of the customer address; wherein the address information corresponds one-to-one with the customer; generating a first matrix based on the address information corresponding to the same geographic information; wherein the element in the i-th row and j-th column of the first matrix indicates whether the j-th address information corresponding to the same geographic information contains the i-th string in the string set corresponding to the geographic information; the number of columns in the first matrix is ​​equal to the number of columns corresponding to the same geographic information. The number of each address information is determined, and the number of rows in the first matrix is ​​the number of strings in the string set. The first matrix is ​​reduced in dimensionality to obtain a second matrix. Based on the second matrix, the address information corresponding to the same regional information is classified, and address information of the same category has similarity. The similarity of each pair of address information in the same category is calculated, and address information pairs with similarity greater than a preset threshold are selected. For each selected address information pair, the two nodes corresponding to the two customers in the graph are determined, and a relationship edge is added between the two determined nodes.

[0028] The specific implementation of the above concept is described below.

[0029] Figure 1 This is a flowchart illustrating a method for adding edge data in one embodiment of the present invention. It is understood that this method can be executed by any device, equipment, platform, or cluster of devices with computing and processing capabilities. See also... Figure 1 The method for adding edge data includes the following steps S102 to S110:

[0030] S102. Each customer address in the full customer data map is split to obtain the geographic information and address information of that customer address; wherein, the address information corresponds one-to-one with the customer.

[0031] S104. Generate a first matrix based on the address information corresponding to the same regional information; wherein, the element in the i-th row and j-th column of the first matrix indicates whether the j-th address information corresponding to the same regional information contains the i-th string in the string set corresponding to that regional information; the number of columns in the first matrix is ​​the number of address information corresponding to the same regional information, and the number of rows in the first matrix is ​​the number of strings in the string set;

[0032] S106. Reduce the dimensionality of the first matrix to obtain the second matrix, and classify the address information corresponding to the same regional information according to the second matrix. Address information of the same category has similarity.

[0033] S108. Calculate the similarity between any two address information in the same category, and select the address information pairs with a similarity greater than a preset threshold.

[0034] S110. For each selected pair of address information, determine the two nodes corresponding to the two customers in the graph, and add a relationship edge between the two determined nodes.

[0035] exist Figure 1 The illustrated method for adding edge data involves splitting customer addresses to obtain geographic information and address information. A first matrix is ​​generated based on the address information corresponding to the same geographic information. This first matrix is ​​then dimensionality-reduced to obtain a second matrix, which is used for classification. This reduces the problem of calculating the number of customers from the entire dataset to the problem of calculating the number of customers within the same geographic information group, significantly reducing computation time and computational load and improving computational efficiency. Furthermore, the similarity between every two address information entries within the same group is calculated. An edge is added between the two nodes corresponding to two customers with high similarity, thus increasing the edge data and enabling the calculation of address similarity for massive numbers of customers to be used in real-world production environments. Therefore, this embodiment of the invention uses address similarity to determine the necessary addition of relational edges, thereby enriching the edge data used in graph computation.

[0036] The following description Figure 1 The execution method of each step.

[0037] S102. Each customer address in the full customer data map is split to obtain the geographic information and address information of that customer address; wherein, the address information corresponds one-to-one with the customer.

[0038] This includes geographical information, such as a country, province, city, or district. Address information refers to the specific address within the district.

[0039] In one embodiment, S102 may specifically include: splitting each customer address in the full customer pool according to the geographical level to obtain geographical information at each level and the address information corresponding to the lowest level of geographical information.

[0040] In other words, by breaking down a customer's address according to geographic level, we can obtain geographic information at each level and the address information corresponding to the lowest level of geographic information.

[0041] For example, Ant C Space on Tianfu Fourth Street, High-tech Zone, Chengdu, Sichuan Province, China, after being split, yields the following geographic information: China, Sichuan Province, Chengdu City, High-tech Zone, and the address information is Ant C Space on Tianfu Fourth Street. Subsequent steps target the individual address information corresponding to the same geographic location. For instance, in the map, the geographic information is all addresses within the High-tech Zone, Chengdu, Sichuan Province, China. This method reduces the computational problem for all customers to the computational problem for customers within the same geographic location, thus reducing computation time and computational load.

[0042] In this system, one address corresponds to one customer, meaning there is a one-to-one correspondence between address information and customers.

[0043] S104. Generate a first matrix based on the address information corresponding to the same regional information; wherein, the element in the i-th row and j-th column of the first matrix indicates whether the j-th address information corresponding to the same regional information contains the i-th string in the string set corresponding to that regional information; the number of columns in the first matrix is ​​the number of address information corresponding to the same regional information, and the number of rows in the first matrix is ​​the number of strings in the string set;

[0044] For example, a geographical location corresponds to a string set; for instance, "Chengdu High-tech Zone, Sichuan Province, China" corresponds to a string set. For each address in "Chengdu High-tech Zone, Sichuan Province, China", a first matrix is ​​generated. The value of the element in the i-th row and j-th column of the first matrix indicates whether the j-th address in "Chengdu High-tech Zone, Sichuan Province, China" contains the i-th string in the corresponding string set.

[0045] For example, if the j-th address information of Chengdu High-tech Zone, Sichuan Province, China contains the ith string in the corresponding string set, then the value of the element in the ith row and j-th column of the first matrix is ​​1. If the j-th address information of Chengdu High-tech Zone, Sichuan Province, China does not contain the ith string in the corresponding string set, then the value of the element in the ith row and j-th column of the first matrix is ​​0.

[0046] The first matrix has the number of columns corresponding to the same geographical information, for example, N represents the number of customers in Chengdu High-tech Zone, Sichuan Province, China. The first matrix has the number of rows corresponding to the number of strings in the string set, for example, the string set contains M strings.

[0047] For example, the steps for generating the first matrix are as follows: if the j-th address information corresponding to the same geographical information does not contain the ith string in the string set, then the element in the ith row and j-th column of the first matrix is ​​set to 0; if the j-th address information corresponding to the same geographical information contains the ith string in the string set, then the element in the ith row and j-th column of the first matrix is ​​set to 1.

[0048] S104 can convert the address information in string form into a first matrix composed of 0s and 1s. The first matrix is ​​a sparse matrix.

[0049] In one embodiment, the method for generating the string set may include the following steps:

[0050] Each address in the address information corresponding to the same geographical location is segmented to obtain multiple strings corresponding to that address information; wherein each string includes characters of a preset length;

[0051] The strings corresponding to each address information with the same regional information are used to form a string set corresponding to that regional information.

[0052] For example, the process of forming a string set corresponding to Chengdu High-tech Zone in Sichuan Province, China, includes: segmenting each address information in Chengdu High-tech Zone, Sichuan Province, to obtain multiple strings corresponding to that address information. For example, the address information "Tianfu Fourth Street Ant C Space" is segmented into the strings "Tian", "Fu", "Si", "Jie", "Ma", "Yi", "C", "Kong", and "Jian". Each string contains only one character, meaning each string has a length of 1. In practical scenarios, the string length can also be set to other values ​​such as 2 or 3. After segmenting the address information in Chengdu High-tech Zone, Sichuan Province, China, the strings corresponding to each address information are combined into a string set. The string set formed in this way is specific to Chengdu High-tech Zone, Sichuan Province, China.

[0053] Furthermore, forming the string set by combining the strings corresponding to the various address information that share the same geographical information may include:

[0054] Each string corresponding to the same geographical information of each address is added to an initially empty string set. Before adding the string corresponding to each address to the string set, it is determined whether the string corresponding to the address does not exist in the string set. If so, the string corresponding to the address is added to the string set to form a string set that does not contain the same string.

[0055] In other words, duplicate strings are not included in the resulting string set to improve the accuracy of subsequent calculations.

[0056] S106. Reduce the dimensionality of the first matrix to obtain the second matrix, and classify the address information corresponding to the same regional information according to the second matrix. Address information of the same category has similarity.

[0057] Understandably, if the first matrix is ​​used directly for calculation, the computational efficiency will still be relatively low. Therefore, it is necessary to reduce the dimensionality of the first matrix to obtain the second matrix. Using the second matrix after dimensionality reduction for calculation can further improve the computational efficiency.

[0058] In one embodiment, reducing the dimensionality of the first matrix to obtain the second matrix may specifically include: using a minimum hash algorithm to reduce the dimensionality of the first matrix to obtain the second matrix; wherein, the number of elements in the column vector corresponding to each address information in the second matrix is ​​less than the number of elements in the column vector corresponding to the address information in the first matrix.

[0059] Among them, the minimum hash algorithm, also known as the MinHash algorithm, is a technique for quickly determining whether two objects are similar. It can map data objects in a high-dimensional set to a smaller set of hash values ​​while preserving the similarity between objects. It can be seen that using the minimum hash algorithm to reduce dimensionality will not affect the subsequent similarity calculation.

[0060] In one embodiment, classifying the address information corresponding to the same geographical information may specifically include:

[0061] Select at least one locality-sensitive hash function;

[0062] Each locality-sensitive hash function is used to perform hash calculation on the column vector corresponding to each address information in the second matrix to obtain the bucket number corresponding to the address information under the locality-sensitive hash function. The address information is then assigned to the hash bucket corresponding to the bucket number. Each hash bucket corresponds to a sequence, which includes the set of address information assigned to the hash bucket under the locality-sensitive hash function.

[0063] The sequences that have at least one common address information in each sequence corresponding to each hash bucket under each locality sensitive hash function are merged into one category.

[0064] Locality Sensitive Hashing (LSH) is an algorithm that hashes an input into a fixed-length string digest. When the inputs are similar, the calculated digests also have similarity.

[0065] The number of locality-sensitive hash functions can be H, and the size of H can be configured.

[0066] In other words, the first of H locality-sensitive hash functions (LSH functions) is used to hash the column vector corresponding to each address in the second matrix, resulting in a bucket number. This bucket number indicates which hash bucket the address is assigned to. For example, if there are 5 hash buckets and the bucket number corresponding to an address is 3, then that address is assigned to the 3rd hash bucket. By hashing each address using the first LSH function, the bucket number corresponding to each address under the calculation of the first LSH function can be obtained, and then each address is assigned to the hash bucket corresponding to its respective bucket number. Each hash bucket corresponds to a sequence containing the set of address information assigned to that hash bucket under the calculation of the first LSH function. For example, under the calculation of the first LSH function, the sequence corresponding to the 1st hash bucket is {address 1, address 3}.

[0067] Similarly, the second of the H locality-sensitive hash functions is used to hash the column vectors corresponding to each address in the second matrix, obtaining the bucket number for each address. Then, each address is assigned to the hash bucket corresponding to its respective bucket number. This process continues until the Hth locality-sensitive hash function is used to hash the column vectors corresponding to each address in the second matrix, obtaining the bucket number for each address, and then assigning each address to the hash bucket corresponding to its respective bucket number.

[0068] Assuming H is 2, after two locality-sensitive hash functions (LSH) calculations and two bucketing operations, it is found that under the first LSH calculation, the sequence corresponding to the first hash bucket is {address information 1, address information 3}. Under the second LSH calculation, the sequence corresponding to the second hash bucket is {address information 1, address information 2}. This indicates that there is a common address information, address information 1, between these two hash bucket sequences. Therefore, address information 1, address information 2, and address information 3 are grouped into one category, resulting in {address information 1, address information 2, address information 3}. Other categories are determined in the same way.

[0069] As can be seen, based on the characteristics of locality-sensitive hashing, as long as multiple address information is assigned to the same bucket number once, then these multiple address information are highly similar with a high probability. Therefore, based on the bucketing situation, the problem of calculating all customers can be reduced to the problem of calculating customers with the same geographical information and the same category.

[0070] S108. Calculate the similarity between any two address information in the same category, and select the address information pairs with a similarity greater than a preset threshold.

[0071] In other words, if the similarity between two address information in the same category is greater than a preset threshold, it means that the two address information have high text similarity, so the two address information are selected as an address information pair.

[0072] In one embodiment, calculating the similarity between every two address information in the same category may specifically include: calculating the Jacobian similarity between every two address information in the same category.

[0073] Among them, Jaccard Similarity is a statistic used to compare the similarity and diversity of sample sets. It can measure the similarity of a finite set of samples and is defined as the ratio between the size of the intersection and the size of the union of two sets.

[0074] That is, Jacobi similarity is used as the textual similarity between two address information.

[0075] S110. For each selected pair of address information, determine the two nodes corresponding to the two customers in the graph, and add a relationship edge between the two determined nodes.

[0076] In other words, each address information pair selected from S108 corresponds to a customer pair. The two nodes corresponding to this customer pair are found in the graph, and an edge is added between these two nodes. Jacobi similarity can be added to the edge to represent the specific text similarity. After adding edge data to the graph, the updated graph can be used for graph computations such as gang detection. Graph computation refers to the process of using a graph as a data model to express and solve problems.

[0077] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0078] According to the second aspect, see Figure 2 This invention provides a side data augmentation device 200, comprising:

[0079] Address splitting module 202 is used to split each customer address in the full customer map to obtain the geographic information and address information of the customer address; wherein, the address information corresponds one-to-one with the customer;

[0080] The matrix generation module 204 is used to generate a first matrix based on the address information corresponding to the same regional information; wherein, the element in the i-th row and j-th column of the first matrix indicates whether the j-th address information corresponding to the same regional information contains the i-th string in the string set corresponding to that regional information; the number of columns in the first matrix is ​​the number of address information corresponding to the same regional information, and the number of rows in the first matrix is ​​the number of strings in the string set;

[0081] The dimensionality reduction and classification module 206 is used to reduce the dimensionality of the first matrix to obtain a second matrix, and classify the address information corresponding to the same regional information according to the second matrix, wherein the address information of the same category has similarity.

[0082] The similarity calculation module 208 is used to calculate the similarity between every two address information in each class and select address information pairs with a similarity greater than a preset threshold.

[0083] The edge-adding module 210 is used to determine, for each selected address information pair, the two nodes corresponding to the two customers in the graph, and add a relationship edge between the two determined nodes.

[0084] In one embodiment, the address splitting module is specifically used to: split each customer address in the full customer pool according to the geographic level, to obtain geographic information at each level and the address information corresponding to the lowest level geographic information.

[0085] In one embodiment, a set-forming module is further included for generating a string set, the set-forming module comprising:

[0086] The address segmentation unit is used to segment each address information corresponding to the same geographical area information to obtain multiple strings corresponding to that address information; wherein each string includes characters of a preset length;

[0087] The set forming unit is used to form a string set corresponding to the same regional information by combining the strings corresponding to each address information.

[0088] Furthermore, the set forming unit is specifically used to: add each string of each address information corresponding to the same regional information to an initially empty string set, and before adding the string corresponding to each address information to the string set, determine whether the string corresponding to the address information does not exist in the string set; if so, add the string corresponding to the address information to the string set to form the string set that does not contain the same string.

[0089] In one embodiment, the matrix generation module is specifically used to: reduce the dimensionality of the first matrix using a minimum hash algorithm to obtain a second matrix; wherein the number of elements in the column vector corresponding to each address information in the second matrix is ​​less than the number of elements in the column vector corresponding to the address information in the first matrix.

[0090] In one embodiment, the dimensionality reduction classification module is specifically used to: select at least one locality-sensitive hash function; perform hash calculation on the column vector corresponding to each address information in the second matrix using each locality-sensitive hash function to obtain the bucket number corresponding to the address information under the locality-sensitive hash function, and classify the address information into the hash bucket corresponding to the bucket number, each hash bucket corresponding to a sequence, the sequence including the set of address information classified into the hash bucket under the locality-sensitive hash function; and merge the sequences with at least one identical address information in each sequence corresponding to each hash bucket under each locality-sensitive hash function into a classification.

[0091] In one embodiment, the similarity calculation module is specifically used to: calculate the Jacobian similarity between every two address pieces of address information of the same class.

[0092] It is understood that explanations, specific implementation methods, beneficial effects, examples, etc. of the contents of the apparatus provided in the embodiments of the present invention can be found in the corresponding parts of the method provided in the first aspect, and will not be repeated here.

[0093] According to a third aspect, one embodiment of this specification provides a computer-readable storage medium having a computer program stored thereon that, when executed in a computer, causes the computer to perform the methods of any embodiment of the specification.

[0094] Specifically, a system or apparatus equipped with a storage medium may be provided, on which software program code implementing the functions of any of the embodiments described above is stored, and the computer (or CPU or MPU) of the system or apparatus may read and execute the program code stored in the storage medium.

[0095] In this case, the program code read from the storage medium can itself implement the function of any of the above embodiments, and therefore the program code and the storage medium storing the program code constitute part of the present invention.

[0096] Furthermore, it should be clear that not only can the program code read by the computer be executed, but also the operating system or other components operating on the computer can be instructed based on the program code to perform some or all of the actual operations, thereby realizing the function of any of the embodiments described above.

[0097] Furthermore, it is understood that the program code read from the storage medium is written to the memory set in the expansion board inserted into the computer or to the memory set in the expansion module connected to the computer. Then, based on the instructions of the program code, the CPU or other components installed on the expansion board or expansion module execute some and all of the actual operations, thereby realizing the function of any of the above embodiments.

[0098] It is understood that explanations, specific implementation methods, beneficial effects, examples, etc. of the contents in the computer-readable medium provided in the embodiments of the present invention can be found in the corresponding parts of the method provided in the first aspect, and will not be repeated here.

[0099] According to a fourth aspect, one embodiment of this specification provides a computing device including a memory and a processor, wherein the memory stores executable code, and the processor, when executing the executable code, implements the method of any embodiment of the specification.

[0100] It is understood that explanations, specific implementation methods, beneficial effects, examples, etc. of the computing device provided in the embodiments of the present invention can be found in the corresponding parts of the method provided in the first aspect, and will not be repeated here.

[0101] It is understood that the structures illustrated in the embodiments of this specification do not constitute a specific limitation on the apparatus of the embodiments of this specification. In other embodiments of the specification, the above-described apparatus may include more or fewer components than illustrated, or combine some components, or split some components, or have different component arrangements. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0102] The information interaction and execution process between the modules in the above-mentioned device and system are based on the same concept as the method embodiments in this specification, and the specific details can be found in the descriptions in the method embodiments in this specification, so they will not be repeated here.

[0103] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the apparatus embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0104] Those skilled in the art will recognize that, in one or more of the examples above, the functions described in this invention can be implemented using hardware, software, widgets, or any combination thereof. When implemented in software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium.

[0105] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made on the basis of the technical solution of the present invention should be included within the scope of protection of the present invention.

Claims

1. A side data adding method, characterized by, The method comprises the following steps: splitting each customer address in the full quantity of customers in the graph to obtain regional information and address information of the customer address; wherein the address information and the customer are in one-to-one correspondence; generating a first matrix according to each address information corresponding to the same regional information; wherein an element in the ith row and the jth column of the first matrix represents whether the jth address information in each address information corresponding to the same regional information contains the ith string in the string set corresponding to the regional information; the number of columns of the first matrix is the number of each address information corresponding to the same regional information, and the number of rows of the first matrix is the number of strings in the string set; dimension reduction is performed on the first matrix to obtain a second matrix, and each address information corresponding to the same regional information is classified according to the second matrix, and address information in the same class has similarity; calculating the similarity of each two address information in the same class, and selecting an address information pair with a similarity greater than a preset threshold; for each address information pair selected, determining two nodes corresponding to the two customers in the graph, and adding a relationship edge between the two determined nodes; The method comprises the following steps: splitting each customer address in the full quantity of customers in the graph to obtain regional information and address information of the customer address; wherein the address information and the customer are in one-to-one correspondence; The method comprises the following steps: splitting each address information corresponding to the same regional information to obtain a plurality of strings corresponding to the address information; wherein each string comprises characters of a preset length; forming a string set corresponding to the regional information by using the strings corresponding to each address information corresponding to the same regional information.

2. The method of claim 1, wherein, The method comprises the following steps: adding each string of each address information corresponding to the same regional information to an initially empty string set one by one, and before adding the string corresponding to each address information to the string set, judging whether the string corresponding to the address information exists in the string set; if yes, adding the string corresponding to the address information to the string set to form the string set without containing the same string.

3. The method of claim 1, wherein, The method comprises the following steps: using a minimum hash algorithm to perform dimension reduction on the first matrix to obtain a second matrix; wherein the number of elements of a column vector corresponding to each address information in the second matrix is less than the number of elements of a column vector corresponding to the address information in the first matrix.

4. The method of claim 1, wherein, The method comprises the following steps: selecting at least one local sensitive hash function; adopting each local sensitive hash function to perform hash calculation on each column vector corresponding to each address information in the second matrix, to obtain a corresponding bucket sequence number of the address information under the local sensitive hash function, and divide the address information into a hash bucket corresponding to the bucket sequence number, each hash bucket corresponding to a number sequence, which includes a set of address information divided into the hash bucket under the local sensitive hash function; merge a number sequence having at least one same address information in each number sequence corresponding to each hash bucket under each local sensitive hash function into one category.

5. The method of claim 1, wherein, The similarity calculation on each two address information in the same category includes: performing Jacobian similarity calculation on each two address information in the same category.

6. A side data adding apparatus characterized by comprising: comprise: The address splitting module is configured to split each customer address in the full customer of the graph to obtain regional information and address information of the customer address; wherein the address information and the customer are in one-to-one correspondence. The matrix generating module is configured to generate a first matrix according to each address information corresponding to the same regional information; wherein an element in the ith row and the jth column of the first matrix represents whether the jth address information in each address information corresponding to the same regional information contains the ith string in a string set corresponding to the regional information; the number of columns of the first matrix is the number of each address information corresponding to the same regional information, and the number of rows of the first matrix is the number of strings in the string set. The dimension reduction classification module is configured to perform dimension reduction on the first matrix to obtain a second matrix, and classify each address information corresponding to the same regional information according to the second matrix, and the address information in the same category has similarity. The similarity calculation module is configured to perform similarity calculation on each two address information in the same category, and select an address information pair with a similarity greater than a preset threshold. The edge adding module is configured to determine two nodes corresponding to the two customers corresponding to each selected address information pair in the graph, and add a relationship edge between the two determined nodes. The address splitting module is configured to split each customer address in the full customer according to regional levels to obtain regional information of each level and address information corresponding to regional information of the lowest level. The device further comprises a set forming module configured to generate a string set, and the set forming module comprises: The address splitting unit is configured to split each address information in each address information corresponding to the same regional information to obtain a plurality of strings corresponding to the address information; wherein each string includes characters of a preset length. The set forming unit is configured to form a string set corresponding to the regional information from each string corresponding to each address information corresponding to the same regional information. 7.A computer readable storage medium having a computer program stored thereon, when the computer program is executed in a computer, the computer program causes the computer to perform the method in any one of claims 1-5.

8. A computing device comprising a memory having executable code stored therein and a processor that, when executing the executable code, implements the method of any one of claims 1-5.

Images