A dynamic information correlation method

Abstract

The application discloses a dynamic information association method, comprising the following steps: 1) receiving a new information record, and determining a similar information entity set according to a static attribute; 2) if there is no similar information entity, a new information entity is created, and the process is ended; if there is a similar information entity, a history information record of a dynamic attribute of the similar information entity is obtained; 3) a prediction model for each dynamic attribute of the information entity is constructed, a reasonable value range of the dynamic attribute is determined, a correlation feature of the prediction model of each dynamic attribute of the information entity is constructed, and a prediction value range of the dynamic attribute of the similar information entity is calculated; and 4) if there is no similar information entity, a new information entity is created; if there is a similar information entity, the new information record is associated to the most matched candidate information entity. According to the application, the change trend of the dynamic attribute and the correlation between the dynamic attributes are utilized to associate information from different data sources and / or different time points, so that the accuracy and recall rate of the association are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of computers, and particularly relates to a dynamic information correlation method. BACKGROUND

[0002] Participants (such as natural persons, companies) in an information society will generate a large amount of information every day, which may come from different systems, different collection times, examples of information from different systems: Zhang San has invested in company A, and is also a special expert of company B, the sources of the two pieces of information may be different systems, such as the former is information from the national enterprise credit information public system, and the latter may come from the official website of company B. Examples of information from different collection times: financial institution C regularly inquires about the external debt situation of mortgage user Li Si, and at the end of June 2021, it is found that Li Si has a loan with a balance of 50,000 yuan, and at the end of December 2021, it is found that Li Si has a loan with a balance of 40,000 yuan, and the two pieces of loan information may correspond to the same loan. It is necessary to correlate the information that exists in business connection, and through information correlation, a complete information chain can be established to reduce data redundancy. For example, when constructing an enterprise graph, if Zhang San is regarded as the same natural person, the association between company A and company B can be mined, and at the same time, the repeated data of storing two "Zhang Sans" is avoided. When doing loan default risk management, the loan records of the user at different times can be correlated to analyze the repayment behavior of the user.

[0003] In order to clearly distinguish the concepts involved in the scheme of the application, the following professional terms are defined.

[0004] 1. Information entity: content with specific business attributes, such as a natural person in an enterprise graph (business attributes such as name, ID number, external investment relationship, etc.), a bank loan (business attributes such as loan balance, loan institution, loan period, etc.);

[0005] 2. Information record: records the content of an information entity from different sources or different time points, and one information entity can correspond to one or more information records. Examples of information from different sources, information record 1: "Shareholder of company A - Zhang San", information record 2: "Special expert of company B - Zhang San", if the two Zhang Sans correspond to the same person, then the two information records of the information entity "Zhang San" correspond to the same person; examples of information from different time points, a bank loan respectively in information records "balance 50,000 yuan, loan type: credit loan, loan period 24 periods" collected in June 2021 and "balance 40,000 yuan, loan type: credit loan, loan period 24 periods" collected in December 2021.

[0006] It should be noted that the attributes of information entities can change dynamically, that is, different information records of the same information entity are not completely the same, and associating dynamic information with business contact is a challenging technical problem. Prior art CN110019542A, method for generating enterprise relationship in enterprise graph, method and device for generating organization member database and identifying same-name members, prior art one uses the identification information of the member to determine whether the same-name members are the same person, wherein the identification attributes include individual attributes and / or relationship attributes. When determining whether it is the same person based on individual attributes, it is required that the attribute values are the same; when determining whether it is the same person based on relationship attributes, it is required that the number of associated same members meets a set threshold. For example, there are two records about "Zhang San", and the two Zhang Sans have 3 common contacts, Li Si, Wang Wu and Zhao Liu. Assuming that the threshold condition for information association is >= 3, then it can be considered that the "Zhang San" of the two records is the same person. Prior art CN106487825B, information association method and device, prior art two is mainly used for associating user information uploaded by different clients, and the association is mainly based on the attribute of address location. If the distance between the geographical positions of two information records is less than a threshold, they are associated. The prior art has the following problems:

[0007] The prior art does not consider the change trend of dynamic attributes and the correlation between dynamic attributes when associating information records, which affects the accuracy (information without business contact is associated) and recall rate (information with business contact is not associated) of information association. For example, when determining whether the same-name members correspond to the same entity, the relationship attributes that can be used include investment relationship, and the solution of prior art one is to compare whether the two same-name members have several completely same investment companies, without using the investment relationship of the entity at historical time and the investment preference and expansion trend that can be extracted therefrom. Prior art two only determines whether the association can be performed according to the distance between the physical positions of the user at different time, without considering the motion direction, acceleration and other information contained in the historical path of the user. SUMMARY

[0008] To achieve the above purpose, the technical solution of the present application is as follows: a dynamic information association method, the solution of the present application uses the change trend of dynamic attributes and the correlation between dynamic attributes to associate information from different data sources and / or different time, improving the accuracy and recall rate of association, and the system module architecture involved in the present application includes:

[0009] Information processing unit: storing the identified information entity content, at least containing the unique identifier of the information entity, static attributes, etc.

[0010] Information entity database: store the information records corresponding to the identified information entities, at least containing the unique identification of the corresponding information entities, dynamic attributes, etc. The dashed line in the architecture diagram can associate the information records and the information entities through the unique identification of the information entities.

[0011] Information record database: responsible for receiving new information records, querying the content of the information entity database and / or the information record database, and processing the relevant content to generate information association results, and writing the relevant results to the corresponding database.

[0012] The method comprises the following steps:

[0013] Step 1) receiving a new information record, containing static attributes and dynamic attributes, and determining a similar information entity set according to the static attributes;

[0014] Step 2) if there is no similar information entity, create a new information entity, and the process ends, if there is a similar information entity, obtain the historical information records of the dynamic attributes of the similar information entity;

[0015] Step 3) constructing a prediction model for each dynamic attribute of the information entity, and determining a reasonable value range of the dynamic attribute, constructing a correlation feature of the prediction model of each dynamic attribute of the information entity, and calculating the prediction value range of the dynamic attribute of the similar information entity;

[0016] Step 4) determining the candidate information entity according to the matching situation of the prediction value range, judging whether there is a candidate information entity, if not, creating a new information entity, and the process ends, if there is, associating the new information record to the most matched candidate information entity, and the process ends.

[0017] Based on the above technical solution, the static attributes include static individual attributes and / or static relationship attributes, such as ID number, name, enterprise unified credit code, etc. Static relationship attributes such as foreign investment, enterprise position, family relationship, etc. The dynamic attributes include dynamic individual attributes and / or dynamic relationship attributes, such as financing amount, age, etc. Dynamic relationship such as the list of enterprises invested by the company and the holding ratio, the list of natural persons with direct or indirect association, etc.

[0018] As an improvement of the present application, the step 1) further comprises: the step of determining the similar information entity set according to the static attributes is: comparing one or more static attributes of the new information record with the static attributes of the information entities in the information entity database. The static attributes with the same value form a similar information entity set of the new information record, and the similar information entity set ID is {ID1, ID2,..., IDn}.

[0019] As an improvement of the present application, the step 2) further comprises: if there is no similar information entity, creating a new information entity and adding it to the information entity database, adding the new information record to the information record database and associating it to the newly created information entity, and ending the process; if there is a similar information entity, obtaining the historical data set of each similar information entity dynamic attribute from the information record database, the historical data being collected from different historical moments and / or different information sources, the historical data set of a specific dynamic attribute being as follows: {x1, x2, x3,.... x t}, wherein Xi represents the value at different moments.

[0020] As an improvement of the present application, the step 3) further comprises: based on the historical information record of the dynamic attribute, refining the change information of the dynamic attribute, and using the time series method to fit the historical data to construct a prediction model for each dynamic attribute of the information entity, for calculating the predicted value range of the dynamic attribute at a future moment.

[0021] Based on the above technical solution, the change information of the dynamic attribute includes change trend and change amplitude, etc., and the following prediction model can be used: x t =f(T t ,C t ,I t ), wherein Xt is the value of the dynamic attribute at moment t, Tt is the trend factor of the dynamic attribute, Ct is the regular fluctuation factor of the dynamic attribute, and It is the random fluctuation factor of the dynamic attribute.

[0022] As an improvement of the present application, the step 3) further comprises: combining the business rules and / or historical data to determine the reasonable value range δ of the dynamic attribute.

[0023] As an improvement of the present application, the step 3) further comprises: constructing the correlation feature of the prediction model of each dynamic attribute of the information entity, and describing the correlation ρ between the dynamic attributes, such as positive correlation and negative correlation, wherein the correlation is represented by ρ(X, Y), X representing dynamic attribute one and Y representing dynamic attribute two.

[0024] As an improvement of the present application, the step 3) comprises: calculating the predicted value range of the dynamic attribute of the similar information entity at the current moment, wherein the predicted value range is obtained by adding and subtracting the fluctuation value to the predicted value calculated by the generated prediction model, and here the fluctuation value can be obtained based on the change amplitude of the dynamic attribute and / or the experience value and / or the reasonable value range.

[0025] As an improvement of the present application, the step 4) further comprises: if one or more dynamic attributes of the new information record are within the predicted value range and meet the correlation characteristics, the corresponding similar information entity is taken as a candidate information entity; if the dynamic attributes of the new information record all exceed the predicted value range, it is considered as a new information entity.

[0026] As an improvement of the present application, the step 4) further comprises: determining the candidate information entity with the highest matching degree with the new information record as a target information entity, and adding the new information record into the information record database and associating it to the target information entity, wherein the matching degree is used to measure the similarity between the new information record and one or more dynamic attributes of the information entity. The candidate information entity with the highest matching degree is selected as the target information entity.

[0027] As an improvement of the present application, the step 4) further comprises: if several dynamic attributes of the information record exceed the reasonable value range, the relevant dynamic attributes are corrected using the correlation information between the dynamic attributes and / or the predicted value range.

[0028] Compared with the prior art, the present application has the beneficial effects that: the change trend of the dynamic attributes and the correlation between the dynamic attributes are used to associate the information from different data sources and / or different time points, so as to improve the accuracy and recall rate of the association. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1 A schematic diagram of a background technical comparison file 1.

[0030] Figure 2 A schematic diagram of a background technical comparison file 2.

[0031] Figure 3 A system architecture diagram of the present application.

[0032] Figure 4 A flowchart of the present application.

[0033] Figure 5 A U001 holding ratio prediction model-A company in embodiment 1 of the present application. DETAILED DESCRIPTION

[0034] The present application will be further illustrated below in combination with the drawings and specific embodiments, and it should be understood that the following specific embodiments are only used to illustrate the present application and not to limit the scope of the present application.

[0035] Embodiment 1: This embodiment takes the construction of an enterprise graph as the background, and the enterprise graph is composed of entities and relationships between entities, such as natural persons, enterprises, telephone numbers, addresses, etc., and relationships between entities, such as investment, employment, etc.

[0036] The system architecture diagram in this embodiment is as followsFigure 3 As shown, the information entity database stores information entities of the type of natural person, and it is assumed that the following content already exists, that there are two natural persons with the name Zhang San, and the unique identifiers are U001 and U002 respectively.

[0037] Table 1 Example One Information Entity Database - Initial Content of Natural Person

[0038] Unique identification Name Certificate number U001 Zhang San ID0001 U002 Zhang San ID0002

[0039] (1.1) The information record database stores information records of the type of investment relationship, and it is assumed that the following content already exists, that there are 3 information records of investment in Company A and 4 information records of investment in Company B for the information entity with the unique identifier U001, collected from different times and information sources, and there are 4 information records of investment in Company A and 3 information records of investment in Company B for the information entity with the unique identifier U002, collected from different times and information sources respectively.

[0040] Table 2 Example One Information Record Database - Initial Content

[0041]

[0042]

[0043] The following new information record is received, in which the static individual attribute and value are "Name: Zhang San", and the dynamic relationship attributes and values are "A Company Shareholding: 500,000" and "A Company Shareholding Ratio: 0.5%", "B Company Shareholding: 4,500,000" and "B Company Shareholding Ratio: 4.5%".

[0044] Table 3 Example One New Information Record

[0045]

[0046] (1.2) Determine the similar information entity set according to the static attribute, compare the static attribute "Name: Zhang San" of the new information record with the static attribute of the information entity in the information entity database (Table 1).

[0047] (1.3) Determine that the information entities with the same static attribute have unique identifiers U001 and U002, and these two information entities form the similar information entity set.

[0048] Similar Information Entity Set ID: {U001, U002}

[0049] Obtain the historical data of the dynamic attributes of each similar information entity from the information record database, that is, the content in Table 2.

[0050] (1.4) Based on the historical information record of dynamic attributes, the change trend, change range and volatility rate of dynamic attributes are extracted, and the prediction model of each dynamic attribute of information entity is constructed.

[0051] (1.5) In addition, combined with data and business rule restrictions, it can be determined that the reasonable value range of the holding ratio of dynamic attributes is 0-100%.

[0052] (1.6) For similar information entity U001, the change trend of the holding ratio of A company is decreasing, that is, "reducing holding". The following information can be extracted from Table 2 to construct a prediction model( Figure 5 ), and in this embodiment, the time series method is used to fit the historical data to generate the model (assuming that the model only has a trend term, the model is: where is the predicted value of the next moment, x t the value of the latest moment in the historical data). There are many decomposition methods in time series analysis, which are not introduced here.

[0053] Table 4 Example one U001 training prediction model data-A company

[0054] Collection time 1 2 3 Shareholding ratio 3％ 2％ 1％

[0055] For similar information entity U002, the change trend of the holding ratio of A company is rising, that is, "increasing holding", and the model curve change trend is opposite to U001.

[0056] Similarly, the holding ratio of similar information entities U001 and U002 to B company can also be constructed by using the same method to construct a prediction model.

[0057] The correlation characteristics of the prediction model of each dynamic attribute of the information entity are constructed. Specifically, the holding ratio of similar information entity U001 to A company (reducing trend) and the holding ratio to B company (increasing trend) are negatively correlated, and the holding ratio of similar information entity U002 to A company (increasing trend) and the holding ratio to B company (increasing trend) are positively correlated.

[0058] Based on the prediction model generated in step 3, the predicted value range of the dynamic attribute of the similar information entity at the current moment is calculated.

[0059] The prediction model of similar information entity U001 holding A company predicts that the value at the next moment is 0 (=0.01-0.01), combined with the reasonable value range and the average change range (1%), a predicted value interval [0, 1%] can be further obtained. In order to improve the recall rate of matching, a range is predicted instead of an accurate value.

[0060] (1.7) Similarly, the prediction ranges of similar information entities U001 and U002 for the holding ratio of company A and company B are as follows:

[0061] Table 5 Prediction of holding ratio in example one

[0062] Information entity unique identification Shareholding company Predicted shareholding ratio U001 Company A [0，1％] U001 Company B [4％，6％] U002 Company A [4％，6％] U002 Company B [3％，5％]

[0063] Referring to Table 2, the time series of U001 for the holding ratio of company B is 0.5% -> 1.5% -> 3% -> 4%, the trend is rising, so the predicted value at the next time is 5%, plus or minus a fluctuation value (1%) to get the range 【4%, 6%】, referring to Table 2, the time series of U002 for the holding ratio of company B is 1% -> 2% -> 3%, the trend is rising, so the predicted value at the next time is 4%, plus or minus a fluctuation value (1%) to get the range 【3%, 5%】.

[0064] (1.8) If the dynamic attributes of the new information record (Table 3), the holding ratio of company A and the holding ratio of company B, are in the prediction value range and meet the correlation characteristics (trend negative correlation), the corresponding similar information entity U001 is taken as a candidate information entity. The holding ratio of company A and company B of the new information record does not match the prediction value range and the correlation characteristics of the similar information entity U002, so U002 is not a candidate information entity.

[0065] (1.9) The candidate information entity with the highest matching degree with the new information record is determined as the target information entity, and the new information record is added to the information record database and associated with the target information entity. In this embodiment, there is only one candidate information entity U001, so the information entity is the target information entity, and the new information record is added to the information record database to obtain the following content:

[0066] Table 6 Information record database in example one - updated content

[0067]

[0068]

[0069] If there are multiple candidate information entities, the similarity between the new information record and the candidate information entities needs to be calculated and compared, for example, the similarity is the standardized Euclidean distance between the predicted values of the dynamic attributes of the new information record and the candidate information entities. Since the values of the dynamic attributes of the new information record are all in the reasonable value range (step 4: 0-100%), no correction is needed, and the process is ended.

[0070] Example two: information correction is needed:

[0071] Assume that the initial state of embodiment two is consistent with that of embodiment one (see Table 1 and Table 2).

[0072] (2.1) Receive the following new information record, in which the static individual attribute and value are "Name: Zhang San", and the dynamic relationship attribute and value are "A Company's shareholding: 500,000" and "A Company's shareholding ratio: 0.5%", "B Company's shareholding: 4,500,000" and "B Company's shareholding ratio: 450%". Among them, "B Company's shareholding ratio: 450%" is different from embodiment one.

[0073] Table 7 New information record of embodiment two

[0074]

[0075] (2.2) to (2.6) are the same as (1.2) to (1.6) of embodiment one;

[0076] (2.7) The dynamic attribute "A Company's shareholding ratio" of the new information record (Table 7) is in the predicted value range, and the corresponding similar information entity U001 is selected as a candidate information entity. The shareholding ratios of A Company and B Company in the new information record do not match the predicted value range of similar information entity U002, so U002 is not selected as a candidate information entity.

[0077] (2.8) Determine the candidate information entity with the highest matching degree with the new information record as the target information entity, and add the new information record to the information record database and associate it to the target information entity. In this embodiment, there is only one candidate information entity U001, so this information entity is the target information entity, and the new information record is added to the information record database to obtain the following content:

[0078] Table 8 Information record database of embodiment two - updated content (before correction)

[0079]

[0080]

[0081] (2.9) Because the value of the dynamic attribute "B Company's shareholding ratio" in the new information record exceeds the reasonable value range (step 4: 0-100%), it needs to be corrected. According to the correlation between the shareholding number and the shareholding ratio refined from historical data (shareholding ratio = shareholding number * 0.0001), the abnormal data 450% is corrected to 4.5%.

[0082] It should be noted that the above content only illustrates the technical idea of the present application, and cannot limit the protection scope of the present application. For ordinary skilled in the art, without departing from the principle of the present application, a number of improvements and refinements can be made, which fall within the protection scope of the claims of the present application.

Claims

1. A method of dynamically associating information, characterized by, The method comprises the following steps: Step 1) receiving an added information record, containing static attributes and dynamic attributes, and determining a similar information entity set according to the static attributes; Step 2) if there is no similar information entity, creating a new information entity, and ending the process, if there is a similar information entity, obtaining historical information records of dynamic attributes of the similar information entity; Step 3) constructing a prediction model for each dynamic attribute of the information entity, determining a reasonable value range of the dynamic attribute, constructing a correlation feature of the prediction model of each dynamic attribute of the information entity, and calculating a prediction value range of the dynamic attribute of the similar information entity; In step 3), the correlation feature of the prediction model of each dynamic attribute of the information entity is also constructed, which describes the correlation between the dynamic attributes ρ, wherein the correlation is represented by ρ(X, Y), X represents dynamic attribute one, and Y represents dynamic attribute two; In step 3), the prediction value range of the dynamic attribute of the similar information entity at the current time is calculated respectively, wherein the prediction value range is generated by adding or subtracting the generated prediction value of the prediction model; Step 4) determining a candidate information entity according to the matching of the prediction value range, judging whether there is a candidate information entity, if not, creating a new information entity, and ending the process, if there is, associating the added information record to the most matched candidate information entity, and ending the process, in step 1), the step of determining the similar information entity set according to the static attributes is: comparing one or more static attributes of the added information record with the static attributes of the information entities in the information entity database, and the several information entities with the same static attribute value constitute the similar information entity set of the added information record, the similar information entity set ID: {ID1, ID2,..., IDn}, in step 2), if there is no similar information entity, a new information entity is created and added to the information entity database, the added information record is added to the information record database and associated with the newly created information entity, and the process is ended; if there is a similar information entity, the historical data set of each dynamic attribute of the similar information entity is obtained from the information record database, the historical data is collected from different historical time and / or different information source, in step 3), the change information of the dynamic attribute is extracted based on the historical information record of the dynamic attribute, the prediction model for each dynamic attribute of the information entity is constructed by using the time series method to fit the historical data, which is used to calculate the prediction value range of the dynamic attribute at the future time, in step 4), if one or more dynamic attributes of the added information record are within the prediction value range and meet the correlation feature, the corresponding similar information entity is taken as the candidate information entity, if the dynamic attributes of the added information record all exceed the prediction value range, it is considered as a new information entity; The change information of the dynamic attribute includes the change trend and the change amplitude information, The following prediction model is used: xt f t t It wherein Xt is the value of the dynamic attribute at time t, Tt is a trend factor of the dynamic attribute, Ct is a regular fluctuation factor of the dynamic attribute, and It is a random fluctuation factor of the dynamic attribute.​​​​ 2. The method of claim 1, wherein, In step 3), the reasonable value range δ of the dynamic attribute is determined by combining the business rules and / or historical data.

3. The method of claim 1, wherein, The step 4) further comprises: determining and adding a candidate information entity with the highest matching degree to the information record as a target information entity, and adding the information record to the information record database and associating it to the target information entity, wherein the matching degree is used to measure the similarity between the information record and one or more dynamic attributes of the information entity, and the candidate information entity with the highest matching degree is selected as the target information entity.

4. The method of claim 1, wherein, The step 4) further comprises: if several dynamic attributes of the information record exceed a reasonable value range, correcting the related dynamic attributes by using the correlation information between the dynamic attributes and / or the predicted value range.

Images