CN115238170A - User portrait processing method and system based on block chain finance - Google Patents

Abstract

The embodiment of the disclosure discloses a user portrait processing method and system based on block chain finance, which can not only efficiently and accurately obtain a preference portrait, but also accurately perform linkage analysis processing according to the number K1 of a plurality of block chain financial service users, and can balance two conditions of 1 block chain financial service user and a plurality of block chain financial service users. Based on the method, the timeliness and the precision of preference portrait mining can be improved, and the flexibility of portrait analysis for different numbers of block chain financial business users can be improved.

Description

User portrait processing method and system based on block chain finance

Technical Field

The disclosure relates to the technical field of block chain finance, in particular to a user portrait processing method and system based on block chain finance.

Background

Products in the digital economy era are no longer limited to physical products, more digital products and services are involved, and further, the global economy has new economic characteristics of globalization, virtualization, socialization and economy sharing. On this basis, the blockchain finance not only relates to virtual currency and distributed ledger technology, but also focuses more on the upgrading of financial business interaction modes. As the size and class of businesses proliferate, the challenges facing blockchain finance continue to expand. For example, it is often difficult to meet current needs for traditional user analysis techniques for blockchain finance, which are still required to improve the efficiency and accuracy of portrait mining, for example, in terms of user portrait analysis.

Disclosure of Invention

One objective of the present disclosure is to provide a method and system for user representation processing based on block chain finance.

The technical scheme of the disclosure is realized by at least some of the following embodiments.

A user portrait processing method based on block chain finance is realized by a block chain finance service system, and the method at least comprises the following steps: obtaining a first session activity description phrase for reflecting big data of a plurality of blockchain financial business users in a plurality of groups of service sessions, and obtaining a second session activity description phrase for reflecting labels of the plurality of blockchain financial business users relative to target preference portrayal; performing linkage analysis processing based on the first session activity description phrase and the second session activity description phrase to obtain global activity description phrase distribution of each block chain financial business user in each service session big data; the method comprises the following steps that the linkage analysis processing type is related to the number K1 of a plurality of block chain financial service users; performing preference portrait mining based on global activity description phrase distribution to obtain a preference portrait knowledge base of a plurality of block chain financial service users relative to a target preference portrait label; the preference portrait knowledge base comprises a plurality of groups of service session big data, and the service session big data carry preference portrait detail fields of financial service users of each block chain.

The method is applied to the embodiment, a first session activity description phrase for reflecting big data of a plurality of block chain financial service users in a plurality of groups of service sessions is obtained, a second session activity description phrase for reflecting a plurality of block chain financial service users relative to a target preference portrait label is obtained, further, linkage analysis processing is carried out based on the first session activity description phrase and the second session activity description phrase, global activity description phrase distribution of each block chain financial service user in each service session big data is obtained, the type of linkage analysis processing is associated with the number K1 of the plurality of block chain financial service users, preference mining is carried out based on the global activity description phrase distribution, a preference portrait knowledge base of the plurality of block chain financial service users relative to the target preference portrait label can be obtained, the preference knowledge base comprises a plurality of groups of service session big data, the service session big data comprises preference detail portrait fields of each block chain financial service user, not only can obtain preference accurately, but also can accurately carry out linkage analysis processing according to the number K1 of the plurality of block chain financial service users, and conditions of two types of the plurality of block chain financial service users can be balanced. Based on the method, the timeliness and the precision of the preference portrait mining can be improved, and the flexibility of portrait analysis for different numbers of block chain financial business users can be improved.

In some independent embodiments, on the basis that the number K1 of the plurality of blockchain financial service users is 1, the linkage analysis processing includes generating a phase passing network between service session big data; and/or on the basis that the number K1 of the plurality of blockchain financial service users is X, the linkage analysis processing comprises the steps of generating an active transmission network among the plurality of blockchain financial service users in each service session big data and generating a stage transmission network among the service session big data.

Applied to the embodiment, the linkage analysis processing includes generating a phase delivery network between service session big data on the basis that the number K1 of a plurality of block chain financial service users is 1, so that the continuity of the service phase between the service session big data can be improved by constructing the phase delivery network, which is helpful for guaranteeing the authenticity of the preference profile knowledge base, and the linkage analysis processing includes generating an activity delivery network between a plurality of block chain financial service users in each service session big data and generating a phase delivery network between each service session big data on the basis that the number K1 of a plurality of block chain financial service users is X, which is helpful for guaranteeing the interactive analysis quality between the block chain financial service users by constructing the activity delivery network, and the continuity of the service phase between the service session big data can be guaranteed by constructing the phase delivery network, which is helpful for guaranteeing the authenticity of the preference profile knowledge base.

In some independent embodiments, on the basis that the linkage analysis processing includes constructing a phase delivery network, performing linkage analysis processing based on the first session activity description phrase and the second session activity description phrase to obtain a global activity description phrase distribution of each blockchain financial business user in each service session big data, including: screening the blockchain financial service users as target blockchain financial service users, and regarding a first session activity description phrase and a second session activity description phrase corresponding to the target blockchain financial service users as phase session activity description phrases of the target blockchain financial service users in different service phases; sequentially screening all service stages to be regarded as a first current service stage, and screening a stage session activity description phrase of the first current service stage to be regarded as an activity characteristic of the first current service stage; obtaining global activity description phrase distribution corresponding to the activity characteristics of the first current service phase by utilizing each first service phase activity characteristic template and the PPMCC of the activity characteristics of the first current service phase; the first business stage activity characteristic template comprises a stage session activity description phrase of the target blockchain financial business user in each business stage.

In some independent embodiments, on the basis that the linkage analysis processing includes constructing an activity delivery network, performing linkage analysis processing based on the first session activity description phrase and the second session activity description phrase to obtain a global activity description phrase distribution of each block chain financial business user in each service session big data, includes: screening the blockchain financial service users as target blockchain financial service users, and regarding a first session activity description phrase and a second session activity description phrase corresponding to the target blockchain financial service users as phase session activity description phrases of the target blockchain financial service users in different service phases; sequentially screening each service stage as a second current service stage, and screening the stage session activity description phrase of the second current service stage as the activity characteristic of the second current service stage; obtaining global activity description phrase distribution corresponding to the second current service phase activity characteristic by utilizing each second service phase activity characteristic template and the PPMCC of the second current service phase activity characteristic respectively; the second business stage activity characteristic template comprises stage session activity description phrases of each block chain financial business user in the second current business stage.

The method is applied to the embodiment, a screening blockchain financial service user is regarded as a target blockchain financial service user, a first session activity description phrase and a second session activity description phrase corresponding to the target blockchain financial service user are regarded as phase session activity description phrases of the target blockchain financial service user in different service phases, the phase session activity description phrases in different service phases are regarded as current service phase activity characteristics respectively based on the phase session activity description phrases, then, global activity description phrase distribution corresponding to the current service phase activity characteristics is obtained based on PPMCC of each service phase activity characteristic template and the current service phase activity characteristics respectively, on the basis of constructing a phase transfer network, the service phase activity characteristic template comprises the phase session activity description phrases of the target blockchain financial service user in each service phase, on the basis of constructing the activity transfer network, the service phase activity characteristic template comprises the phase session activity description phrases of each blockchain financial service user in an exemplary service phase respectively, and the exemplary service phases are service phases corresponding to the current service phase activity characteristics, so that the phase transfer network and the intelligent service chain can be constructed by similar construction ideas, and the flexibility of the service chain financial service users and the mining flexibility of the service can be guaranteed.

In some independent embodiments, on the basis that the linkage analysis processing includes constructing an activity delivery network and a phase delivery network, performing linkage analysis processing based on the first session activity description phrase and the second session activity description phrase to obtain a global activity description phrase distribution of each blockchain financial business user in each service session big data, including: constructing a first transfer description based on the first session activity description phrase and the second session activity description phrase, obtaining a session activity description phrase out feature of the first transfer description, constructing a second transfer description based on the session activity description phrase out feature, and obtaining a global activity description phrase distribution; the first transfer is described as an active transfer network and the second transfer is described as a phase transfer network, and for example, the first transfer is described as a phase transfer network and the second transfer is described as an active transfer network.

The method is applied to the embodiment, and on the basis that linkage analysis processing comprises the construction of the activity transfer network and the stage transfer network, the conversation activity description phrase out feature of the activity transfer network constructed in advance is the conversation activity description phrase raw material of the stage transfer network constructed in the following, so that under the condition of a plurality of block chain financial service users, the activity transfer network and the stage transfer network are constructed in sequence, the distribution of all global activity description phrases is respectively matched with the activity transfer network and the stage transfer network, and the method is favorable for ensuring the sorting quality of the activity transfer network and the stage transfer network.

In some independent embodiments, the preference profile knowledge base is obtained by a preference profile processing network, the preference profile processing network including a linkage analysis processing layer, and the linkage analysis processing layer including a business phase analysis layer for constructing a phase delivery network and an activity impact analysis layer for constructing an activity delivery network.

The preference portrait knowledge base is obtained by the preference portrait processing network, the working processing network comprises a linkage analysis processing layer, the linkage analysis processing layer comprises a business phase analysis layer and an activity influence analysis layer, the business phase analysis layer is used for constructing a phase transmission network, and the activity influence analysis layer is used for constructing an activity transmission network, so that preference portrait mining can be completed through an AI algorithm, and preference portrait mining timeliness and accuracy can be guaranteed.

In some independent embodiments, the first session activity description phrase is based on a mining of the iterative analysis task.

The method is applied to the embodiment, the first session activity description phrase is mined based on the iterative analysis task, the acquisition difficulty of the first session activity description phrase can be obviously reduced, and the acquisition precision and the authenticity of the description phrase under the preference portrait data with sufficient tags can be improved.

In some independent embodiments, obtaining a first session activity description phrase reflecting big data of a plurality of blockchain financial transaction users in a plurality of groups of service sessions comprises: in a plurality of iterative analysis tasks, mining is carried out based on the number K2 respectively, and first basic description phrases used for reflecting K2 service session big data are obtained; the scale constraint value of the first basic description phrase is the same as the number of the iterative analysis tasks, and the task configurations of the iterative analysis tasks are not consistent with each other; obtaining K3 first session activity description phrases based on the number K1 and the first base description phrases; wherein, the number K3 is the set operation result of the number K1 and the number K2.

The method is applied to the embodiment, in multiple rounds of iterative analysis tasks, mining is performed based on the number K2 respectively to obtain first basic description phrases for reflecting K2 service session big data, the scale constraint value of the first basic description phrases is the same as the number of the iterative analysis tasks, the task configurations of the iterative analysis tasks are not consistent with each other, then, based on the number K1 and the first basic description phrases, K3 first session activity description phrases are obtained, the number K3 is the set operation result of the number K1 and the number K2, due to the fact that the task configurations of the iterative analysis tasks are not consistent with each other, and the execution of each round of iterative analysis tasks can obtain the significant features of the big data of each service session, and therefore the precision and the reliability of each first session activity description phrase can be guaranteed.

In some independent embodiments, the second session activity description phrase is obtained based on target preference portrait tag recognition.

The method is applied to the embodiment, the second conversation activity description phrase is obtained based on the identification of the target preference portrait tag, so that the second conversation activity description phrase can be obtained only by the operations of identification of the feature vector and the like, and the difficulty of preference portrait mining can be obviously reduced.

In some independent embodiments, obtaining a second session activity description phrase reflecting a plurality of blockchain financial transaction users portrait tags relative to a target preference includes: carrying out description transformation on the target preference portrait label to obtain a second basic description phrase; based on the number K1 and the second base description phrases, a number K1 of second session activity description phrases is obtained.

The method is applied to the embodiment, description transformation is carried out on the target preference portrait label to obtain the second basic description phrase, the number K1 of second conversation activity description phrases are obtained based on the number K1 and the second basic description phrases, the number K1 of second conversation activity description phrases can be obtained by carrying out description transformation on the structured data and combining the number K1 for processing, and the difficulty in obtaining the second conversation activity description phrases can be obviously reduced.

In some independent embodiments, both the first session activity description phrase and the second session activity description phrase carry distribution characteristics; the distribution characteristics comprise service stage distribution characteristics on the basis that the plurality of block chain financial service users are 1 block chain financial service user, and the distribution characteristics comprise block chain financial service user distribution characteristics and service stage distribution characteristics on the basis that the plurality of block chain financial service users are X block chain financial service users.

The first session activity description phrase and the second session activity description phrase both carry distribution characteristics, and the distribution characteristics include service phase distribution characteristics on the basis that the multiple blockchain financial service users are 1 blockchain financial service user, and the distribution characteristics include blockchain financial service user distribution characteristics and service phase distribution characteristics on the basis that the multiple blockchain financial service users are X blockchain financial service users, so that different session activity description phrases can be distinguished by adopting different distribution characteristic rules under two types of conditions of 1 blockchain financial service user and multiple blockchain financial service users, the distribution characteristics of the session activity description phrases are not consistent with each other, and the accuracy of the session activity description phrases is favorably guaranteed.

In some independent embodiments, the preference profile knowledge base is obtained by a preference profile processing network, and the distribution features are optimized in parallel with variable data of the preference profile processing network in a debugging step of the preference profile processing network until debugging of the preference profile processing network tends to be stable.

The preference portrait knowledge base is obtained by the preference portrait processing network, the distribution characteristics are optimized in parallel with variable data of the preference portrait processing network in the debugging step of the preference portrait processing network until the debugging of the preference portrait processing network tends to be stable, and the distribution characteristics are debugged in parallel with an AI algorithm, so that the detail output quality of the distribution characteristics can be improved, and the precision and the richness of the preference portrait knowledge base can be guaranteed.

In some embodiments, the preference profile details field of the blockchain financial transaction user in the service session big data comprises: in the service session big data, a first relative distribution of service preference items of the blockchain financial business user and a session state of the blockchain financial business user, wherein the session state comprises a second relative distribution of a plurality of business activity links of the blockchain financial business user.

Applied to the embodiment, the preference portrait detail field of the block chain financial service user in the service session big data comprises: the first relative distribution of the service preference items of the blockchain financial service users and the session state information of the blockchain financial service users are obtained in the service session big data, and the session state information comprises the second relative distribution of a plurality of business activity links of the blockchain financial service users, so that the preference portrait of the blockchain financial service users can be reflected through the relative distribution of the service preference items and the business activity links, and the accuracy of preference portrait detail fields is favorably ensured.

In some independent embodiments, the preference representation knowledge base is obtained by a preference representation processing network, and the preference representation processing network and the verification network are obtained by positive and negative case debugging.

The method is applied to the embodiment, the preference portrait processing network and the verification network are jointly debugged through positive and negative example debugging, so that the preference portrait processing network and the verification network can make up for deficiencies in a joint debugging step, and the network quality of the preference portrait processing network is further favorably guaranteed.

In some independent embodiments, the idea of positive and negative case debugging is as follows: obtaining an authenticated preference profile knowledge base of a plurality of authenticated blockchain financial service users about authenticated preference profile tags; the authenticated preference sketch knowledge base comprises a set number K0 of authenticated service session big data, and a priori annotation is added to the authenticated preference sketch knowledge base and indicates the authenticity of the authenticated preference sketch knowledge base generated and obtained by a preference sketch processing network; sequentially carrying out division treatment on big data of each authenticated service session in the authenticated preference portrait knowledge base to obtain authenticated visual information; the authenticated visual information comprises K0 groups of unit relation networks, the unit relation networks are obtained by combining knowledge units, the knowledge units comprise service preference items and business activity links, the unit relation networks comprise knowledge unit session activity description phrases of the knowledge units, and the distribution characteristics of the knowledge units are obtained by combining the distribution characteristics of a plurality of authenticated block chain financial business users at corresponding knowledge units respectively; verifying the authenticated visual information and the authenticated preference image tag based on a verification network to obtain verification data; wherein the verification data comprises first verification content of the authenticated preference profile knowledge base and second verification content, the first verification content represents a quantization score that the authenticated preference profile knowledge base estimates to be output by the preference profile processing network, and the second verification content represents a quantization score that the authenticated preference profile knowledge base belongs to the authenticated preference profile label; based on the prior annotation, the first verification content and the second verification content, the variable data of one of the preference portrait processing network and the verification network is improved.

The method is applied to the embodiment, the authenticated preference portrait detail field is divided and treated into authenticated visual information, the verification of the preference portrait knowledge base can be converted into the verification of the visual information, and the debugging difficulty and the deployment complexity of a verification network can be obviously reduced.

In some independent embodiments, based on the authenticated preference profile knowledge base collected from the historical financial transaction environment, the distribution characteristics of the knowledge units are combined from the distribution characteristics of the authenticated blockchain financial transaction users at the corresponding knowledge units according to the set relationship of the authenticated blockchain financial transaction users.

The method is applied to the embodiment, the distribution characteristics of the knowledge unit are obtained by combining the distribution characteristics of a plurality of authenticated block chain financial service users at the corresponding knowledge unit according to the set relationship of the plurality of authenticated block chain financial service users, so that the preference portrait processing network regards the conditions of different queues actually corresponding to the same authenticated preference portrait knowledge base as different authentication templates in the debugging step, sample amplification interference can be realized, and the anti-interference performance of the network can be guaranteed.

A blockchain financial services system comprising: a memory for storing an executable computer program, a processor for implementing the above method when executing the executable computer program stored in the memory.

A computer-readable storage medium, on which a computer program is stored which, when executed, performs the above-described method.

Drawings

Fig. 1 is a schematic diagram illustrating one communication configuration of a blockchain financial services system in which embodiments of the present disclosure may be implemented.

FIG. 2 is a flow diagram illustrating a method of user representation processing based on block chain finance in which embodiments of the disclosure may be implemented.

FIG. 3 is an architectural diagram illustrating an application environment for a method of user representation processing based on block chain finance in which embodiments of the disclosure may be implemented.

Detailed Description

The technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the drawings in the embodiments of the present disclosure. To further clarify the objects, technical solutions and advantages of the present disclosure, the present disclosure will be described in further detail with reference to the accompanying drawings, and the described embodiments should not be construed as limiting the present disclosure, and all other embodiments that can be obtained by a person of ordinary skill in the art without making an inventive effort fall within the scope of protection of the present disclosure. In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. The terminology used herein is for the purpose of describing embodiments of the disclosure only and is not intended to be limiting of the disclosure.

Fig. 1 is a block diagram illustrating a communication configuration of a blockchain financial services system 100 that can implement an embodiment of the present disclosure, where the blockchain financial services system 100 includes a memory 101 for storing an executable computer program, and a processor 102 for implementing a user representation processing method based on blockchain finance in the embodiment of the present disclosure when the executable computer program stored in the memory 101 is executed.

Fig. 2 is a flowchart illustrating a user representation processing method based on blockchain finance, which may be implemented by the blockchain financial service system 100 shown in fig. 1 and further may include the technical solutions described in the following related steps.

The Process11: obtaining a first session activity description phrase for reflecting big data of a plurality of blockchain financial business users in a plurality of groups of service sessions, and obtaining a second session activity description phrase for reflecting labels of the plurality of blockchain financial business users relative to target preference portrayal.

For some possible examples, the number of multiple blockchain financial transaction users K1 and the target preference profile tag may be agreed upon by the fsp platform system prior to mining the preference profile. For example, the financial service provider platform system may agree that the target preference portrait label is "cross-border e-commerce information security", and set the number K1 of the plurality of block chain financial service users to 2; for another example, the financial service provider platform system may agree that the target preference portrait label is "financial service push interception", and set the number K1 of multiple block chain financial service users to be 1; for another example, the platform system of the financial service provider may agree that the target preference portrait label is "private user intrusion prevention", and set the number K1 of the multiple financial service users with blockchain to 3. Further, the service session big data may be a service session record/log, and the session activity description phrase may be understood as a session activity feature or a session activity vector. A preferred portrait label may also be understood as a preferred portrait category.

For other possible examples, the target preference profile tag may be contracted by the fsp platform system prior to mining the preference profile, and the number of multiple blockchain fsp users K1 may be derived based on the target preference profile tag adaptive parsing. For example, the financial service provider platform system may agree that the target preference portrait label is "anonymous interaction", and based on the target preference portrait label, the number K1 of the plurality of block chain financial service users may be obtained by adaptive analysis as 2; for another example, the financial service provider platform system may agree that the target preference portrait label is "service authority authentication", and then the number K1 of the plurality of block chain financial service users may be obtained by adaptive analysis based on the target preference portrait label as 2; for another example, the financial service provider platform system may agree that the target preference portrait label is "service information recommendation", and then the number K1 of the plurality of blockchain financial service users may be obtained by adaptive analysis based on the target preference portrait label as 1.

For some other possible examples, the target preference profile tag may be agreed upon by the fsp platform system before mining the preference profile, and the number K1 of the plurality of blockchain fsp users may be obtained based on the target preference profile tag adaptive parsing, and the obtained number K1 of adaptive parsing may be modified in response to an indication of an adjustment of the adaptively parsed number K1 by the fsp platform system. For example, the financial facilitator platform system may agree that the target preference portrait label is "private user intrusion protection", then the number K1 of the multiple block chain financial service users may be obtained by adaptive parsing based on the target preference portrait label, and the number K1 obtained by adaptive parsing is modified to 4 in response to an adjustment instruction of the financial facilitator platform system to the number K1; for another example, the financial service provider platform system may agree that the target preference portrait label is "service browsing", and the number K1 of the multiple block chain financial service users may be obtained by adaptive analysis based on the target preference portrait label as 1, and the number K1 obtained by adaptive analysis is modified to 2 in response to an adjustment instruction of the financial service provider platform system to the number K1 obtained by adaptive analysis.

It is to be understood that the plurality of blockchain financial transaction users may all be individual users, and the plurality of blockchain financial transaction users may also include both individual users and enterprise users. For example, a target preference profile tag may be set to "government and enterprise item recommendation," and the plurality of users of the blockchain financial service may include personal users and enterprise users.

For some possible examples, the number of sets of service session big data K2 may be set in advance.

For some possible examples, a first session activity description phrase for big data per service session per blockchain financial transaction user may be obtained. For example, under the condition that the number K1 of the plurality of blockchain financial service users is 1 (for example, under the condition of preference portrait mining for 1 blockchain financial service user), the first session activity description phrases of the 1 blockchain financial service user in each service session big data can be obtained; for another example, under the condition that the number K1 of the plurality of block chain financial transaction users is 2 (i.e. under the condition of mining the preference profile of 2 block chain financial transaction users), the first session activity description phrase of each block chain financial transaction user in the big data of each service session may be obtained, and these two block chain financial transaction users may be respectively referred to as "clientA" and "clientB", so that the first session activity description phrase of "clientA" in the big data of each service session may be obtained, and the first session activity description phrase of "clientB" in the big data of each service session may be obtained.

For some possible examples, it can be understood that the service session big data is included in the preference profile knowledge base that the disclosed user profile processing method based on blockchain finance disclosed in the embodiments of the present disclosure is expected to output, that is, the service session big data is not generated in essence when the first session activity description phrase is obtained, and the first session activity description phrase can be regarded as the original session activity description phrase of each blockchain financial business user in each service session big data. In general, the resulting first session activity description phrase may be mined based on an iterative analysis task. It is understood that the iterative analysis task is an algorithm based on statistical selection, for example, the iterative analysis task may be implemented based on a stochastic process (stochastic process), which is not described herein.

Under some possible design ideas, mining can be performed on the basis of the number K2 in multiple rounds of iterative analysis tasks respectively to obtain first basic description phrases for reflecting K2 service session big data, the scale constraint value of the first basic description phrases is the same as the number of the iterative analysis tasks, and the task configurations of the iterative analysis tasks are not consistent with each other. Further, based on the number K1 and the first basic description phrase, K3 first session activity description phrases are obtained, and the number K3 is the set operation result of the number K1 and the number K2. For example, the number K2 of the multiple sets of service session big data may be regarded as num, and values of task configuration configurations of the iterative analysis tasks are not limited, then num rounds are performed on the iterative analysis tasks with task configuration of 1 to obtain a 1-dimensional array with size constraint value of num, and such repetition is performed, so that each iterative analysis task with task configuration configurations of 10, 100, and 1000 may mine the obtained 1-dimensional array with size constraint value of num, and members at the same position on the 1-dimensional array with size constraint value of num obtained by mining each of the 4 iterative analysis tasks are combined, so as to obtain num first base description phrases with size constraint value of 4, and the num first base description phrases are respectively matched with num service session big data one by one, that is, the first base description phrase corresponds to the first service session big data, and the second first base description phrase corresponds to the second service session big data, based on which the first base description phrase corresponds to the num big service session data.

In some examples, the size constraint value of the first base description phrase obtained by the mining may be regarded as (number 0), and thus the first base description phrase for reflecting the large data of the multiple groups of service sessions may be regarded as (num, restriction 0). Further, the first base description phrase (num, restriction 0) may be mined (e.g., the first base description phrase may be identified based on an AI model) to adjust the scale of the previous first base description phrase (num, restriction 0). The number of first base description phrases after mining is num. Because the task configurations of the iterative analysis tasks are mutually inconsistent and the execution of each iterative analysis task can obtain the significant characteristics of the big data of each service session, the precision and the credibility of each first session activity description phrase can be guaranteed.

Under some possible design considerations, after obtaining the first basic descriptive phrase reflecting the K2 service session big data, whether to continue processing the first basic descriptive phrase reflecting the respective service session big data may be determined based on whether the number K1 is equal to 1 or greater than 1, so as to obtain the first basic descriptive phrases of the plurality of blockchain financial business users in the respective service session big data. For example, on the basis that the number K1 is equal to 1, the condition that the portrait mining is preferred to be 1 blockchain financial service user can be determined, and then the first basic description phrase obtained by mining and reflecting the big data of each service session can be directly regarded as the first basic description phrase of the 1 blockchain financial service user in the big data of each service session; for another example, on the basis that the number K1 is greater than one, the condition that the preferred image is mined into a plurality of block chain financial service users may be determined, then the first basic description phrases obtained by the mining and reflecting the preferred images may be respectively subjected to K1-round copying to obtain the first basic description phrases of the block chain financial service users in the service session big data, and on the basis that the number K1 is 2, the first basic description phrases reflecting the 1 st service session big data may be converted into two first basic description phrases, and the two first basic description phrases may be respectively reflected in the first basic description phrases of the two block chain financial service users in the 1 st service session big data.

Under some possible design ideas, in order to distinguish different first basic description phrases on the basis of 1 blockchain financial service user and X blockchain financial service users, the relative distribution of each first basic description phrase can be mined on the basis of the first basic description phrases to obtain the corresponding first session activity description phrase, in other words, the first session activity description phrases are paired with distribution features, and the distribution features are mutually inconsistent. Generally, on the basis that the plurality of blockchain financial service users are 1 blockchain financial service user, the distribution characteristics include service phase distribution characteristics, in other words, on the basis that the plurality of blockchain financial service users are 1 blockchain financial service user, different first base description phrases are distinguished by mining service session big data of different service phases, so as to obtain a first session activity description phrase. For example, still taking num service session big data as an example, on the basis of 1 blockchain financial service user, the first basic descriptive phrase of the num service session big data may be sequentially matched to the service phase distribution features (for example, 1, 2, etc., num), so as to obtain the first session activity descriptive phrase for reflecting the num service session big data of the 1 blockchain financial service user. On the basis that the plurality of blockchain financial service users are X blockchain financial service users, the distribution characteristics may include service stage distribution characteristics and blockchain financial service user distribution characteristics, in other words, on the basis that the plurality of blockchain financial service users are X blockchain financial service users, not only service session big data of different service stages need to be mined, but also a plurality of blockchain financial service users in each service session big data need to be mined, so as to distinguish different first basic description phrases, thereby obtaining a first session activity description phrase. For example, still taking num service session big data as an example, on the basis of X blockchain financial service users, the first basic descriptive phrase of the 1 st service session big data may be matched to the service phase distribution feature (e.g., 1), and further, the multiple blockchain financial service users of the 1 st service session big data may be matched to the blockchain financial service user distribution features (e.g., 1, 2, etc.), so that the service phase distribution feature and the blockchain financial service user distribution feature are combined to be regarded as the distribution features, so that the first session activity descriptive phrases reflecting the multiple blockchain financial service users in the 1 st service session big data are respectively paired with different distribution features (e.g., 1-1, 1-2, etc.). Further, the first base description phrase of the 2 nd service session big data may be matched to a business phase distribution feature (e.g., 2), and further the plurality of blockchain financial business users may be matched to a blockchain financial business user distribution feature (e.g., 1, 2, etc.) of the 1 st service session big data, so that the business phase distribution feature and the blockchain financial business user distribution feature are combined to be regarded as a distribution feature, such that the first session activity description phrases reflecting the plurality of blockchain financial business users in the 2 nd service session big data are respectively paired with different distribution features (e.g., 2-1, 2-2, etc.), and other service session big data may be repeated as such.

Further, the distribution characteristics are only taken as examples, generally speaking, a preference portrait processing network can be debugged in advance, and the distribution characteristics can be optimized in parallel with variable data of the preference portrait processing network in the debugging step of the preference portrait processing network until the debugging of the preference portrait processing network tends to be stable, and then the adjusted distribution characteristics can be used. Based on the method, under the two conditions of 1 blockchain financial service user and a plurality of blockchain financial service users, different session activity description phrases can be distinguished based on different distribution characteristic rules, so that the distribution characteristics of the session activity description phrases are inconsistent, and the accuracy of the session activity description phrases is guaranteed.

For some possible examples, similar to the first session activity description phrase, for a second session activity description phrase, a second session activity description phrase may be obtained for each blockchain financial service user that portrays tags relative to target preferences. For example, under the condition that the number K1 of the multiple blockchain financial service users is one (under the condition of preference profile mining for 1 blockchain financial service user), a second session activity description phrase of the 1 blockchain financial service user relative to the target preference profile tag can be obtained; for another example, under the condition that the number K1 of the plurality of block chain financial service users is two (under the condition of preference profile mining for two block chain financial service users), a second session activity description phrase of each block chain financial service user with respect to the target preference profile tag may be obtained, and these two block chain financial service users may be respectively referred to as "clientA" and "clientB", and then a second session activity description phrase of "clientA" with respect to the target preference profile tag may be obtained, and a second session activity description phrase of "clientB" with respect to the target preference profile tag may be obtained.

For some possible examples, the target preference profile tag may be agreed upon by the financial facilitator platform system, and after determining the target preference profile tag, recognition may be performed based on the target preference profile tag to obtain a second session activity description phrase.

Under some possible design ideas, description transformation can be carried out on the target preference portrait label to obtain a second basic description phrase, and further, a number K1 of second conversation activity description phrases are obtained based on the number K1 and the second basic description phrase. It will be appreciated that the above description transform functions to convert a target preference portrait label into an array field. For example, tag array fields of different preference portrait tags may be set in advance, for example, on the basis of twenty-six different preference portrait tags, tag array fields of twenty-six preference portrait tags may be set in advance (the scale constraint value of each tag array field may be 200), and after a target preference portrait tag is determined, a tag array field of a preference portrait tag that is consistent with the target preference portrait tag may be regarded as a second basic description phrase of the target preference portrait tag; for example, the target preference image tag may be mined first and then subjected to linear regression processing to obtain the second basic phrase of the target preference image tag, and if twenty-six different preference image tags are used, the target preference image tag may be adjusted to a twenty-six-dimensional array first and then subjected to linear regression processing to obtain the second basic phrase of the target preference image tag.

Under some possible design ideas, similar to obtaining the first session activity description phrase, after obtaining the second basic description phrase reflecting the target preference portrait tag, it may also be determined whether to continue processing on the first basic description phrase based on whether the number K1 is equal to 1 or greater than 1, so as to obtain the second basic description phrases of the plurality of blockchain financial service users respectively relative to the target preference portrait tag. For example, on the basis that the number K1 is equal to one, the condition that the mining of the preference portrait is 1 blockchain financial service user can be determined, and then the second basic description phrase obtained by mining and reflecting the target preference portrait label can be directly regarded as the second basic description phrase of the 1 blockchain financial service user about the target preference portrait label; for another example, on the basis that the number K1 is greater than one, the condition that the preference portrait is mined into a plurality of blockchain financial service users can be determined, then K1 round copying can be performed on the second basic description phrases reflecting the target preference portrait tag, so as to obtain second basic description phrases of the plurality of blockchain financial service users relative to the target preference portrait tag, for example, on the basis that the number K1 is 2, the second basic description phrases reflecting the target preference portrait tag can be converted into two second basic description phrases, and the two second basic description phrases respectively reflect the second basic description phrases of the two blockchain financial service users relative to the target preference portrait tag.

Under some possible design ideas, similar to obtaining the first session activity description phrase, in order to distinguish different second basic description phrases on the basis of 1 blockchain financial business user and X blockchain financial business users, the relative distribution of each second basic description phrase can be mined on the basis of the second basic description phrases, and the corresponding second session activity description phrases are obtained. In other words, similar to the first session activity descriptive phrase, the second session activity descriptive phrase is also paired with a distribution characteristic, and the distribution characteristics are mutually inconsistent. It is to be understood that not only the distribution characteristics of the respective second session activity description phrase pairs are mutually inconsistent, but also the distribution characteristics of the second session activity description phrase pairs and the distribution characteristics of the first session activity description phrase pairs are mutually inconsistent. Generally, on the basis that the plurality of blockchain financial transaction users are 1 blockchain financial transaction user, the distribution characteristics include a business phase distribution characteristic, in other words, on the basis that the plurality of blockchain financial transaction users are 1 blockchain financial transaction user, the second session activity description phrase can be distinguished from the first session activity description phrase of the big data of the different service sessions at the business phase level. For example, taking num service session big data as an example, on the basis of 1 blockchain financial service user, a first basic description phrase of the num service session big data may be sequentially matched to the service phase distribution feature (e.g., 1, 2, etc., num) to obtain a first session activity description phrase reflecting the num service session big data of the 1 blockchain financial service user, and then a second basic description phrase of the target preference profile tag may be matched to the service phase distribution feature (e.g., num + 1) to obtain a second session activity description phrase of the 1 blockchain financial service user with respect to the target preference profile tag.

On the basis that the plurality of blockchain financial service users are X blockchain financial service users, the distribution characteristics may include a service stage distribution characteristic and a blockchain financial service user distribution characteristic, in other words, on the basis that the plurality of blockchain financial service users are X blockchain financial service users, a distinction needs to be made between a service stage level and a blockchain financial service user level at the same time. For example, still taking num service session big data as an example, on the basis of X block chain financial service users, a second basic description phrase of a plurality of block chain financial service users with respect to a target preference profile tag may be first matched to a service phase distribution feature (e.g., num + 1), and further a second basic description phrase of a 1 st block chain financial service user with respect to a target preference profile tag may be matched to a block chain financial service user distribution feature (e.g., 1), and further a second basic description phrase of a 2 nd block chain financial service user with respect to a target preference profile tag may be matched to a block chain financial service user distribution feature (e.g., 2), which are repeated, so as to combine the service phase distribution feature and the block chain financial service user distribution feature, so as to reflect that the plurality of block chain financial service users are respectively paired with different distribution features (e.g., num +1-1, num 1-2, etc.) with respect to the target preference profile tag. Generally speaking, a preference portrait processing network can be debugged in advance, and the distribution characteristics can be optimized in parallel with variable data of the preference portrait processing network in the debugging step of the preference portrait processing network until the debugging of the preference portrait processing network tends to be stable, and then the adjusted distribution characteristics can be used. Based on the above, under two types of conditions of 1 blockchain financial service user and a plurality of blockchain financial service users, different distribution characteristic rules are adopted to distinguish different session activity description phrases, so that the distribution characteristics of the session activity description phrases are inconsistent with each other, and the accuracy of the session activity description phrases is favorably ensured.

For some possible examples, the first session activity description phrase and the second session activity description phrase both carry distribution characteristics, and the distribution characteristics include business phase distribution characteristics on the basis that the plurality of blockchain financial business users are 1 blockchain financial business user, and the distribution characteristics include blockchain financial business user distribution characteristics and business phase distribution characteristics on the basis that the plurality of blockchain financial business users are X blockchain financial business users.

Further, the distribution characteristics may be different from each other for the convenience of distinction. Taking num service session big data and R (R equals to 1, for example, R is greater than 1) blockchain financial service users as an example, session activity description phrases of (num + 1) × R can be obtained, where num × R session activity description phrases are included for reflecting the first session activity description phrases of the respective blockchain financial service users in the respective service session big data, and R session activity description phrases are included for reflecting the target preference profile labels of the respective blockchain financial service users.

It is understood that, for each blockchain financial service user, the first basic description phrase of the blockchain financial service user in the service session big data and the second basic description phrase of the blockchain financial service user relative to the target preference portrait label can be regarded as the original service stage activity characteristics of the blockchain financial service user in different service stages. Still with num service session big data, for the r blockchain financial business user, the first basic description phrase of the num service session big data and the second basic description phrase of the target preference portrait label can be regarded as the original business phase activity characteristics of the num service session big data and the num +1 business phase. Further, under the condition of preference portrait mining of 1 blockchain financial service user, on the basis that the interval of the service phase y is 1 to num, the service phase distribution characteristic characteriodic of the ith service phase and the activity characteristic of the ith original service phase can be weighted to obtain a first session activity description phrase of the ith service phase, and on the basis that the service phase y is num +1, the service phase distribution characteristic characteriodic of the ith service phase and the activity characteristic of the ythe original service phase can be weighted to obtain a second session activity description phrase of the ith service phase. Similarly, in the case of mining the preference profile of multiple blockchain financial service users, the blockchain financial service user distribution feature vector of the r-th blockchain financial service user may be combined with the service phase distribution feature of the y-th service phase to obtain a distribution feature CC (vector) of the r-th blockchain financial service user in the y-th service phase, where CC represents the combining process, and then on the basis of the interval of the service phase y being 1 to num, the distribution feature location (y, r) of the r-th blockchain financial service user in the y-th service phase may be weighted with the original service phase activity feature of the r-th blockchain financial service user in the y-th service phase to obtain the first session activity description phrase of the r-th blockchain financial service user in the y-th service phase, and on the basis of the service phase y being num +1, the distribution feature location (y, r) of the r-th blockchain financial service user in the y-th service phase and the second session activity feature phrase of the second blockchain financial service user in the y-th service phase may be weighted with the distribution feature phrase of the first blockchain financial service user in the original service phase.

The Process12: and performing linkage analysis processing based on the first session activity description phrase and the second session activity description phrase to obtain the global activity description phrase distribution of each block chain financial business user in each service session big data.

For some possible examples, the kind of the linkage analysis processing (modeling processing) is associated with the number K1 of the plurality of blockchain financial service users, and in general, on the basis that the number K1 of the plurality of blockchain financial service users is 1, the linkage analysis processing includes generating a phase transfer network between service session big data, so as to promote uninterrupted service phases between the service session big data through the construction phase transfer network, thereby contributing to ensuring the authenticity of the preference profile knowledge base, and on the basis that the number K1 of the plurality of blockchain financial service users is X, the linkage analysis processing includes generating an activity transfer network between the plurality of blockchain financial service users in each service session big data, and generating a phase transfer network between the service session big data, thereby ensuring the quality of interaction analysis between the blockchain financial service users through the construction activity transfer network, and promoting uninterrupted service phases between the service session big data through the construction phase transfer network, thereby contributing to ensuring the authenticity of the preference profile knowledge base.

For some possible examples, on the basis that the number K1 of the multiple blockchain financial service users is 1, only a phase pass-through network needs to be constructed, based on which, the 1 blockchain financial service user can be directly screened to be regarded as a target blockchain financial service user, and the first session activity description phrase and the second session activity description phrase corresponding to the target blockchain financial service user are regarded as phase session activity description phrases of the target blockchain financial service user at different service phases. For example, still taking num service session big data as an example, the first session activity description phrase of the target blockchain financial service user in the 1 st service session big data may be regarded as a first phase session activity description phrase, the first session activity description phrase of the target blockchain financial service user in the 2 nd service session big data may be regarded as a second phase session activity description phrase, and so on, the first session activity description phrase of the target blockchain financial service user in the num service session big data may be regarded as a num phase session activity description phrase, and the second session activity description phrase of the target blockchain financial service user with respect to the target preference portrait tag may be regarded as a num +1 phase session activity description phrase. Further, each service phase may be sequentially screened to be regarded as the current service phase, and the phase session activity description phrase of the current service phase is screened to be regarded as the current phase session activity description phrase, and the global activity description phrase distribution corresponding to the current service phase activity feature is obtained by using each service phase activity feature template and the PPMCC of the current service phase activity feature. In other words, on the basis that the u-th stage session activity description phrase is regarded as the current service stage activity feature, the stage session activity description phrases of the target blockchain financial service user in each service stage (1 to num + 1) can be regarded as the service stage activity feature templates, and based on the PPMCC between the service stage activity feature templates and the u-th stage session activity description phrases, the global activity description phrase distribution corresponding to the u-th stage session activity description phrase is obtained, so that in the case of mining the preference portrait of 1 blockchain financial service user, num +1 global activity description phrase distributions can be further obtained, where the num +1 global activity description phrase distributions include: session activity description phrases of num service session big data respectively after the 1 blockchain financial service user convergence phase delivery network, and session activity description phrases of the target preference portrait label after the 1 blockchain financial service user convergence phase delivery network. It can be understood that, in order to facilitate distinction from the construction steps of the subsequent activity delivery network, in the service phase construction, the current service phase may be regarded as a first current service phase, the phase session activity description phrase of the current service phase may be regarded as a first current service phase activity feature, and the service phase activity feature template may be regarded as a first service phase activity feature template.

Under some possible design ideas, in order to improve the timeliness and the precision of preference portrait mining, a preference portrait processing network (such as a GCN model) can be debugged in advance, and the preference portrait processing network can comprise a linkage analysis processing layer, and the linkage analysis processing layer can further comprise a business phase analysis layer. Further, on the basis of screening the y-th stage session activity description phrase as the current business stage activity feature, the PPMCC (pearson correlation coefficient) between the query session activity description phrase corresponding to the y-th stage session activity description phrase and the key session activity description phrase of the y 0-th (value interval 1 to num + 1) stage session activity description phrase may be obtained. After obtaining the PPMCC, the value session activity description phrases of the y0 th (value interval 1 to num + 1) stage session activity description phrase can be summed based on the PPMCC, so as to obtain the global activity description phrase distribution after the y stage session activity description phrase fusion stage delivery network.

Under some possible design ideas, the business phase analysis layer may be formed by cascading a D (D is greater than or equal to 1) layer CNN, further, after obtaining the global activity description phrase distribution output by the f-th layer CNN, it may be regarded as the input of the f + 1-th layer CNN, and perform the business phase construction process again, so as to obtain the global activity description phrase distribution output by the f + 1-th layer CNN, and repeat this way, and then may regard the global activity description phrase distribution output by the last layer CNN as the final global activity description phrase distribution. Furthermore, after the final global activity description phrase distribution is obtained, the num +1 final global activity description phrase distribution associated with the target preferred portrait tag may be filtered before subsequent processing 13 preferred portrait mining because the 1 st through num final global activity description phrase distributions have been fully matched to the target preferred portrait tag.

For some possible examples, on the basis that a plurality of blockchain financial service users are X blockchain financial service users, a stage delivery network and an active delivery network need to be constructed, and the active delivery network and the stage delivery network may be constructed in sequence, for example, the active delivery network may be constructed first, and then the stage delivery network may be constructed; for another example, the stage delivery network may be constructed first, and then the active delivery network may be constructed. Further, the session activity description phrase out feature of a previously constructed relationship is a session activity description phrase material of a subsequently constructed relationship. In other words, on the basis that the linkage analysis processing includes constructing an activity delivery network and a phase delivery network, a first delivery description may be constructed based on the first session activity description phrase and the second session activity description phrase to obtain a session activity description phrase out feature of the first delivery description, and then a second delivery description may be constructed based on the session activity description phrase out feature to obtain a global activity description phrase distribution. It is to be understood that the first transfer is described as an active transfer net and the second transfer is described as a phase transfer net, and that for another example, the first transfer is described as a phase transfer net and the second transfer is described as a phase transfer net.

Under some possible design ideas, in order to improve the timeliness and the precision of preference portrait mining, a preference portrait processing network can be debugged in advance, and the preference portrait processing network can comprise a linkage analysis processing layer which can comprise a business phase analysis layer and an activity influence analysis layer. For example, the business phase analysis layer and the activity impact analysis layer may both be generated based on CNN. Similar to the preferred portrait mining situation of 1 blockchain financial service user, in the preferred portrait mining situations of a plurality of blockchain financial service users, one blockchain financial service user can be screened as a target blockchain financial service user. For example, the R-th blockchain financial transaction user of the R blockchain financial transaction users may be filtered as the target blockchain financial transaction user. Further, the first session activity description phrase and the second session activity description phrase corresponding to the target blockchain financial service user may be regarded as phase session activity description phrases of the target blockchain financial service user at different service phases respectively. When the activity transfer network is generated firstly, similar to the generation stage transfer network, after the stage session activity description phrases of target block chain financial service users in different service stages are obtained, each service stage can be screened to be respectively regarded as a current service stage, the stage session activity description phrases of the current service stage are screened to be regarded as current service stage activity characteristics, and global activity description phrase distribution corresponding to the current service stage activity characteristics is obtained by utilizing each service stage activity characteristic template and the PPMCC of the current service stage activity characteristics.

It is to be understood that, for the purpose of distinguishing from the above-mentioned construction steps of the phase delivery network, the current business phase may be regarded as a second current business phase, the phase session activity description phrase of the second current business phase may be regarded as a second current business phase activity feature, and the business phase activity feature template may be named as a second business phase activity feature template. Under some possible design ideas, after obtaining global activity description phrase distributions after each blockchain financial business user fuses the activity delivery networks in each business phase respectively, the global activity description phrase distributions can be regarded as session activity description phrase raw materials of the generation phase delivery networks to continue to generate the phase delivery networks.

Under some possible design considerations, the network node 1 for generating the active delivery network and the network node 2 for generating the phase delivery network may be combined into a group CNN to jointly generate the active delivery network and the phase delivery network, and then the relevant model layer may include D groups CNN. Further, for the r-th blockchain financial service user in the service session big data of the y-th service phase, after the global activity description phrase output by the f-th group CNN is distributed, the global activity description phrase output by the f-th group CNN can be regarded as the input of the f + 1-th layer CNN, and the service phase construction process is executed again, so that the global activity description phrase distribution output by the f + 1-th layer CNN is obtained, and repeated, and further, the global activity description phrase distribution output by the last layer CNN can be regarded as the final global activity description phrase distribution. After the final global activity description phrase distribution is obtained, the num +1 final global activity description phrase distribution associated with the target preferred portrait label may be filtered before subsequent processing 13 preferred portrait mining, given that the 1 st through num final global activity description phrase distributions have been sufficiently matched to the target preferred portrait label.

For example, the block chain financial service user may be first screened as a target block chain financial service user, the first session activity description phrase and the second session activity description phrase corresponding to the target block chain financial service user may be regarded as the phase session activity description phrases of the target block chain financial service user in different service phases, then each service phase may be screened as the current service phase, the phase session activity description phrase of the current service phase may be screened as the current service phase activity feature, and then the global activity description phrase distribution corresponding to the current service phase activity feature may be obtained by using each service phase activity feature template and the PPMCC of the current service phase activity feature. On the basis of constructing a stage transmission network, the service stage activity characteristic template comprises stage session activity description phrases of target block chain financial service users in each service stage, and on the basis of constructing the activity transmission network, the service stage activity characteristic template comprises stage session activity description phrases of each block chain financial service user in the current service stage. Based on the method, the stage delivery network and the activity delivery network can be constructed through similar construction ideas, so that the flexibility and the intelligent degree of preference mining for different numbers of block chain financial business users can be guaranteed.

The Process13: and performing preference portrait mining based on the global activity description phrase distribution to obtain a preference portrait knowledge base of a plurality of block chain financial business users relative to a target preference portrait label.

For some possible examples, the preference profile repository includes sets of service session big data, and the service session big data carries preference profile detail fields for each blockchain financial transaction user. For example, the preference profile repository may include num service session big data, and the plurality of blockchain financial service users are R blockchain financial service users, and each service session big data includes preference profile detail fields of the R blockchain financial service users, so that a dynamic preference profile with uninterrupted service stage may be generated.

For some possible examples, to improve the preference profile mining timeliness and accuracy, the preference profile processing network may be debugged in advance, and may include a preference profile mining module, where the structure of the preference profile mining module is not limited. Further, the global activity description phrase distribution of each blockchain financial business user in each service session big data can be input to the preference profile mining module, so that a preference profile knowledge base of a plurality of blockchain financial business users relative to the target preference profile tag can be obtained. Taking num service session big data and R block chain financial service users as examples, num × R global activity description phrase distributions can be obtained, and then the num × R global activity description phrase distributions can be input to the preference portrait mining module, so that num service session big data can be obtained, and each service session big data contains preference portrait detail fields of the R block chain financial service users, so that the num service session big data can be combined according to the service stage sequence to obtain a preference portrait knowledge base.

For some possible examples, the preference profile detail field of the blockchain financial business user in the service session big data may include: in the service session big data, a first relative distribution of service preference items of the blockchain financial transaction user and a session state of the blockchain financial transaction user, and the session state may specifically include a second relative distribution of a plurality of business activity links of the blockchain financial transaction user.

For some possible examples, the preference portrait knowledge base output by the preference portrait processing network only carries preference portrait detail fields of big data of each blockchain financial business user in each service session, but does not carry clustering keywords and preference portrait scenes of each blockchain financial business user, so that after the preference portrait knowledge base is obtained, the clustering keywords of each blockchain financial business user can be adjusted as required.

It can be understood that, a first session activity description phrase for reflecting big data of a plurality of block chain financial service users in a plurality of groups of service sessions is obtained, a second session activity description phrase for reflecting a plurality of block chain financial service users relative to a target preference portrait label is obtained, further, linkage analysis processing is carried out based on the first session activity description phrase and the second session activity description phrase, global activity description phrase distribution of each block chain financial service user in each service session big data is obtained, the type of linkage analysis processing is associated with the number K1 of the plurality of block chain financial service users, preference mining is carried out based on the global activity description phrase distribution, a preference portrait knowledge base of the plurality of block chain financial service users relative to the target preference portrait label can be obtained, the preference knowledge base comprises a plurality of groups of service session big data, the service session big data comprises preference detail portrait fields of each block chain financial service user, not only can bias be obtained efficiently and accurately, but also portrait can be accurately carried out according to the number K1 of the plurality of block chain financial service users and the condition of two types of block chain financial service users. Based on the method, the timeliness and the precision of the preference portrait mining can be improved, and the flexibility of portrait analysis for different numbers of block chain financial business users can be improved.

Under some independent design ideas, the preference portrait knowledge base is obtained by a preference portrait processing network, and in order to guarantee debugging quality, the preference portrait processing network and a verification network (an authentication model) can be obtained through positive and negative example debugging, and the following scheme can be exemplarily realized.

The Process41: an authenticated preference profile knowledge base of authenticated blockchain financial business users about authenticated preference profile tags is obtained.

For some possible examples, the authenticated preference profile knowledge base includes a set number K0 of authenticated service session big data, and the authenticated preference profile knowledge base adds a priori annotations indicating the authenticity of the preference profile knowledge base generated by the preference image processing network. In general, the authenticated preference profile knowledge base may be generated by a preference profile processing network or collected in a historical financial transaction environment.

For some possible examples, several authenticated business data may be obtained that are authenticated with respect to authenticated preference profile tags. Further, the authenticated preference profile detail field of each authenticated blockchain financial transaction user in the authenticated business data may be extracted, for example, the authenticated preference profile detail field of each authenticated blockchain financial transaction user may include a service preference item of the authenticated blockchain financial transaction user and a relative distribution of the plurality of business activity links. Further, each set of authenticated transaction data may be represented as an authenticated service session big data, and the authenticated preference profile of each authenticated blockchain financial transaction user in each authenticated service session big data.

The Process42: and sequentially carrying out treatment of different authentication service conversation big data in the authenticated preference portrait knowledge base to obtain authenticated visual information.

For some possible examples, the authenticated visualization information (such as graphical Data) includes K0 sets of unit relationship networks, the unit relationship networks are obtained by combining knowledge units, the knowledge units include service preference items and business activity links, the unit relationship networks include knowledge unit session activity description phrases of each knowledge unit, and distribution characteristics of the knowledge units are obtained by combining distribution characteristics of a plurality of authenticated blockchain financial business users at corresponding knowledge units respectively. Continuing to take the example that the authenticated preference profile detail field of each authenticated blockchain financial service user includes the service preference item of the authenticated blockchain financial service user and the relative distribution of a plurality of business activity links, the E-dimensional array of the authenticated preference profile detail field may be treated as W Z-dimensional arrays (the arrays reflect the relative distribution), and E = W × Z, where W is the statistical value of the service preference item of the authenticated blockchain financial service user and the business activity links, for example, the statistical value of the service preference item of each authenticated blockchain financial service user and the business activity links is eighteen.

For some possible examples, for the condition of 1 authenticated blockchain financial transaction user, each group of the unit relationship nets only needs to represent 1 authenticated blockchain financial transaction user, so each group of the unit relationship nets is obtained by combining W knowledge units, and each knowledge unit on the unit relationship nets is reflected by the Z-dimensional array of the knowledge unit, so each group of the unit relationship nets can be a linear variable of (W, Z), and the authenticated visual information can be a linear variable of (num, W, Z) based on the linear variable.

For other possible examples, unlike the condition of 1 authenticated blockchain financial transaction user, each set of cellular relationship networks needs to represent multiple authenticated blockchain financial transaction users under the condition of multiple authenticated blockchain financial transaction users. In addition, for a plurality of authenticated blockchain financial service users in the authenticated preference profile knowledge base, if the queues are different, the verification data of the subsequent verification network may be different, and based on the fact that the authenticated preference profile knowledge base is obtained by collecting historical financial service environments, the distribution characteristics of the knowledge units are obtained by combining the distribution characteristics of the authenticated blockchain financial service users at the corresponding knowledge units according to the set relationship of the authenticated blockchain financial service users, so that the preference profile processing network takes the situations that different queues are actually matched with the same authenticated preference profile knowledge base as different authentication templates in the debugging step, and sample amplification can be achieved and the anti-interference performance of the network can be guaranteed.

Process43: and verifying the authenticated visual information and the authenticated preference image tag based on the verification network to obtain verification data.

For some possible examples, the verification data includes first verification content and second verification content of the authenticated preference profile repository, the first verification content representing a quantization score that the authenticated preference profile repository estimates to output via the preference profile processing network, the second verification content representing a quantization score that the authenticated preference profile repository belongs to the authenticated preference profile tag. It is to be appreciated that the first authentication content and the second authentication content can be expressed in percentages, and the greater the percentages, the higher the corresponding quantization scores.

The Process44: and adjusting variable data of one of the preference portrait processing network and the verification network based on the prior annotation, the first verification content and the second verification content.

In general, the verification cost of the verification network can be weighed by the first verification content and the prior annotation, the generation cost of the preference portrait processing network can be weighed by the second verification content and the prior annotation, and in the debugging step, the preference portrait processing network (the variable data of the preference portrait processing network is improved) can be debugged J2 times every time the verification network (the variable data of the verification network is improved) is debugged J1 times, for example, the preference portrait processing network is debugged 1 time every 4 times the verification network is debugged. Furthermore, by debugging the verification network, the verification capability of the verification network on the verified preference portrait knowledge base (namely, the performance of the verified preference portrait knowledge base output by the distinguishing model and the actually determined verified preference portrait knowledge base) can be improved, so that the authenticity of the preference portrait knowledge base generated by the preference portrait processing network can be improved, and by debugging the preference portrait debugging model, the authenticity of the preference portrait knowledge base generated by the preference portrait processing network can be improved, so that the verification network is prompted to guarantee the verification precision, and the verification network and the preference portrait processing network are complementary, therefore, the network quality of the preference portrait processing network can be continuously improved, and the verification network is difficult to distinguish the preference portrait knowledge base output by the preference portrait processing network from the actual preference knowledge base, so that the debugging can be stopped. It is understood that positive and negative examples of debugging may be understood as countertraining, and the technical details thereof are not described herein. In the process of mining the preference portrait, distribution characteristic labeling can be carried out, and the distribution characteristic can be optimized in parallel with variable data of the preference portrait processing network in the debugging step of the preference portrait processing network.

In this way, the preference portrait processing network and the verification network are debugged in a combined manner through positive and negative case debugging, so that the preference portrait processing network and the verification network can make up for deficiencies in a combined debugging step, and the network quality of the preference portrait processing network is further guaranteed. Furthermore, by processing the authenticated preference portrait detail field into authenticated visual information in a sub-treatment mode, the verification of the preference portrait knowledge base can be converted into the verification of the visual information, and the debugging difficulty and the deployment complexity of a verification network can be obviously reduced.

It will be appreciated that the blockchain financial services system communicates with the corresponding electronic terminals of the blockchain financial transaction users for preference mining, but does not participate in the transaction interactions between the electronic terminals of the blockchain financial transaction users to ensure "decentralized" of the blockchain financial transaction.

In some independent design concepts, after obtaining a knowledge base of preference profiles of a plurality of blockchain financial service users with respect to the target preference profile tag, the method may further include: determining a pushing interest analysis result of a designated blockchain financial business user based on the preference portrait knowledge base; and pushing the service information according to the pushing interest analysis result.

It can be appreciated that if the preference portrait carried in the preference portrait knowledge base includes a pushed aspect of the portrait, then push interest analysis can be performed, resulting in accurate push interest analysis results.

Under some independent design ideas, the push interest analysis result of the financial service user of the designated block chain is determined based on the preference portrait knowledge base, and the method can be realized through the following technical scheme: acquiring a first preference portrait knowledge set and a second preference portrait knowledge set corresponding to a target preference portrait knowledge base based on a user tag of a designated block chain financial service user, wherein the first preference portrait knowledge set comprises a knowledge characteristic matrix which does not contain an interception preference field in the target preference portrait knowledge base, and the second preference portrait knowledge set comprises a knowledge characteristic matrix which contains the interception preference field in the target preference portrait knowledge base; carrying out interest prediction on the first preference portrait knowledge set to obtain an interest keyword corresponding to the first preference portrait knowledge set; carrying out interest prediction on the second preference portrait knowledge set to obtain an interception intention field corresponding to the second preference portrait knowledge set; weighting the interception intention field and the interest keywords to obtain a push interest vector corresponding to the target preference portrait knowledge base; matching the pushed interest vectors to obtain classification probability corresponding to the target preference portrait knowledge base; and when the classification probability falls into a set probability interval, determining a pushing interest analysis result of the appointed block chain financial service user based on a pushing strategy corresponding to the set probability interval.

By means of the design, the complete and accurate push interest vector can be obtained by performing combined analysis on the push characteristics and the interception characteristics, so that a push interest analysis result is obtained in a targeted manner by combining with a push strategy corresponding to a set probability interval, and the reliability of the push interest analysis result is ensured.

Fig. 3 is a schematic block chain finance-based user representation processing method, which may be implemented according to an embodiment of the present disclosure, and the system 100 and the platform 200 may be included in the application environment. Based on this, the blockchain financial service system 100 and the financial service platform system 200 implement or partially implement the user representation processing method based on blockchain finance according to the embodiment of the disclosure when running.

The above description is only for the preferred embodiment of the present disclosure, and is not intended to limit the scope of the present disclosure.

Claims (10)

1. A user portrait processing method based on block chain finance is characterized in that the method is realized by a block chain finance service system, and the method at least comprises the following steps:

obtaining a first session activity description phrase for reflecting big data of a plurality of blockchain financial business users in a plurality of groups of service sessions, and obtaining a second session activity description phrase for reflecting labels of the plurality of blockchain financial business users relative to target preference figures;

performing linkage analysis processing by combining the first session activity description phrase and the second session activity description phrase to obtain global activity description phrase distribution of each block chain financial service user in each service session big data; wherein the type of the linkage analysis processing is associated with the number K1 of the plurality of blockchain financial service users;

performing preference portrait mining in combination with the global activity description phrase distribution to obtain a preference portrait knowledge base of the plurality of block chain financial service users relative to the target preference portrait tag; the preference portrait knowledge base covers the plurality of groups of service session big data, and the service session big data carries preference portrait detail fields of each block chain financial business user.

2. The method of claim 1, wherein the category of the linkage analysis process is associated with a number K1 of the plurality of blockchain financial transaction users, including at least one of:

on the basis that the number K1 of the plurality of block chain financial service users is 1, the linkage analysis processing comprises generating a stage transmission network among the service session big data;

on the basis that the number K1 of the plurality of blockchain financial service users is X, the linkage analysis processing comprises generating an activity transmission network among the plurality of blockchain financial service users in each service session big data and generating a stage transmission network among each service session big data, wherein X is an integer greater than 1.

3. The method of claim 2, wherein said performing a linkage analysis process in combination with said first session activity description phrase and said second session activity description phrase to obtain a global activity description phrase distribution for each of said blockchain financial transaction users in each of said service session big data based on said linkage analysis process including generating said phase delivery network comprises:

screening the blockchain financial service user to be a target blockchain financial service user, and regarding a first session activity description phrase and a second session activity description phrase corresponding to the target blockchain financial service user as phase session activity description phrases of the target blockchain financial service user in different service phases;

sequentially screening each service stage as a first current service stage, and screening a stage session activity description phrase of the first current service stage as a first current service stage activity characteristic;

obtaining global activity description phrase distribution corresponding to the first current service stage activity characteristic by utilizing each first service stage activity characteristic template and the PPMCC of the first current service stage activity characteristic respectively; the first business stage activity characteristic template covers the stage session activity description phrases of the target block chain financial business user in each business stage.

4. The method of claim 2, wherein said performing a linkage analysis process in combination with said first session activity description phrase and said second session activity description phrase to obtain a global activity description phrase distribution of each said blockchain financial transaction user in each said service session big data based on said linkage analysis process including generating said activity delivery network comprises: screening the blockchain financial service user to be a target blockchain financial service user, and regarding a first session activity description phrase and a second session activity description phrase corresponding to the target blockchain financial service user as phase session activity description phrases of the target blockchain financial service user in different service phases; screening each service stage in sequence to be regarded as a second current service stage, and screening the stage session activity description phrases of the second current service stage to be regarded as second current service stage activity characteristics respectively; obtaining global activity description phrase distribution corresponding to the second current service stage activity characteristics by utilizing each second service stage activity characteristic template and the PPMCC of the second current service stage activity characteristics; the second service stage activity feature template comprises a stage session activity description phrase of each block chain financial service user in the second current service stage;

on the basis that the linkage analysis processing includes generation of the activity delivery network and the phase delivery network, the linkage analysis processing is performed in combination with the first session activity description phrase and the second session activity description phrase to obtain global activity description phrase distribution of each blockchain financial service user in each service session big data, and the method includes: constructing a first transfer description by combining the first session activity description phrase and the second session activity description phrase, and obtaining a session activity description phrase out feature of the first transfer description; constructing a second transfer description in combination with the session activity description phrase out feature, and obtaining the global activity description phrase distribution; wherein the first transfer description is the active transfer net and the second transfer description is the phase transfer net, or the first transfer description is the phase transfer net and the second transfer description is the active transfer net.

5. The method of claim 2, wherein the preference profile knowledge base is obtained by a preference profile processing network, the preference profile processing network comprising a linkage analysis processing layer, and the linkage analysis processing layer comprising a business phase analysis layer and an activity impact analysis layer, the business phase analysis layer for generating the phase delivery network, the activity impact analysis layer for generating the activity delivery network.

6. The method of claim 1, wherein the first session activity description phrase is based on a mining of iterative analysis tasks;

wherein the obtaining a first session activity description phrase reflecting big data of a plurality of blockchain financial service users in a plurality of groups of service sessions comprises: in the multiple rounds of iterative analysis tasks, mining is respectively carried out on the basis of the number K2, and first basic description phrases used for reflecting K2 service session big data are obtained; the scale constraint value of the first basic descriptive phrase is the same as the number of the iterative analysis tasks, and the task configurations of the iterative analysis tasks are not consistent with each other; obtaining K3 first session activity description phrases in combination with the number K1 and the first basic description phrase; wherein the number K3 is a set operation result of the number K1 and the number K2.

7. The method of claim 1, wherein the second session activity description phrase is obtained in conjunction with the target preference portrait tag identification; wherein the obtaining a second session activity description phrase reflecting the plurality of blockchain financial service users relative to the target preference profile tag comprises: performing description transformation on the target preference portrait label to obtain a second basic description phrase; obtaining the number K1 of the second session activity description phrases by combining the number K1 and the second basic description phrases.

8. The method of claim 1, wherein the first session activity description phrase and the second session activity description phrase both carry a distribution characteristic; wherein, on the basis that the plurality of blockchain financial service users are 1 blockchain financial service user, the distribution characteristics comprise service stage distribution characteristics, and on the basis that the plurality of blockchain financial service users are X blockchain financial service users, the distribution characteristics comprise blockchain financial service user distribution characteristics and the service stage distribution characteristics;

wherein the preference profile knowledge base is obtained by a preference profile processing network, and the distribution characteristics are optimized in parallel with variable data of the preference profile processing network in a debugging step of the preference profile processing network until the debugging of the preference profile processing network tends to be stable.

9. The method of claim 1, wherein the preference profile detail field of the blockchain financial service user in the service session big data comprises: in the service session big data, a first relative distribution of service preference items of the blockchain financial service user and a session state of the blockchain financial service user, wherein the session state covers a second relative distribution of a plurality of business activity links of the blockchain financial service user;

the preference portrait knowledge base is obtained by a preference portrait processing network, and the preference portrait processing network and the verification network are obtained by positive and negative case debugging;

the idea of debugging the positive and negative examples is as follows: obtaining an authenticated preference profile knowledge base of a plurality of authenticated blockchain financial service users about authenticated preference profile tags; the authenticated preference sketch knowledge base comprises a set number K0 of authenticated service session big data, and a priori annotation is added to the authenticated preference sketch knowledge base and reflects the authenticity of the authenticated preference sketch knowledge base generated and obtained by the preference sketch processing network; sequentially carrying out division and treatment on each authenticated service session big data in the authenticated preference portrait knowledge base to obtain authenticated visual information; the authenticated visual information covers the K0 group of unit relation networks, the unit relation networks are obtained by combining knowledge units, the knowledge units cover service preference items and service activity links of authenticated block chain financial service users, the unit relation networks comprise knowledge unit session activity description phrases of each knowledge unit, and the distribution characteristics of the knowledge units are obtained by combining the distribution characteristics of the authenticated block chain financial service users at the positions corresponding to the knowledge units; verifying the authenticated visual information and the authenticated preference image tag based on a verification network to obtain verification data;

wherein the verification data encompasses first verification content and second verification content of the authenticated preference profile knowledge base, the first verification content reflecting a quantization score that the authenticated preference profile knowledge base estimates to output by the preference profile processing network, the second verification content reflecting a quantization score that the authenticated preference profile knowledge base belongs to an authenticated preference profile tag; improving variable data of one of the preference representation processing network and the verification network in combination with the prior annotation, the first verification content and the second verification content;

on the basis that the authenticated preference portrait knowledge base is obtained by collecting historical financial service environments, the distribution characteristics of the knowledge units are obtained by combining the distribution characteristics of the authenticated blockchain financial service users at the positions corresponding to the knowledge units respectively according to the set relations of the authenticated blockchain financial service users.

10. A blockchain financial services system, comprising:

a memory for storing an executable computer program, a processor for implementing the method of any one of claims 1-9 when executing the executable computer program stored in the memory.

Images