Loan qualification assessment method adopting digital assets

Abstract

The invention relates to the technical field of loan qualification evaluation adopting digital assets, and discloses a loan qualification evaluation method adopting digital assets. According to the scheme, 30 days are taken as an observation period, a unique identifier of a borrower is set, contribution and interaction identifiers are extracted from an original digital asset record, an effectiveness judgment function is constructed, and the problem that contribution information and interaction behaviors are difficult to quantify is solved through square root integer part comparison and integer division judgment. Effective items and change trends in an observation period are counted through contribution counting, interaction counting and an increment function, constant standardization processing is introduced, information density, interaction activeness and future kinetic energy are set, a digital asset vector is formed, and data comparability and stability are enhanced. And finally, a loan qualification score is formed by the product of the three indexes, so that standardized and quantitative evaluation of the borrower digital assets is realized, and stable and explainable support is provided for loan decision.

Description

Technical Field

[0001] The present invention relates to the technical field of loan qualification evaluation using digital assets, and in particular to a loan qualification evaluation method using digital assets. Background Art

[0002] Existing credit risk assessment methods primarily rely on traditional financial data, such as a borrower's historical credit history, income verification, and balance sheet, using empirical formulas or statistical models to assess loan eligibility. With the development of the internet and information technology, more and more financial institutions are exploring the use of digital information to comprehensively assess borrowers. While traditional credit data has a long-term cumulative effect on borrowers' real-world transactions, it also has limitations in reflecting individual performance and interactions in the digital economy.

[0003] Currently, some systems attempt to extract borrowers' social network behavior, online activity records, payment behavior, and mobile device usage from the internet and big data platforms, and use machine learning models, neural networks, or cluster analysis to construct borrower credit scoring models. While these methods can capture some aspects of a borrower's digital activity, due to the diverse data sources and complex data processing, they often require a large number of empirical parameters for weighting, normalization, and preprocessing, lacking a clear and definitive mathematical model. Some existing technologies also rely on statistical weights and fuzzy factors, resulting in scoring results that often lack transparency and traceability, making it difficult to intuitively explain the inherent logical relationships between the data. Furthermore, because digital asset data often exists in the form of unstructured text and log records, conventional data preprocessing methods are subject to data redundancy, noise interference, and information silos. As a result, the final assessment model may not fully reflect the borrower's true creditworthiness. For example, some existing credit assessment schemes based on social media or online platform data generally use fuzzy logic or machine learning methods based on big data mining. These rely on pre-set weighting parameters to normalize features such as the borrower's interactive behavior, social relationships, and posted content before inputting them into the model. However, the weights and normalization parameters used in these methods are mostly empirical values, lacking strict mathematical definitions. Furthermore, the parameters fluctuate significantly under different data environments, potentially leading to insufficient model stability and difficulty in reproducibly verifying the scoring results. Furthermore, because existing methods often rely on readily available text processing techniques during the feature extraction phase, these methods have poor applicability across different languages ​​and domains, and can impact the fidelity of textual information. In the digital asset sector, there is currently no unified standard for clearly defining digital assets. In particular, the quantification of original content or interactive behaviors posted online by borrowers in credit assessments is relatively vague. Existing technologies generally describe content contributions and interactive behaviors in an abstract manner, without providing a complete computational process for extracting specific numerical values ​​from text or interactive logs. Due to the lack of clear metrics, existing methods are prone to data bias when processing original content and interactive behaviors, resulting in insufficient interpretability and stability in the final loan qualification scores.

[0004] Therefore, this case aims to propose a loan qualification assessment method using digital assets, which can accurately extract the specific numerical values ​​of original contributions and interactive behaviors in digital assets, and establish a unified and clear data model and graph structure through self-created mathematical formulas, thereby providing a technical solution for loan qualification assessment with complete mathematical expression, transparent data processing, and traceable calculation process. Summary of the Invention

[0005] The present invention provides a loan qualification assessment method using digital assets, which helps solve the problems mentioned in the above background technology.

[0006] The present invention provides the following technical solution: a loan qualification assessment method using digital assets, comprising:

[0007] Let i be the unique integer identifier of the borrower, and satisfy i∈{1,2,...,I}; where I is the total number of borrowers;

[0008] Assume an observation period of 30 days;

[0009] The total number of digital asset records generated by borrower i in an observation period is recorded as n i , the serial number of the record is j, and j∈{1,2,...,n i};

[0010] The original digital asset record of borrower i with serial number j is recorded as d i,j , where d i,j Contains: Digital Contribution ID i,j , is a positive integer, for record d i,j The corresponding contribution value; digital interaction identifier IM i,j , is a positive integer, for record d i,j The corresponding interactive prompt value;

[0011] Set the effective contribution function and calculate the digital contribution ID in the j-th record of borrower i i,j The judgment value is:

[0012]

[0013] Among them, φ(ID i,j ) is the ID i,j The judgment value of Indicates getting ID i,j The integer part of the square root; 1 means valid, 0 means invalid;

[0014] Set the effective interaction function and calculate the digital interaction identifier IM in the j-th record of borrower i i,j The judgment value is:

[0015]

[0016] Among them, ψ(IM i,j ) is IM i,j The judgment value; mod is the remainder operation; Indicates taking IM i,j The integer part of the square root; 1 means valid, 0 means invalid;

[0017] Set up a digital contribution counting function to calculate the total count of valid contribution items in all records of borrower i in an observation period, specifically:

[0018] Among them, N i is the total count of valid contribution items in all records of borrower i in an observation period;

[0019] Set the digital interaction counting function to calculate the total count of valid interaction items in all records of borrower i in an observation period, specifically:

[0020] Among them, M i is the total count of valid interaction items in all records of borrower i in an observation period;

[0021] Set the contribution growth function to calculate the contribution growth of borrower i, specifically:

[0022] in, is the total count of valid contributions obtained by borrower i in the current observation period (t); is the total number of valid contributions obtained by borrower i in the previous observation period (t-1);

[0023] Set the digital asset vector of borrower i to: V i =(I i ,R i ,P i );

[0024] Among them, I i is the information density, The constant 5 is used for normalization; R i For the level of interactive activity, The constant 10 is used as a normalization factor; R i For future momentum, The constant 30 represents the number of days of the fixed observation period;

[0025] Set the initial loan qualification score of borrower i to S i , S i =I i ×R i ×P i .

[0026] Optionally, the digital contribution identifier specifically includes:

[0027] From the jth record d of borrower i i,j Extract the original online text content and record it as string P i,j ;

[0028] Remove string P i,jAll spaces, newlines and non-alphanumeric characters in the string are processed and recorded as P' i,j ;

[0029] Get string P' i,j The total number of characters is L i,j ;

[0030] For string P' i,j Each character in is weighted according to its position in the string, specifically:

[0031] ω(c i,j,f )=f×U(c i,j,f );

[0032] Among them, ω(c i,j,f ) is the string P' i,j The weight of the fth character in ; c i,j,f For string P' i,j The fth character in the string; f is the position of the character in the string, f∈{1,2,...L i,j}; U(c i,j,f ) is the character c i,j,f Unicode code value;

[0033] Calculate the cumulative sum of all character weights, specifically: Among them, S i,j Extract the total weight of the online original text content characters from the j-th record of borrower i;

[0034] Set the offset preset constant to C1;

[0035] ID i,j =S i,j +C1.

[0036] Optionally, the digital interactive identifier specifically includes:

[0037] From the jth record d of borrower i i,j Get the interactive log data from the set H i,j ;

[0038] The event types in the record include browsing events v, liking events l, commenting events c, and sharing events s;

[0039] Give set H i,j Each interactive event in is assigned a weight, where the browsing event v is assigned a weight v =1, the like event l is given weight a l =3, comment event c is assigned weight a c =5, the sharing event s is given weight a s =7;

[0040] Count the events in the set H i,j The number of occurrences of the browsing event is recorded as n v , the number of like events is recorded as n l , record the number of comment events as n c , the number of sharing events is recorded as n s ;

[0041] Calculating interaction points for:

[0042] Set the integral offset constant C2;

[0043] calculate

[0044] Optionally, it also includes pre-processing of the original digital asset data, specifically:

[0045] Record d i,j Do the following:

[0046] Call function φ(ID i,j ) and ψ(IM i,j ) respectively determine the value of the digital contribution identifier and the digital interaction identifier;

[0047] Construct preprocessing record as r i,j =(φ(ID i,j ),ψ(IM i,j ));

[0048] Among them, r i,k It is a binary group, the first component represents whether the digital contribution is effective; the second component represents whether the digital interaction is effective;

[0049] For record d i,j If any of the following conditions exist, the application will be deemed invalid and will not be included in the subsequent processing:

[0050] Record d i,j Missing ID i,j or IM i,j ;ID i,j ≤0 or IM i,j ≤0;

[0051] Finally, the preprocessed data set is set to

[0052] Optionally, it also includes constructing digital asset graph nodes and creating a self-created semantic association algorithm, specifically:

[0053] From the preprocessed data set Generate graph nodes:

[0054] For each record r i,j =(φ(ID i,j ), ψ(IM i,j )), denoted as graph node A i,j ;

[0055] Each node A i,j is set with the following properties:

[0056] Contribution property α i,j , α i,j =φ(ID i,j );

[0057] Interaction property β i,j , β i,j =ψ(IM i,j );

[0058] Set the semantic association function and calculate the association strength between any two nodes A i,j and A i,k , where j < k, specifically:

[0059]

[0060] where, Λ(A i,j , A i,k ) is the association strength between the two nodes A i,j and A i,k ; α i,j , α i,k are respectively the contribution properties of nodes A i,j and A i,k ; β i,j , β i,k are respectively the interaction properties of nodes A i,j and A i,k ; min(α i,j , α i,k ) returns the minimum value between α i,j and α i,k ;

[0061] When generating nodes, establish edges between nodes through the following rules:

[0062] If for any j < k, Λ(A i,j , A i,k ) > 0, then establish an edge pointing from node A i,j to node A i,k ;

[0063] Denote this edge as e i,j,k , and set the value on the edge to L i,j,k =Λ(A i,j , Ai,k );

[0064] Set the node set V i as:

[0065] Set the edge set as: E i ={e i,j,k | j < k and Λ(A i,j , A i,k ) > 0}.

[0066] Optionally, it also includes constructing a holographic map and node attribute mapping, specifically:

[0067] S11. Holographic map definition:

[0068] For each borrower i, construct a holographic map G i , G i =(V i , E i );

[0069] S12. Node attribute injection:

[0070] Set γ i,j as the loan qualification score injected into node A i,j , and γ i,j = S i ;

[0071] S13. Map generation process:

[0072] Generate nodes A i,j in the order of the preprocessing record r i,j , and only generate records that meet α i,j [[ID=6l]]= 1;

[0073] For all node pairs that meet j < k, calculate Λ(A i,j , A i,k ); if the result is greater than zero, establish the corresponding edge e i,j,k and assign the edge value L i,j,k ;

[0074] Thus, the completed map G i =(V i , E i ).

[0075] Optionally, it also includes calculating the topological value of the map and generating the loan qualification score, specifically:

[0076] For any node A i in the spectral map G i,j ∈ V i , set the topological value Θ i,j as:

[0077] Average the topological values ​​of all nodes in the holographic graph of borrower i and set the overall graph index Ω i for:

[0078] Among them, |V i | is the node set V i The total number of nodes in ;

[0079] The overall spectrum index Ω i Directly mapped to the final loan qualification score Q i , specifically: Q i =Ω i .

[0080] Optionally, it also includes two-dimensional coordinate mapping of graph nodes and data interface output, specifically:

[0081] The atlas G i Each node A i,j ∈V i Mapped onto a two-dimensional plane, where:

[0082] Horizontal coordinate x i,j Set to the node's sequence number, i.e. x i,j =j;

[0083] vertical coordinate y i,j Set to the topological value of the node, y i,j =Θ i,j ;

[0084] According to the above coordinates, each node is drawn as a point (x i,j ,y i,j );

[0085] At the same time, for any edge e in the graph i,j,k , if L i,j,k >0, then draw a straight line on the plane connecting A i,j With A i,k The two corresponding coordinate points; otherwise, no connection A is drawn i,j With A i,k The line connecting the two corresponding coordinate points;

[0086] Set the output interface as the collection of hologram data of borrower i, denoted as Ξ i , specifically:

[0087] Ξ i ={(j,x i,j ,y i,j ,Θ i,j )|Ai,j ∈V i}∪{Q i}; Each tuple (j,x i,j ,y i,j ,Θ i,j ) records the serial number, two-dimensional coordinates and topological value of the corresponding node; and also adds the final loan qualification score Q i .

[0088] The present invention has the following beneficial effects:

[0089] 1. By establishing an effective contribution function, which compares the integer portion of the square root with the residual value of the digital contribution identifier, the difficulty in quantifying the contribution information of the borrower's original online content is addressed. This method accurately distinguishes the quality and effectiveness of the contributed content, thus providing a clear numerical representation of the contribution information for each digital asset record. This step accurately quantifies the contribution level of the borrower's online content creation, laying a solid data foundation for the subsequent comprehensive assessment of loan eligibility, thereby achieving data standardization and objective assessment of contribution quality. Furthermore, by establishing an effective interaction function, which accurately determines whether the digital interaction identifier is divisible by the integer portion of its square root, the effective interaction behavior is determined, addressing the issues of ambiguous parameters and unclear judgment basis in traditional interaction data processing. This step ensures that the borrower's online interaction behavior within an observation period can be expressed in a deterministic and digital manner, ensuring consistency and comparability in subsequent assessments. Furthermore, by constructing digital contribution count, digital interaction count, and contribution growth functions, which accumulate the effective contribution and effective interaction terms within each observation period and calculate the difference between the two periods, the problem of fragmented data and unclear changes in loan eligibility assessment is addressed. By using fixed observation periods and normalization constants (e.g., constants 5, 10, and 30), the raw count data is mapped to a standard interval, stabilizing the data distribution while ensuring that the overall score is not biased by extreme data. This normalization step ensures comparability across borrowers and ensures that the information density, interactive activity, and future momentum in the digital asset vector objectively reflect the borrower's performance in the digital world. Furthermore, by multiplying the standardized values ​​in the digital asset vector into a preliminary loan qualification score, this overcomes the unstable and poorly interpretable scoring issues inherent in traditional assessment methods, which rely solely on historical data or fuzzy weighting. This step ensures that the entire evaluation process utilizes fixed mathematical operations and clear numerical mappings, ensuring transparency, stability, and traceability of the evaluation results. Overall, through these steps, the present invention overcomes the existing issues of fragmented data sources, unclear indicators, and inconsistent parameters, achieving an objective, standardized, and quantitative assessment of borrowers' digital assets, providing solid and reliable data support for loan decisions.

[0090] 2. By extracting original online text from each borrower's record and subjecting it to a series of rigorous preprocessing steps, this solution addresses the interference caused by formatting noise, redundant symbols, and unnecessary characters in the raw text data. Specifically, this solution first extracts the text content and stores it as a string. It then removes all spaces, line breaks, and non-alphanumeric characters from the string, resulting in a processed string containing only core characters. This ensures that the subsequent weight calculation is based solely on valid information. This preprocessing step addresses issues such as inconsistent formatting, high data noise, and redundant characters that may exist in the raw text, resulting in a cleaner and more standardized character data for the calculation, laying the foundation for accurate contribution calculation. In the weight calculation step, the solution multiplies each processed character by its specific position in the string and its Unicode code point to generate a weight for each character. This step introduces a character position factor to differentiate the contribution of each character at different positions, addressing the problem that simple character counts cannot reflect differences in content structure. The weights of all characters are summed to obtain a total weight value, which provides a more detailed representation of the record's overall contribution. This calculation method ensures that the actual value of each character in the original content can be effectively quantified, avoiding the data trivialization that can result from simple word counts. Finally, to ensure that all generated digital contribution identifiers fall within a reasonable and stable numerical range, the scheme sets a fixed offset constant. The introduction of this constant not only shifts the cumulative sum to ensure that the value is large enough to prevent data overlap, but also makes the absolute differences between adjacent records more obvious, facilitating subsequent comparison and sorting. By rationally selecting the fixed offset constant, the practical problem of identifier values ​​being too low or difficult to distinguish due to fluctuations in the original weights is resolved, thereby effectively improving the discreteness and stability of the data across borrowers. In summary, through text preprocessing, character weight calculation, weight accumulation, and the application of a fixed offset constant, this method addresses the problems of high noise in the original online text, uneven data distribution, and difficulty in distinguishing subtle contribution differences. It achieves the effect of accurately and standardizedly converting digital contribution identifiers into a sufficiently large, stable, and easily distinguishable positive integer, providing reliable and comparable basic data for subsequent loan qualification assessments.

[0091] 3. By obtaining interaction log data from the borrower's digital asset records and making fine distinctions between various interaction events in the logs, the digital interaction identifiers in this solution achieve accurate quantification of the borrower's online interactive activity. First, the solution uses specific steps to extract interaction log data from the borrower's records, forming a collection of browsing, like, comment, and sharing events. This categorized data collection solves the problems of unclear interaction data sources and mixed event types in traditional methods. Next, a fixed and clear weight is assigned to each interaction event in the collection. By assigning different weights to browsing, like, comment, and sharing events, this solution ensures the importance of different events in the score calculation, thereby solving the problem of the inability to quantitatively distinguish between various interaction events in traditional methods. Subsequently, by counting the number of occurrences of each interaction event in the collection and recording them as the number of occurrences of browsing, like, comment, and sharing events, this step ensures the accuracy and comprehensiveness of the data statistics. Next, an interaction score is calculated based on the product of each event's weight and the number of occurrences. This score truly reflects the borrower's total interactions and level of activity during an observation period, resolving the issue of discontinuous and unstable values ​​in traditional interaction data identification. To further ensure the comparability and distinguishability of interaction score data obtained between different records, the scheme also incorporates an integral offset constant. By introducing the integral offset constant, the different interaction scores are shifted overall to a stable and expectedly high numerical range, unifying the numerical baseline and effectively avoiding data confusion or unclear score differences caused by too low a base value. This step addresses the issue of uneven distribution of raw interaction scores and overly narrow data intervals. At the same time, by properly adjusting the integral offset constant, the sensitivity of the interaction indicator's contribution to the overall score can be effectively adjusted. Specifically, when the integral offset constant is set to a moderate value, it not only prevents the identification value from being too low and losing its discernibility, but also allows the subtle differences generated by different interaction events to be reflected in the overall score, ensuring the balance and stability of the scoring results. In summary, by extracting interaction logs, assigning weights to events, counting the number of events, calculating interaction scores, and introducing a fixed score offset constant, this solution solves problems such as uneven data distribution, insufficient sensitivity control, and inconsistent result intervals in traditional interaction data processing. Ultimately, it achieves the expression of borrowers' online interactive behaviors in the form of deterministic and standardized positive integers, providing an objective, comparable, and stable data foundation for subsequent overall loan qualification scoring, and significantly improving the transparency and accuracy of data processing.

[0092] 4. By preprocessing the raw digital asset data, this solution effectively addresses the issues of uneven data quality, missing fields, and noisy data that interfere with subsequent evaluations, thereby achieving data standardization and extracting effective information. Specifically, the preprocessing step first calls a custom function to calculate the judgment values ​​for the digital contribution indicator and the digital interaction indicator. This step assigns clear validity rules to these two key indicators in the raw data. The digital contribution and interaction information in each record are converted into binary values ​​(1 or 0), clearly indicating the validity of each data component in the record. Through this operation, the preprocessing system distinguishes ineffective contributions or interactions in the raw records, providing a clean, standardized data source for subsequent comprehensive scoring and graph construction. Next, the system constructs each record, after passing the above function, into a binary tuple, where the first component represents whether the digital contribution is valid, and the second component represents whether the digital interaction is valid. This binary record structure not only simplifies data storage but also ensures that each record retains only information that is useful for loan eligibility assessment, significantly reducing the impact of noisy data on the overall evaluation results. This process addresses issues such as redundant information, confusing formats, and invalid information in the raw data, improving data cleansing and consistency throughout the entire evaluation process. Furthermore, during preprocessing, if a record contains missing required fields or non-positive integer values, the preprocessing system will deem the record invalid and exclude it from subsequent processing. This step ensures that the dataset entering the subsequent calculation phase consists entirely of high-quality data that meets the established standards, effectively preventing assessment bias and inaccuracy caused by missing or incorrect data input. Ultimately, by screening and integrating all valid records, a precise preprocessed dataset is constructed, providing a reliable and standardized data foundation for subsequent digital asset map construction, indicator calculation, and final loan eligibility scoring. In summary, by invoking functions to determine the validity of digital contribution and interaction information, constructing standardized two-tuple records, and rigorously eliminating invalid or erroneous data, this preprocessing step addresses issues such as inconsistent data formats, excessive noise, and incomplete information in the raw digital asset data. Ultimately, the entire loan eligibility assessment system operates on a high-quality, standardized data foundation, significantly improving the accuracy and stability of assessment results.

[0093] 5. By constructing digital asset graph nodes and creating a self-created semantic association algorithm, this solution effectively solves the problem of unclear relationships between data in the original pre-processed records and difficulty in fully aggregating information, thereby providing structured and quantitative data support for subsequent loan qualification assessments. Specifically, this solution first generates graph nodes from the pre-processed data set. Each node corresponds to a pre-processed digital asset record, and sets two attributes for each node: a contribution attribute and an interaction attribute. Both attributes can only take values ​​of 0 or 1, reflecting the contribution judgment result and the interaction judgment result of the record, respectively, thereby clearly classifying digital asset information in a binary manner, solving the problem of ambiguous record information and inconsistent judgment criteria in traditional methods. Next, this solution introduces a self-created semantic association function, which is used to calculate the association strength between any two nodes. The calculation steps of the semantic association function include: first, calculating the Euclidean distance between two nodes in the attribute space. The first metric is obtained by taking the square root of the sum of the squared differences, which reflects the degree of attribute similarity between the nodes. Second, when both nodes have valid contributions (i.e., their contribution attributes are both 1), the function's return value is added to the smaller of the two node's contribution attributes. This strengthens the effect of shared contribution information, thereby quantitatively reflecting the semantic similarity between the nodes. This innovative function addresses the problem of traditional graph construction lacking a precise quantitative representation of semantic relationships between nodes, enabling a quantitative description of the inherent connections between digital assets. Furthermore, during node generation, this solution further compares any two nodes. When the result of the semantic association function is greater than zero, the system establishes a directed edge from the previous node to the next node, and sets the edge's weight to the function's output value, thus constructing a complete node-to-node association graph. This step not only addresses the lack of connections between records and the severe data silos in the original data, but also achieves the aggregation and structuring of digital asset information by clearly expressing the relationships between nodes and edges in the graph. Ultimately, the establishment of the graph node set and edge set provides intuitive and clear structured data support for subsequent loan qualification scoring, ensuring that the semantic relationship between digital assets is fully utilized, thereby improving the transparency, objectivity and replicability of the entire evaluation system.

[0094] 6. By constructing a holographic map and mapping node attributes, this solution systematically solves the problem of lack of a global unified structure representation for borrowers’ multi-period digital asset records, significantly improving the integrity of information aggregation and the consistency of evaluation results in loan qualification assessment. First, in step S11, an independent holographic map G is defined for each borrower. i, the graph collects all the digital asset records of the borrower that have been strictly screened during the current observation period and represents them as a set of nodes and edges in the graph structure. This structured aggregation process ensures that each borrower's information is independently modeled, avoiding the problem of cross-interference of multiple borrower data, and achieving user isolation and graph independence in the evaluation modeling. Then in step S12, the loan qualification score S i The mapping is injected into each graph node as one of the important attributes of the node. The score is generated by the aforementioned standardized dimensions based on digital contribution, interaction and growth trend. Through the injection process, a deep integration of the original scoring system and the graph structure is achieved. This step effectively solves the problem that the traditional scoring system is difficult to link with the behavior graph, making the score no longer an isolated indicator, but a part of the graph node attribute. Finally, in step S13, by traversing the preprocessed data, only the nodes that meet the condition α are included. i,j = 1 is constructed as a node to ensure that the graph is constructed based only on high-quality and high-validity digital asset information; further, for all legal node pairs (A i,j ,A i,k )Calculate the semantic association strength Λ(A i,j ,A i,k ), and dynamically generates edges and edge weights based on whether the result is greater than zero. This construction mechanism not only avoids the generation of redundant edges, improving graph sparsity, but also ensures that nodes are connected only based on real semantic commonalities.

[0095] 7. By introducing the steps of calculating graph topology values ​​and generating loan qualification scores, this solution effectively addresses the difficulty in quantifying the intensity of interactions between a borrower's digital assets over multiple periods and the difficulty in reflecting the overall network of behavior in the assessment results. This enhances the structural rationality and interpretability of the loan qualification assessment results. First, by setting a topology value for each node—directly summing the edge values ​​of all edges originating from that node—this step quantifies the propagation influence of a single record in the digital asset graph. This topology value not only reflects the node's own level of interaction and contribution, but also further reflects the strength of its semantic connections with other records, addressing the problem that traditional metrics cannot capture the internal interaction structure. Next, by averaging the topology values ​​of all nodes in the borrower's graph, an overall graph index is formed, which uniformly measures the comprehensive topological characteristics of all of the borrower's valid digital assets during the current observation period. This average value, a quantitative indicator of the graph's overall connectivity and interaction cohesion, reflects the organizational quality of the borrower's digital asset behavior, addressing the problem of over-reliance on local behaviors in the assessment process and the lack of global visibility. Finally, the graph index Ω is used. i Directly mapped to loan qualification score Q iThis mapping maintains the interpretability and numerical consistency of the original graph calculation logic, without the need to introduce external complex models, effectively avoiding the model black box problem. In this process, not only is the subjectivity in the weighted fusion of multi-dimensional data avoided, but the graph structure information is also directly converted into usable decision-making scoring results. In summary, this step solves the problem of the difficulty of incorporating the behavioral structure of digital assets into the evaluation system through the path of "node topology accumulation → full graph average aggregation → numerical mapping scoring", and ultimately realizes a quantitative loan qualification scoring mechanism driven by structured network information, significantly improving the accuracy, objectivity and scalability of the scoring process.

[0096] 8. By introducing the design of two-dimensional coordinate mapping of graph nodes and data interface output, this solution effectively solves the problem of complex information structure of borrower digital asset graph, difficulty in intuitive presentation and difficulty in direct access to external systems, thereby improving the visual expression and data interactivity of graph analysis, and enhancing the application efficiency and transparency of the final loan qualification scoring results. First, by setting the horizontal coordinate of the node to its sequential number x in the original record i,j =j, the vertical coordinate is set to the topological value y of the node i,j =Θ i,j , a structured two-dimensional coordinate mapping system was constructed. This step intuitively displays the network structure effect of the borrower records evolving over time through the linkage arrangement between the number and the contribution influence, effectively solving the problem that the semantics of multiple nodes are difficult to visualize and focus. Secondly, through the method of node drawing and edge value threshold filtering and drawing lines, only the edge value L i,j,k >0, thus forming a concise graph view that preserves strong correlation structure while avoiding information overload. This strategy effectively selects node pairs with significant interactive correlation in digital behavior, avoids the problem of over-crowding the graph due to drawing all edges, and makes the final graph more visually interpretable and structurally expressive. Finally, by constructing a structured data interface output format Ξ i ={(j,x i,j ,y i,j ,Θ i,j )|A i,j ∈V i}∪{Q iThis output not only returns visualization information (node ​​number, coordinates, topology value) along with the loan score, but also provides a standardized, clearly structured data entry point for external credit risk control systems or management platforms, resolving the difficulty integrating graph results into business systems. In summary, this step, through the three-step linkage of "two-dimensional mapping → visualization construction → interface output," effectively addresses the difficulties of interpreting graph structures and reusing behavioral analysis results. It achieves a seamless integration of data analysis, risk assessment, and visual expression, significantly enhancing the practicality, transparency, and integration capabilities of the loan qualification assessment system. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0098] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0099] Example, see Figure 1 , a loan qualification assessment method using digital assets, comprising:

[0100] Let i be the unique integer identifier of the borrower, and satisfy i∈{1,2,...,I}; where I is the total number of borrowers;

[0101] Assume an observation period of 30 days;

[0102] The total number of digital asset records generated by borrower i in an observation period is recorded as n i , the serial number of the record is j, and j∈{1,2,...,n i};

[0103] The original digital asset record of borrower i with serial number j is recorded as d i,j , where d i,j Contains: Digital Contribution ID i,j , is a positive integer, for record d i,j The corresponding contribution value; digital interaction identifier IM i,j , is a positive integer, for record d i,j The corresponding interactive prompt value;

[0104] Set the effective contribution function and calculate the digital contribution ID in the j-th record of borrower i i,j The judgment value is:

[0105]

[0106] Among them, φ(ID i,j ) is the ID i,j The judgment value of Indicates getting ID i,j The integer part of the square root; 1 indicates valid, 0 indicates invalid; this function determines whether the remaining value of the digital contribution identifier in the record is greater than or equal to the integer part of the square root, thereby determining whether the contribution of the record is valid;

[0107] Set the effective interaction function and calculate the digital interaction identifier IM in the j-th record of borrower i i,j The judgment value is:

[0108]

[0109] Among them, ψ(IM i,j ) is IM i,j The judgment value; mod is the remainder operation; Indicates taking IM i,j The integer part of the square root; 1 means valid, 0 means invalid; this function judges IM i,j Can it be Divisibility is used to determine whether the interaction is effective;

[0110] Set up a digital contribution counting function to calculate the total count of valid contribution items in all records of borrower i in an observation period, specifically:

[0111] Among them, N i is the total count of valid contribution items in all records of borrower i in an observation period;

[0112] Set the digital interaction counting function to calculate the total count of valid interaction items in all records of borrower i in an observation period, specifically:

[0113] Among them, M i is the total count of valid interaction items in all records of borrower i in an observation period;

[0114] Set the contribution growth function to calculate the contribution growth of borrower i, specifically:

[0115] in, is the total count of valid contributions obtained by borrower i in the current observation period (t); is the total count of valid contribution items obtained by borrower i in the previous observation period (t-1); this difference directly represents the increase in the borrower's contribution in the time dimension;

[0116] Set the digital asset vector of borrower i to: V i =(I i ,R i ,P i );

[0117] Among them, I i is the information density, The constant 5 is used for normalization to map the contribution count to the expected range; if the constant 5 is larger than the actual set value, then I i The calculated result will be smaller, which may make the initial loan qualification score S i Reduce; if constant 5 is smaller than the actual set value, then I i The calculated result will be larger, which may make the initial loan qualification score S i High; R i For the level of interactive activity, The constant 10 is used as a normalization factor to control the interaction count within a reasonable range; if the constant 10 is set too large, the interaction data R after normalization will be i will decrease, which may reduce the overall score; on the contrary, if the constant 10 is small, then R i High, overall score S i Rising; P i For future momentum, The constant 30 represents the number of days in the fixed observation period, which directly determines the growth value per unit time;

[0118] Set the initial loan qualification score of borrower i to S i , S i =I i ×R i ×P i ; The above three standardized components are combined by multiplication to obtain the comprehensive score S i , the score is subsequently used to initialize the graph node attributes.

[0119] By establishing an effective contribution function, which compares the integer portion of the square root with the residual value of the digital contribution identifier, the difficulty in quantifying the contribution information of borrowers' original online content is addressed. This method accurately distinguishes the quality and effectiveness of the contributed content, thus providing a clear numerical representation of the contribution information for each digital asset record. This step accurately quantifies the contribution level of the borrower's online content creation, laying a solid data foundation for the subsequent comprehensive assessment of loan eligibility, thereby achieving data standardization and objective assessment of contribution quality. Furthermore, by establishing an effective interaction function, which accurately determines whether the digital interaction identifier is divisible by the integer portion of its square root, the effective interaction behavior is determined, addressing the issues of ambiguous parameters and unclear judgment basis in traditional interaction data processing. This step ensures that the borrower's online interaction behavior within an observation period can be expressed in a deterministic and digital manner, ensuring consistency and comparability in subsequent assessments. Furthermore, by constructing digital contribution count, digital interaction count, and contribution growth functions, which accumulate the effective contribution and effective interaction terms within each observation period and calculate the difference between the two periods, the problem of fragmented data and unclear changes in loan eligibility assessment is addressed. By using fixed observation periods and normalization constants (e.g., constants 5, 10, and 30), the raw count data is mapped to a standard interval, stabilizing the data distribution while ensuring that the overall score is not biased by extreme data. This normalization step ensures comparability across borrowers and ensures that the information density, interactive activity, and future momentum in the digital asset vector objectively reflect the borrower's performance in the digital world. Furthermore, by multiplying the standardized values ​​in the digital asset vector into a preliminary loan qualification score, this overcomes the unstable and poorly interpretable scoring issues inherent in traditional assessment methods, which rely solely on historical data or fuzzy weighting. This step ensures that the entire evaluation process utilizes fixed mathematical operations and clear numerical mappings, ensuring transparency, stability, and traceability of the evaluation results. Overall, through these steps, the present invention overcomes the existing issues of fragmented data sources, unclear indicators, and inconsistent parameters, achieving an objective, standardized, and quantitative assessment of borrowers' digital assets, providing solid and reliable data support for loan decisions.

[0120] The digital contribution identifier specifically includes:

[0121] From the jth record d of borrower i i,j Extract the original online text content and record it as string P i,j ;

[0122] Remove string P i,j All spaces, newlines and non-alphanumeric characters in the string are processed and recorded as P'i,j ;

[0123] Get string P' i,j The total number of characters is L i,j ;

[0124] For string P' i,j Each character in is weighted according to its position in the string, specifically:

[0125] ω(c i,j,f )=f×U(c i,j,f );

[0126] Among them, ω(c i,j,f ) is the string P' i,j The weight of the fth character in ; c i,j,f For string P' i,j The fth character in the string; f is the position of the character in the string, f∈{1,2,...L i,j}; U(c i,j,f ) is the character c i,j,f Unicode code value;

[0127] Calculate the cumulative sum of all character weights, specifically: Among them, S i,j Extract the total weight of the online original text content characters from the j-th record of borrower i;

[0128] Set the offset preset constant to C1; baseline adjustment and value range offset: The main function of C1 is to convert the accumulated sum S i,j Shift to a predetermined value range to ensure that all generated digital contribution identifiers are large enough positive integers; if the value of C1 is small, the overall calculation result ID i,j It may fall in a lower numerical range. Low values ​​may make the relative differences between different contents unclear in absolute value, thereby reducing the ability to distinguish. If C1 takes a larger value, all results will be shifted upward to ensure that the results are in a higher range, while avoiding the problem of S i,j Fluctuations in the C1 value may lead to the risk that the marker may approach zero or overlap; the impact on sensitivity to differences: C1 is essentially a constant bias, and its value directly affects the relative differences between markers in subsequent comparisons; when the value of C1 is much larger than S between different records, the markers may overlap. i,j If the difference between different records is too large, the relative difference ratio between different records will be diluted, which may weaken the impact of the original difference in the contribution content in some comparison or sorting algorithms; if C1 is selected to be moderate, so that S i,jIf the changes in C1 can still be adequately reflected, it can ensure that the identification value fully prevents the risk of being too low while still maintaining sensitivity to content differences; numerical range and data standardization: by selecting an appropriate C1 value, the digital contribution identification can be normalized to the desired numerical range, which is convenient for subsequent data processing, storage and comparison; if the numerical range is too wide, it may increase the complexity of processing large number operations; if the range is too narrow, it may cause data distribution to overlap, which is not conducive to subsequent differentiation; therefore, C1 should comprehensively consider the typical fluctuation range of the original text length and the character encoding weight, so as to ensure that the final identification has good discreteness and stability;

[0129] ID i,j =S i,j +C1.

[0130] By extracting original online text from each borrower's record and subjecting it to a series of rigorous preprocessing steps, this solution addresses the issues that formatting noise, redundant symbols, and unnecessary characters present in the raw text data can interfere with numerical calculations. Specifically, this solution first extracts the text content and stores it as a string. It then removes all spaces, line breaks, and non-alphanumeric characters from the string, resulting in a processed string containing only core characters. This ensures that subsequent weight calculations are based solely on valid information. This preprocessing step addresses issues such as inconsistent formatting, high data noise, and redundant characters that may exist in the raw text, resulting in a cleaner and more standardized character data for accurate contribution calculation. In the weight calculation step, the solution multiplies each processed character by its position in the string and its Unicode code point to generate a weight for each character. This step incorporates a character position factor to differentiate the contribution of each character at different positions, addressing the problem that simple character counts fail to reflect differences in content structure. The weights of all characters are summed to produce a total weight that more accurately reflects the overall contribution of the record. This calculation method ensures that the actual value of each character in the original content can be effectively quantified, avoiding the data trivialization that can result from simple word counts. Finally, to ensure that all generated digital contribution identifiers fall within a reasonable and stable numerical range, the scheme sets a fixed offset constant. The introduction of this constant not only shifts the cumulative sum to ensure that the value is large enough to prevent data overlap, but also makes the absolute differences between adjacent records more obvious, facilitating subsequent comparison and sorting. By rationally selecting the fixed offset constant, the practical problem of identifier values ​​being too low or difficult to distinguish due to fluctuations in the original weights is resolved, thereby effectively improving the discreteness and stability of the data across borrowers. In summary, through text preprocessing, character weight calculation, weight accumulation, and the application of a fixed offset constant, this method addresses the problems of high noise in the original online text, uneven data distribution, and difficulty in distinguishing subtle contribution differences. It achieves the effect of accurately and standardizedly converting digital contribution identifiers into a sufficiently large, stable, and easily distinguishable positive integer, providing reliable and comparable basic data for subsequent loan qualification assessments.

[0131] The digital interactive identifier specifically includes:

[0132] From the jth record d of borrower i i,j Get the interactive log data from the set H i,j ;

[0133] The event types in the record include browsing events v, liking events l, commenting events c, and sharing events s;

[0134] Give set H i,j Each interactive event in is assigned a weight, where the browsing event v is assigned a weight v =1, the like event l is given weight a l =3, comment event c is assigned weight a c =5, the sharing event s is given weight a s =7;

[0135] Count the events in the set H i,j The number of occurrences of the browsing event is recorded as n v , the number of like events is recorded as n l , record the number of comment events as n c , the number of sharing events is recorded as n s ;

[0136] Calculating interaction points for:

[0137] Set the integral offset constant C2; unify the numerical benchmark and interval shift: the introduction of C2 is mainly used to integrate different interactive events The whole is shifted to a stable and expected higher value range; when the C2 value is small, the final interaction mark of some records with low interaction score sum may still be in a lower range, which may be confused with the situation with smaller data fluctuations in subsequent comparisons; when the C2 value is large, it can prevent all mark values ​​from being too low and enhance the recognition between different records, but an excessively large C2 value may also make the contribution of interaction score relatively unobvious; Difference amplification and sensitivity adjustment: The value of C2 directly determines the baseline level of interaction mark in the overall evaluation; if C2 greatly exceeds Internal differences, the differences between different records caused by the counts of interactive events will be relatively weakened in the fixed offset, reducing the sensitivity of the interactive characteristics; on the contrary, if C2 is selected moderately, it can ensure that small changes can still better reflect the differences in interactive activity under a fixed benchmark, making it easier to distinguish different records; numerical standardization and subsequent connection: by adjusting the value of C2, the numerical distribution of digital interaction identifiers and digital contribution identifiers can be kept within a mutually coordinated range, facilitating the subsequent comprehensive calculation of the overall loan qualification score; if the numerical ranges of the two differ greatly, it may lead to the contribution of a certain part of the indicators to the final result being suppressed in the subsequent combined scoring; therefore, the selection of C2 should consider the coordination between the expected cumulative value of interactive events and other indicators, so as to ensure the balance and comparability of the overall loan qualification score;

[0138] calculate

[0139] By extracting interaction log data from borrowers' digital asset records and meticulously distinguishing various interaction events within the logs, this digital interaction identifier achieves precise quantification of borrowers' online interactive activity. First, the solution employs specific steps to extract interaction log data from borrowers' records, generating a collection of views, likes, comments, and shares. This categorized data collection addresses the issues of unclear interaction data sources and mixed event types found in traditional methods. Next, a fixed and defined weight is assigned to each interaction event in the collection. By assigning different weights to views, likes, comments, and shares, this solution ensures the importance of different events in point calculation, thus addressing the inability to quantitatively distinguish between interaction events in traditional methods. Subsequently, the number of occurrences of each interaction event in the collection is counted and recorded as the number of views, likes, comments, and shares, respectively. This step ensures the accuracy and comprehensiveness of the data. Next, an interaction score is calculated by multiplying each event's weight by the number of occurrences. This score truly reflects the borrower's total interaction volume and level of activity over an observation period, addressing the discontinuous and unstable values ​​often associated with traditional interaction data identifiers. In order to further ensure the comparability and distinguishability of the interaction score data obtained between different records, the scheme also adopts an integral offset constant. By introducing the integral offset constant, the different interaction scores are shifted as a whole to a stable and expectedly high numerical range, the numerical benchmark is unified, and the situation where data confusion or unclear integral differences are caused by the base value being too low is effectively avoided. This step solves the problem that the original interaction scores may be unevenly distributed and the data interval is too narrow. At the same time, by reasonably adjusting the integral offset constant, the sensitivity of the interaction indicator to the overall score contribution can be effectively adjusted. Specifically, when the integral offset constant is moderate, it can not only prevent the identification value from being too low and losing recognition, but also enable the subtle differences caused by different interaction events to be reflected in the overall score, ensuring the balance and stability of the scoring results. In summary, by extracting interaction logs, assigning weights to events, counting the number of events, calculating interaction scores, and introducing a fixed score offset constant, this solution solves problems such as uneven data distribution, insufficient sensitivity control, and inconsistent result intervals in traditional interaction data processing. Ultimately, it achieves the expression of borrowers' online interactive behaviors in the form of deterministic and standardized positive integers, providing an objective, comparable, and stable data foundation for subsequent overall loan qualification scoring, and significantly improving the transparency and accuracy of data processing.

[0140] It also includes the pre-processing of raw digital asset data, specifically:

[0141] Record d i,j Do the following:

[0142] Call function φ(ID i,j) and ψ(IM i,j ) respectively determine the value of the digital contribution identifier and the digital interaction identifier;

[0143] Construct preprocessing record as r i,j =(φ(ID i,j ),ψ(IM i,j ));

[0144] Among them, r i,j It is a binary group, the first component represents whether the digital contribution is effective; the second component represents whether the digital interaction is effective;

[0145] For record d i,j If any of the following conditions exist, the application will be deemed invalid and will not be included in the subsequent processing:

[0146] Record d i,j Missing ID i,j or IM i,j ;ID i,j ≤0 or IM i,j ≤0;

[0147] Finally, the preprocessed data set is set to

[0148] By preprocessing the raw digital asset data, this solution effectively addresses the issues of uneven data quality, missing fields, and noisy data that interfere with subsequent assessments, thereby achieving data standardization and extracting effective information. Specifically, the preprocessing step first calls a custom function to calculate the judgment values ​​for the digital contribution indicator and the digital interaction indicator. This step assigns clear validity rules to these two key indicators in the raw data. The digital contribution and interaction information in each record are converted into binary values ​​(1 or 0), clearly indicating the validity of each data component in the record. This operation distinguishes ineffective contributions or interactions in the raw records, providing a clean, standardized data source for subsequent comprehensive scoring and graph construction. Next, the system constructs each record, after passing the aforementioned function, into a binary tuple, where the first component indicates whether the digital contribution is valid, and the second component indicates whether the digital interaction is valid. This binary record structure not only simplifies data storage but also ensures that each record retains only information that is relevant to the loan qualification assessment, significantly reducing the impact of noisy data on the overall assessment results. This process addresses issues such as redundant information, confusing formats, and invalid information in the raw data, improving data cleansing and consistency throughout the entire evaluation process. Furthermore, during preprocessing, if a record contains missing required fields or non-positive integer values, the preprocessing system will deem the record invalid and exclude it from subsequent processing. This step ensures that the dataset entering the subsequent calculation phase consists entirely of high-quality data that meets the established standards, effectively preventing assessment bias and inaccuracy caused by missing or incorrect data input. Ultimately, by screening and integrating all valid records, a precise preprocessed dataset is constructed, providing a reliable and standardized data foundation for subsequent digital asset map construction, indicator calculation, and final loan eligibility scoring. In summary, by invoking functions to determine the validity of digital contribution and interaction information, constructing standardized two-tuple records, and rigorously eliminating invalid or erroneous data, this preprocessing step addresses issues such as inconsistent data formats, excessive noise, and incomplete information in the raw digital asset data. Ultimately, the entire loan eligibility assessment system operates on a high-quality, standardized data foundation, significantly improving the accuracy and stability of assessment results.

[0149] It also includes constructing digital asset graph nodes and creating a self-created semantic association algorithm, specifically:

[0150] From the preprocessed data set Generate graph nodes:

[0151] For each record r i,j =(φ(ID i,j ),ψ(IMi,j )), denoted as graph node A i,j ;

[0152] Each node A i,j is set with the following properties:

[0153] Contribution property α i,j , α i,j = φ(ID i,j ); This property only takes 0 or 1, reflecting the record contribution judgment result;

[0154] Interaction property β i,j , β i,j = ψ(IM i,j ); This property only takes 0 or 1, reflecting the interaction judgment result;

[0155] Set a semantic association function to calculate the association strength between any two nodes, and calculate the association strength between any two nodes A i,j and A i,k , where j < k, specifically:

[0156]

[0157] where, Λ(A i,j , A i,k ) is the association strength between two nodes A i,j and A i,k ; α i,j , α i,k are respectively the contribution properties of nodes A i,j and A<000​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​, set the value on the side to L i,j,k = Λ(A i,j , A i,k );

[0161] Set the node set V i as:

[0162] Set the edge set to: E i = {e i,j,k | j < k and Λ(A i,j , A i,k ) > 0}.

[0163] Through the steps of constructing the digital asset graph nodes and the self-created semantic association algorithm, this solution effectively solves the problems of unclear relationships between data and difficult comprehensive aggregation of information in the original preprocessing records, thus providing structured and quantitative data support for subsequent loan qualification assessment. Specifically, this solution first generates graph nodes from the preprocessing data set. Each node corresponds to a preprocessed digital asset record, and two attributes are set for each node: contribution attribute and interaction attribute. Both of these attributes only take values of 0 or 1, respectively reflecting the contribution judgment result and interaction judgment result of the record, thus clearly classifying digital asset information in a binary way and solving the problems of fuzzy record information and inconsistent judgment criteria in traditional methods. Next, this solution introduces a self-created semantic association function, which is used to calculate the association strength between any two nodes. The calculation steps of the semantic association function include: first, calculate the Euclidean distance between the two nodes in the attribute space, and obtain the first metric by taking the square root of the sum of squared differences, which can reflect the similarity degree of attributes between nodes; second, when both nodes are effective contributions, that is, the contribution attributes are both 1, the function return value is added with the smaller value of the contribution attributes of the two nodes, which plays a role in strengthening the effect of shared contribution information, thus quantitatively reflecting the semantic similarity between nodes. Through this innovative function, the problem of lack of precise quantitative expression of the semantic relationship between nodes in the traditional graph construction process is solved, and the quantitative description of the internal connection between digital assets is realized. In addition, when generating nodes, this solution further compares any two nodes. When the result of the semantic association function is greater than zero, the system establishes a directed edge from the previous node to the next node, and sets the weight of this edge to the output value of the function, thus constructing a complete association graph between nodes. This step not only solves the problem of lack of connection between each record in the original data and serious data islands, but also realizes the aggregation and structuring of digital asset information through the clear relationship expression between nodes and edges in the graph. Finally, the establishment of this graph node set and edge set provides intuitive and clear structured data support for subsequent loan qualification scoring, ensures that the semantic relationship between digital assets is fully utilized, and thus improves the transparency, objectivity and replicability of the entire assessment system.

[0164] It also includes constructing a holographic map and node attribute mapping, specifically:

[0165] S11. Holographic map definition:

[0166] For each borrower i, construct a holographic map G i , G i =(V i , E i );

[0167] S12. Node attribute injection:

[0168] Set γ i,j as the loan qualification score injected into node A i,j , and γ i,j =S i ;

[0169] S13. Map generation process:

[0170] Generate nodes A i,j in the order of the preprocessing record r i,j , and only generate records that satisfy α i,j =1;

[0171] For all node pairs that satisfy j < k, calculate Λ(A i,j , A i,k ); if the result is greater than zero, establish the corresponding edge e i,j,k and assign the edge value L i,j,k ;

[0172] Thus, the completed map G i =(V i , E i ) is generated.

[0173] Through the steps of constructing a holographic map and node attribute mapping, this solution systematically solves the problem of the lack of a globally unified structural representation for multi-cycle digital asset records of borrowers, significantly improving the integrity of information aggregation and the coherence of evaluation results in loan qualification assessment. First, in step S11, an independent holographic map G i is defined for each borrower. This map aggregates all strictly screened digital asset records of the borrower in the current observation period and represents them as a set of nodes and a set of edges in a graph structure. This structured aggregation process ensures that each borrower's information is independently modeled, avoiding the problem of cross-interference of multi-borrower data and achieving user isolation and map independence in the evaluation modeling. Subsequently, in step S12, the loan qualification score S iThe mapping is injected into each graph node as one of the important attributes of the node. The score is generated by the aforementioned standardized dimensions based on digital contribution, interaction and growth trend. Through the injection process, a deep integration of the original scoring system and the graph structure is achieved. This step effectively solves the problem that the traditional scoring system is difficult to link with the behavior graph, making the score no longer an isolated indicator, but a part of the graph node attribute. Finally, in step S13, by traversing the preprocessed data, only the nodes that meet the condition α are included. i,j = 1 is constructed as a node to ensure that the graph is constructed based only on high-quality and high-validity digital asset information; further, for all legal node pairs (A i,j ,A i,k )Calculate the semantic association strength Λ(A i,j ,A i,k ), and dynamically generates edges and edge weights based on whether the result is greater than zero. This construction mechanism not only avoids the generation of redundant edges, improving graph sparsity, but also ensures that nodes are connected only based on real semantic commonalities.

[0174] It also includes the calculation of graph topology values ​​and the generation of loan qualification scores, specifically:

[0175] For the spectrum G i Any node A i,j ∈V i , set the topology value Θ i,j for: On all slave nodes A i,j Starting side i,j,k In the example above, we can directly find the boundary value L i,j,k The cumulative sum of is the topological influence value of the node;

[0176] Average the topological values ​​of all nodes in the holographic graph of borrower i and set the overall graph index Ω i for:

[0177] Among them, |V i | is the node set V i The total number of nodes in Ω i Directly and quantitatively describe the closeness of the borrower's internal self-organized network of digital assets;

[0178] The overall spectrum index Ω i Directly mapped to the final loan qualification score Q i , specifically: Q i =Ω i ; Borrower i’s loan qualification score Q i The larger the value, the better the borrower's internal digital asset network performs in terms of contribution and interaction, and the more secure the loan qualification is.

[0179] By introducing the steps of calculating graph topology values ​​and generating loan qualification scores, this solution effectively addresses the difficulties in quantifying the intensity of interactions between a borrower's digital assets over multiple periods and the difficulty in capturing the overall network of behavior in the assessment results. This approach thus enhances the structural rationality and interpretability of loan qualification assessment results. First, by setting a topology value for each node—directly summing the edge values ​​of all edges originating from that node—this step quantifies the influence of individual records in the digital asset graph. This topology value not only reflects the node's own level of interaction and contribution, but also further reflects the strength of its semantic connections with other records, addressing the problem that traditional metrics cannot capture the internal interaction structure. Next, by averaging the topology values ​​of all nodes in the borrower's graph, an overall graph index is formed, which uniformly measures the comprehensive topological characteristics of all of the borrower's valid digital assets during the current observation period. This average value, a quantitative indicator of the graph's overall connectivity and interaction cohesion, reflects the organizational quality of the borrower's digital asset behavior, addressing the problem of over-reliance on local behaviors in the assessment process and the lack of global visibility. Finally, the graph index Ω is used. i Directly mapped to loan qualification score Ω i This mapping maintains the interpretability and numerical consistency of the original graph calculation logic, without the need to introduce external complex models, effectively avoiding the model black box problem. In this process, not only is the subjectivity in the weighted fusion of multi-dimensional data avoided, but the graph structure information is also directly converted into usable decision-making scoring results. In summary, this step solves the problem of the difficulty of incorporating the behavioral structure of digital assets into the evaluation system through the path of "node topology accumulation → full graph average aggregation → numerical mapping scoring", and ultimately realizes a quantitative loan qualification scoring mechanism driven by structured network information, significantly improving the accuracy, objectivity and scalability of the scoring process.

[0180] It also includes the two-dimensional coordinate mapping of graph nodes and data interface output, specifically:

[0181] The atlas G i Each node A i,j ∈V i Mapped onto a two-dimensional plane, where:

[0182] Horizontal coordinate x i,j Set to the node's sequence number, i.e. x i,j =j;

[0183] vertical coordinate y i,j Set to the topological value of the node, y i,j =Θ i,j ;

[0184] This mapping method ensures that the nodes present an intuitive distribution on the plane that is related to their generation order and the degree of mutual influence;

[0185] According to the above coordinates, each node is drawn as a point (x i,j ,y i,j );

[0186] At the same time, for any edge e in the graph i,j,k , if L i,j,k >0, then draw a straight line on the plane connecting A i,j With A i,k The two corresponding coordinate points; otherwise, no connection A is drawn i,j With A i,k The line connecting the two corresponding coordinate points;

[0187] Set the output interface as the collection of hologram data of borrower i, denoted as Ξ i , specifically:

[0188] Ξ i ={(j,x i,j ,y i,j ,Θ i,j )|A i,j ∈V i}∪{Q i}; Each tuple (j,x i,j ,y i,j ,Θ i,j ) records the serial number, two-dimensional coordinates and topological value of the corresponding node; and also adds the final loan qualification score Q i , for use in credit decision-making.

[0189] By introducing the design of two-dimensional coordinate mapping of graph nodes and data interface output, this solution effectively solves the problem of complex information structure of borrower digital asset graph, difficulty in intuitive presentation and difficulty in direct access to external systems, thereby improving the visual expression and data interactivity of graph analysis, and enhancing the application efficiency and transparency of the final loan qualification scoring results. First, by setting the horizontal coordinate of the node to its sequential number x in the original record, i,j =j, the vertical coordinate is set to the topological value y of the node i,j =Θ i,j , a structured two-dimensional coordinate mapping system was constructed. This step intuitively displays the network structure effect of the borrower records evolving over time through the linkage arrangement between the number and the contribution influence, effectively solving the problem that the semantics of multiple nodes are difficult to visualize and focus. Secondly, through the method of node drawing and edge value threshold filtering and drawing lines, only the edge value L i,j,k>0, thus forming a concise graph view that preserves strong correlation structure while avoiding information overload. This strategy effectively selects node pairs with significant interactive correlation in digital behavior, avoids the problem of over-crowding the graph due to drawing all edges, and makes the final graph more visually interpretable and structurally expressive. Finally, by constructing a structured data interface output format Ξ i ={(j,x i,j ,y i,j ,Θ i,j )|A i,j ∈V i}∪{Q i This output not only returns visualization information (node ​​number, coordinates, topology value) along with the loan score, but also provides a standardized, clearly structured data entry point for external credit risk control systems or management platforms, resolving the difficulty integrating graph results into business systems. In summary, this step, through the three-step linkage of "two-dimensional mapping → visualization construction → interface output," effectively addresses the difficulties of interpreting graph structures and reusing behavioral analysis results. It achieves a seamless integration of data analysis, risk assessment, and visual expression, significantly enhancing the practicality, transparency, and integration capabilities of the loan qualification assessment system.

[0190] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0191] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A loan qualification assessment method using digital assets, characterized in that: include: Let i be the unique integer identifier of the borrower, and satisfy i∈{1,2,...,I}; where I is the total number of borrowers; Assume an observation period of 30 days; The total number of digital asset records generated by borrower i in an observation period is recorded as n i , the serial number of the record is j, and j∈{1,2,...,n i }; The original digital asset record of borrower i with serial number j is recorded as d i,j , where d i,j Contains: Digital Contribution ID i,j , is a positive integer, for record d i,j The corresponding contribution value; digital interaction identifier IM i,j , is a positive integer, for record d i,j The corresponding interactive prompt value; Set the effective contribution function and calculate the digital contribution ID in the j-th record of borrower i i,j The judgment value is: Among them, φ(ID i,j ) is the ID i,j The judgment value of Indicates getting ID i,j The integer part of the square root; 1 means valid, 0 means invalid; Set the effective interaction function and calculate the digital interaction identifier IM in the j-th record of borrower i i,j The judgment value is: Among them, ψ(IM i,j ) is IM i,j The judgment value; mod is the remainder operation; Indicates taking IM i,j The integer part of the square root; 1 means valid, 0 means invalid; Set up a digital contribution counting function to calculate the total count of valid contribution items in all records of borrower i in an observation period, specifically: Among them, N i is the total count of valid contribution items in all records of borrower i in an observation period; Set the digital interaction counting function to calculate the total count of valid interaction items in all records of borrower i in an observation period, specifically: Among them, M i is the total count of valid interaction items in all records of borrower i in an observation period; Set the contribution growth function to calculate the contribution growth of borrower i, specifically: in, is the total count of valid contributions obtained by borrower i in the current observation period (t); is the total number of valid contributions obtained by borrower i in the previous observation period (t-1); Set the digital asset vector of borrower i to: V i =(I i ,R i ,P i ); Among them, I i is the information density, The constant 5 is used for normalization; R i For the level of interactive activity, The constant 10 is used as a normalization factor; P i For future momentum, The constant 30 represents the number of days of the fixed observation period; Set the initial loan qualification score of borrower i to S i , S i =I i ×R i ×P i .

2. A loan qualification assessment method using digital assets according to claim 1, characterized in that: The digital contribution identifier specifically includes: From the jth record d of borrower i i,j Extract the original online text content and record it as string P i,j ; Remove string P i,j All spaces, newlines and non-alphanumeric characters in the string are processed and recorded as P' i,j ; Get string P' i,j The total number of characters is L i,j ; For string P' i,j Each character in is weighted according to its position in the string, specifically: ω(c i,j,f )=f×U(c i,j,f ); Among them, ω(C i,j,f ) is the string P' i,j The weight of the fth character in ; c i,j,f For string P' i,j The fth character in the string; f is the position of the character in the string, f∈{1,2,...L i,j }; U(c i,j,f ) is the character c i,j,f Unicode code value; Calculate the cumulative sum of all character weights, specifically: Among them, S i,j Extract the total weight of the online original text content characters from the j-th record of borrower i; Set the offset preset constant to C1; ID i,j =S i,j +C1.

3. The loan qualification assessment method using digital assets according to claim 1, characterized in that: The digital interactive identifier specifically includes: From the jth record d of borrower i i,j Get the interactive log data from the set H i,j ; The event types in the record include browsing events v, liking events l, commenting events c, and sharing events s; Give set H i,j Each interactive event in is assigned a weight, where the browsing event v is assigned a weight v =1, the like event l is given weight a l =3, comment event c is assigned weight a c =5, the sharing event s is given weight a s =7; Count the events in the set H i,j The number of occurrences of the browsing event is recorded as n v , the number of like events is recorded as n l , record the number of comment events as n c , the number of sharing events is recorded as n s ; Calculating interaction points for: Set the integral offset constant C2; calculate 4. The loan qualification assessment method using digital assets according to claim 1, characterized in that: It also includes the pre-processing of raw digital asset data, specifically: Record d i,j Do the following: Call function φ(ID i,j ) and ψ(IM i,j ) respectively determine the value of the digital contribution identifier and the digital interaction identifier; Construct preprocessing record as r i,j =(φ(ID i,j ),ψ(IM i,j )); Among them, r i,j It is a binary group, the first component represents whether the digital contribution is effective; the second component represents whether the digital interaction is effective; For record d i,j If any of the following conditions exist, the application will be deemed invalid and will not be included in the subsequent processing: Record d i,j Missing ID i,j or IM i,j ;ID i,j ≤0 or IM i,j ≤0; Finally, the preprocessed data set is set to 5. The loan qualification assessment method using digital assets according to claim 4, characterized in that: It also includes constructing digital asset graph nodes and creating a self-created semantic association algorithm, specifically: From the preprocessed data set Generate graph nodes: For each record r i,j =(φ(ID i,j ),ψ(IM i,j )), recorded as graph node A i,j ; Each node A i,j Set the following properties: Contribution attribute α i,j , α i,j =φ(ID i,j ); Interactive properties β i,j , β i,j =ψ(IM i,j ); Set up a semantic association function and calculate the association strength between any two nodes A i,j and A i,k where j < k, specifically as follows: Among them, Λ(A i,j ,A i,k ) are two nodes A i,j With A i,k The strength of association; α i,j , α i,k Node A i,j With A i,k Contribution attribute; β i,j , β i,k Node A i,j With A i,k Interaction properties; min(α i,j ,α i,k )Return α i,j With α i,k The one with the smallest median value; When generating nodes, edges between nodes are established according to the following rules: If for any j < k, Λ(A i,j , A i,k ) > 0, then an edge is established from node A i,j to node A i,k . Let this edge be e i,j,k , the value on the edge is set to L i,j,k =Λ(A i,j ,A i,k ); Set the node set V i for: The edge set is set to: E i = {e i,j,k | j < k and Λ(A i,j , A i,k ) > 0}.

6. The loan qualification assessment method using digital assets according to claim 5, characterized in that: It also includes building holographic maps and node attribute mapping, specifically: S11. Definition of hologram: For each borrower i, construct a holographic spectrum G i , G i =(V i ,E i ); S12. Node attribute injection: Setting up i,j For node A i,j The loan qualification score injected in, and γ i,j =S i ; S13, graph generation process: According to the pre-processing record r i,j Generate node A sequentially i,j , only for those satisfying α i,j =1 record generation; For all node pairs that satisfy j < k, calculate Λ(A i,j , A i,k ); if the result is greater than zero, establish the corresponding edge e i,j,k and assign the edge value L i,j,k ; This generates the completed graph G i =(V i ,E i ).

7. The loan qualification assessment method using digital assets according to claim 6, characterized in that: It also includes the calculation of graph topology values ​​and the generation of loan qualification scores, specifically: For the spectrum G i Any node A i,j ∈V i , set the topology value Θ i,j for: Average the topological values ​​of all nodes in the holographic graph of borrower i and set the overall graph index Ω i for: Among them, |V i | is the node set V i The total number of nodes in ; The overall spectrum index Ω i Directly mapped to the final loan qualification score Q i , specifically: Q i =Ω i .

8. The loan qualification assessment method using digital assets according to claim 7, characterized in that: It also includes the two-dimensional coordinate mapping of graph nodes and data interface output, specifically: The atlas G i Each node A i,j ∈V i Mapped onto a two-dimensional plane, where: Horizontal coordinate x i,j Set to the node's sequence number, i.e. x i,j =j; vertical coordinate y i,j Set to the topological value of the node, y i,j =Θ i,j ; According to the above coordinates, each node is drawn as a point (x i,j ,y i,j ); At the same time, for any edge e in the graph i,j,k , if L i,j,k >0, then draw a straight line on the plane connecting A i,j With A i,k The two corresponding coordinate points; otherwise, no connection A is drawn i,j With A i,k The line connecting the two corresponding coordinate points; Set the output interface as the collection of hologram data of borrower i, denoted as Ξ i , specifically: Ξ i ={(j,x i,j ,y i,j ,Θ i,j )|A i,j ∈V i }∪{Q i }; Each tuple (j,x i,j ,y i,j ,Θ i,j ) records the serial number, two-dimensional coordinates and topological value of the corresponding node; and also adds the final loan qualification score Q i .

Images