Server access method for large model and actual service scene

Abstract

The invention relates to the technical field of large-model and actual service scene server access, and discloses a large-model and actual service scene server access method, which comprises the following steps of: configuring access parameters including protocol type identification, interface calling frequency and data transmission format; obtaining a service interaction data set under the parameter constraint; extracting features of the access parameters to obtain parameter identification vectors; performing integration processing based on feature association on the service interaction data set to obtain a service data association feature vector, wherein the process comprises format conversion, association reference point analysis, association direction calculation and integration; performing multi-dimensional feature cooperative processing on the two vectors to obtain a parameter-business cooperative feature vector, wherein the cooperative processing relates to external business knowledge, an attention mechanism and a Transform structure; and completing access based on the collaborative feature vector, wherein access verification comprises dimension alignment, initial verification and deep verification. According to the method, the compatibility, the dynamic adaptability and the reliability of access are improved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, specifically to a server access method for large models and real-world business scenarios. Background Technology

[0002] With the rapid development of artificial intelligence technology, large-scale models have demonstrated powerful capabilities in various fields such as natural language processing, data analysis, and intelligent decision-making. Their integration with real-world business scenarios has become a key direction for promoting the intelligent upgrading of industries. However, significant technical differences and interaction barriers exist between large-scale model systems and the servers in real-world business scenarios, leading to numerous challenges in achieving efficient integration between the two.

[0003] Real-world business scenarios often involve diverse protocol types, interface specifications, and data formats. For example, e-commerce platforms might use the HTTP protocol for transaction data transmission, while core business interactions in financial systems may rely on proprietary protocols. Furthermore, the frequency of interface calls varies significantly across different business scenarios; high-concurrency scenarios (such as flash sales) and low-frequency interaction scenarios (such as backend data statistics) place drastically different demands on server responsiveness. These differences make it easy for large-scale models to encounter issues such as protocol incompatibility, data transmission delays, and interface call conflicts during integration, severely impacting the stability and efficiency of business processes.

[0004] In existing technologies, the integration of large-scale models with business scenarios often employs fixed configuration schemes, relying on manually preset protocol parameters, interface rules, and data conversion methods. This approach lacks the ability to perceive and adapt to dynamic data characteristics during business interactions. When business scenarios change (such as adjustments to interface call frequency or data format updates), parameter reconfiguration and system debugging are required, resulting in high integration costs and poor flexibility. Furthermore, traditional integration methods often overlook the inherent relationships between business data and the collaborative relationship between integration parameters and business data, relying solely on simple data format conversion to complete the interaction. This makes it difficult to guarantee the accuracy and integrity of data transmission, thereby affecting the large-scale model's understanding and response quality to business requirements.

[0005] With the continuous iteration of large-scale model technology and the increasing complexity of business scenarios, existing access methods can no longer meet the access requirements of high compatibility, high dynamism, and high reliability. Therefore, a server access method that can comprehensively consider access parameters and business data characteristics to achieve multi-dimensional collaborative adaptation is needed to break down the technical barriers between large-scale models and actual business scenarios and improve the interaction efficiency and stability between the two. Summary of the Invention

[0006] The purpose of this invention is to provide a server access method for large models and real-world business scenarios, in order to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides a server access method for large models and real-world business scenarios, the method comprising:

[0008] Configure access parameters, which include protocol type identifier, interface call frequency, and data transmission format;

[0009] Under the constraints of the access parameters, obtain the business interaction data between the target business scenario and the large model system to obtain a business interaction data set;

[0010] The access parameters are subjected to feature extraction processing to obtain parameter identifier vectors;

[0011] The business interaction data set is integrated based on feature association to obtain a business data association feature vector;

[0012] The business data association feature vector and the parameter identifier vector are subjected to multi-dimensional feature collaborative processing to obtain the parameter-business collaborative feature vector;

[0013] Based on the aforementioned parameter-business collaboration feature vector, server access between the large model and actual business scenarios is completed.

[0014] Preferably, the business interaction data set is subjected to feature-based integration processing to obtain a business data association feature vector, including:

[0015] Each business interaction data in the business interaction data set is processed by a format conversion unit to obtain a standard business data unit set.

[0016] Based on the characteristic distribution pattern of the standard business data unit set, analyze the business data association reference point representation vector;

[0017] Calculate the associated motion direction of each standard business data unit in the set of standard business data units relative to the business data association reference point representation vector to obtain the set of business data association directions;

[0018] Based on the set of business data association directions, the set of standard business data units is integrated and processed towards the business data association reference point representation vector to obtain the business data association feature vector.

[0019] Preferably, based on the characteristic distribution pattern of the standard business data unit set, the business data association reference point representation vector is analyzed, including:

[0020] Each standard business data unit in the set of standard business data units is input into the feature distribution measurement module to obtain a set of business data distribution measurement coefficients.

[0021] The set of business data distribution measurement coefficients is input into the feature weight filtering and control unit to obtain the set of business data distribution weight factors;

[0022] Based on the set of business data distribution weight factors, the weighted sum of the set of standard business data units is calculated to obtain the business data association reference point representation vector.

[0023] Preferably, each standard business data unit in the standard business data unit set is input into the feature distribution measurement module to obtain a set of business data distribution measurement coefficients, including:

[0024] The standard business data unit is subjected to numerical standardization to obtain a standardized business data unit;

[0025] The sum of the squares of each feature value in the standardized business data unit is calculated and divided by the feature range value of the standardized business data unit to obtain the business data distribution metric coefficient.

[0026] Preferably, the direction of motion associated with the business data is the inverse cosine similarity value between the standard business data unit and the business data associated reference point representation vector.

[0027] Preferably, the business data association feature vector and the parameter identifier vector are subjected to multi-dimensional feature collaborative processing to obtain a parameter-business collaborative feature vector, including:

[0028] Based on external business knowledge, the business data association feature vector and the parameter identification vector are adjusted by feature interaction based on an attention mechanism to obtain the knowledge optimization parameter-business collaboration feature matrix.

[0029] Based on the knowledge optimization parameter-business collaboration feature matrix, the business data association feature vector and the parameter identifier vector are respectively subjected to feature adjustment and optimization to obtain the optimized business data association feature vector and the optimized parameter identifier vector;

[0030] The parameter-business collaboration feature vector is obtained by dividing the positional points between the optimized business data association feature vector and the optimized parameter identifier vector.

[0031] Preferably, based on external business knowledge, the business data association feature vector and the parameter identifier vector are adjusted using an attention mechanism to obtain a knowledge optimization parameter-business collaboration feature matrix, including:

[0032] The business data associated feature vector and the parameter identifier vector are input into the feature collaborative interaction network to obtain the parameter-business collaborative feature matrix;

[0033] The parameter-business collaboration feature matrix is ​​input into the attention processing unit based on external business knowledge to obtain the knowledge optimization parameter-business collaboration feature matrix.

[0034] Preferably, based on the knowledge optimization parameter-business collaboration feature matrix, feature adjustment and optimization are performed on the business data association feature vector and the parameter identifier vector respectively to obtain the optimized business data association feature vector and the optimized parameter identifier vector, including:

[0035] The business data association feature vector is linearly transformed to obtain a first query feature vector and a first value feature vector. The knowledge optimization parameter-business collaboration feature matrix is ​​used as the key matrix. The first query feature vector, the first value feature vector and the key matrix are input into the feature adjustment module based on the Transformer structure to obtain the optimized business data association feature vector.

[0036] The parameter identifier vector is linearly transformed to obtain a second query feature vector and a second value feature vector. The knowledge optimization parameter-business collaboration feature matrix is ​​used as the key matrix. The second query feature vector, the second value feature vector, and the key matrix are input into the feature adjustment module based on the Transformer structure to obtain the optimized parameter identifier vector.

[0037] Preferably, based on the parameter-business collaboration feature vector, the server access between the large model and the actual business scenario is completed, including:

[0038] The parameter-business collaboration feature vector is input into the access verification module based on the decision maker to obtain the access verification result, which is used to indicate whether the large model and the server of the actual business scenario have completed the access.

[0039] Preferably, the parameter-service collaboration feature vector is input into the access verification module based on the determiner to obtain the access verification result, including:

[0040] The parameter-business collaboration feature vector is subjected to dimension alignment processing to obtain an aligned feature vector;

[0041] The aligned feature vector is input into the base verification layer of the determiner for initial verification to obtain an initial verification mark;

[0042] The initial verification mark is input into the deep verification layer of the decision-maker for secondary verification to obtain the access verification result;

[0043] The step of performing dimension alignment processing on the parameter-business collaboration feature vector to obtain an aligned feature vector includes:

[0044] Extract the preset standard feature dimension parameters;

[0045] Calculate the deviation between the current dimension parameter of the parameter-business collaboration feature vector and the standard feature dimension parameter;

[0046] The alignment feature vector is obtained by adjusting the dimension of the parameter-business collaboration feature vector based on the deviation.

[0047] Compared with the prior art, the beneficial effects of the present invention are:

[0048] This method establishes a standardized starting point for the access process by configuring access parameters such as protocol type identifier, interface call frequency, and data transmission format, and then acquiring business interaction data based on these parameters. This parameterized configuration approach can accurately adapt to the personalized needs of different business scenarios, avoiding protocol conflicts and interface incompatibility issues in traditional fixed configuration modes. It significantly improves the compatibility of large models with servers in various real-world business scenarios, enabling the same large model system to flexibly connect to business platforms in different fields such as e-commerce, finance, and healthcare.

[0049] In the business interaction data processing stage, standard business data units are obtained through format conversion, and reference points are associated based on feature distribution trend analysis. Data integration is then achieved through association direction calculation, ultimately yielding a business data association feature vector. This process fully explores the inherent relationships between business data, transforming scattered and heterogeneous business interaction data into association vectors with unified feature dimensions. This not only eliminates interference caused by data format differences but also strengthens the overall feature expression of business data, enabling the large model to more comprehensively and accurately understand the core needs of the business scenario and providing a high-quality data foundation for subsequent interaction responses.

[0050] By extracting features from access parameters to obtain parameter identifier vectors, and then associating these vectors with business data for multi-dimensional collaborative processing, particularly by introducing external business knowledge and attention mechanisms for feature interaction adjustment, deep integration of access parameters and business data is achieved. This collaborative processing approach overcomes the limitations of traditional methods that separate parameter and data processing, enabling access parameters to dynamically adapt to changes in business data features, while business data can also be optimized according to parameter constraints. The synergy between the two effectively improves the dynamic adaptability of the access process. When changes occur in the interface call frequency or data transmission format of the business scenario, the system can automatically adjust the feature interaction strategy to ensure the stability of the access process.

[0051] In the feature collaborative processing, a feature adjustment module based on the Transformer structure is introduced. Through the interactive calculation of query vectors, value vectors, and key matrices, the expression accuracy of business data association feature vectors and parameter identifier vectors is further optimized, enabling the collaborative feature vectors to more accurately reflect the matching relationship between business requirements and access parameters. Based on this collaborative feature vector, the access verification process employs a multi-level verification mechanism of dimension alignment, initial verification, and deep verification to ensure the reliability of the access results. This effectively reduces the probability of data transmission errors and interface call anomalies, improving the accuracy and security of the interaction between the large model and the business server.

[0052] The entire methodology forms a complete closed loop, from parameter configuration, data acquisition, feature extraction, collaborative optimization to access verification. Each stage is closely connected and mutually supportive, ensuring both the standardization and operability of the access process while providing flexibility and robustness to handle complex business scenario changes. Through this method, large-scale models can be quickly and stably connected to servers in various real-world business scenarios, reducing access costs, improving business interaction efficiency, and providing crucial technical support for the large-scale application of large-scale models in industrial sectors. Attached Figure Description

[0053] Figure 1 This is a schematic diagram illustrating the working principle of the server access method for large models and real-world business scenarios described in this invention.

[0054] Figure 2 A flowchart for associating feature vectors with business data;

[0055] Figure 3 A flowchart illustrating the vector representation of reference points for business data;

[0056] Figure 4 A flowchart for parameter-business collaboration feature vectors;

[0057] Figure 5 This is a flowchart illustrating the association of feature vectors and parameter identifier vectors with optimized business data. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Please see Figures 1-5 This invention provides a server access method for large models and real-world business scenarios, with the following specific implementation steps:

[0060] Configure access parameters, which include protocol type identifier, interface call frequency, and data transmission format.

[0061] Under the constraints of the access parameters, obtain the business interaction data between the target business scenario and the large model system to obtain a business interaction data set.

[0062] The access parameters are subjected to feature extraction processing to obtain parameter identifier vectors.

[0063] The business interaction data set is integrated based on feature association to obtain a business data association feature vector.

[0064] The business data association feature vector and the parameter identifier vector are subjected to multi-dimensional feature collaborative processing to obtain the parameter-business collaborative feature vector.

[0065] Based on the aforementioned parameter-business collaboration feature vector, server access between the large model and actual business scenarios is completed.

[0066] Example 1:

[0067] In this embodiment, when performing feature-based integration processing on a set of business interaction data to obtain a business data association feature vector, a specific process must be followed. Each piece of business interaction data in the set is processed, and these data are passed through a format conversion unit. The format conversion unit has preset format conversion rules; after business interaction data of different formats enters this unit, they are converted into a unified standard format according to these rules, thereby forming a standard set of business data units.

[0068] Based on the characteristic distribution pattern of the standard business data unit set, the representation vector of the business data association benchmark point is analyzed. In this process, each standard business data unit in the set is first input into the characteristic distribution measurement module. Within the characteristic distribution measurement module, each standard business data unit undergoes numerical standardization. Specifically, the value range of each feature in the unit is determined, and each feature value is subtracted from its minimum value, then divided by the difference between the maximum and minimum values. This maps the feature values ​​to a specific interval, resulting in a standardized business data unit.

[0069] After processing the standardized business data unit, the sum of the squares of each feature value is calculated. This sum is then divided by the feature range value of the standardized business data unit, where the feature range value is the difference between the maximum and minimum values ​​of the standardized feature. This calculation yields the business data distribution metric coefficient. Each standardized business data unit is processed in this manner, ultimately resulting in a set of business data distribution metric coefficients.

[0070] The set of business data distribution measurement coefficients is input into the feature weight filtering and control unit. This unit has a specific weight filtering algorithm and rules that evaluate each coefficient in the set. Based on factors such as the magnitude and distribution of the coefficients, a corresponding business data distribution weight factor is assigned to each coefficient, thus obtaining a set of business data distribution weight factors.

[0071] Based on the set of business data distribution weight factors, a weighted sum of the standard business data unit sets is calculated. Specifically, each standard business data unit is multiplied by its corresponding business data distribution weight factor, and then all the multiplication results are summed. The sum obtained is the business data association benchmark point representation vector.

[0072] After obtaining the business data association reference point representation vector, the association motion direction of each standard business data unit in the standard business data unit set relative to this vector is calculated. This association motion direction is determined by calculating the inverse cosine similarity value between the standard business data unit and the business data association reference point representation vector. This calculation is performed on each standard business data unit, ultimately forming a set of business data association directions.

[0073] Based on the set of business data association directions, the standard business data unit set is integrated towards the business data association benchmark point representation vector. During integration, a specific integration algorithm is used to adjust and aggregate each standard business data unit according to its association direction. For example, standard business data units with similar association directions are brought closer together in the feature space during integration, while units with significantly different association directions are adjusted accordingly. Ultimately, these units are integrated into a whole, resulting in a business data association feature vector. This vector integrates the characteristics of the standard business data unit set and their relationships, more comprehensively reflecting the characteristics of business interaction data and providing effective data support for subsequent multi-dimensional feature collaborative processing.

[0074] Example 2:

[0075] This embodiment follows specific processing logic and steps when performing multi-dimensional feature collaboration processing on business data association feature vectors and parameter identifier vectors to obtain parameter-business collaborative feature vectors. Based on external business knowledge, the business data association feature vectors and parameter identifier vectors are adjusted through attention-based feature interaction to obtain a knowledge-optimized parameter-business collaborative feature matrix. Specifically, the business data association feature vectors and parameter identifier vectors are input into a feature collaboration interaction network. This network adopts a specific neural network architecture, containing multiple layers of neurons and connection weights, enabling feature interaction processing on the two input vectors. During network operation, the various features in the business data association feature vectors and parameter identifier vectors interact with each other, generating a parameter-business collaborative feature matrix through weighted summation, activation function operations, and other operations. This matrix integrates the feature information of the two vectors, reflecting the initial association between business data features and access parameter features.

[0076] The parameter-business collaboration feature matrix is ​​input into an attention processing unit based on external business knowledge. This unit stores pre-organized external business knowledge, covering industry rules, business processes, data characteristics, and other information related to the target business scenario. The attention processing unit analyzes each feature in the parameter-business collaboration feature matrix based on this external business knowledge, calculating the attention weight for each feature. Features closely related to business knowledge and more important for business understanding and processing are assigned higher weights; features with lower relevance and relatively weaker importance are assigned lower weights. In this way, important features are highlighted and secondary features are suppressed, resulting in a knowledge-optimized parameter-business collaboration feature matrix. This matrix, optimized by business knowledge, more accurately reflects the key correlation features between the business scenario and the access parameters.

[0077] After obtaining the knowledge optimization parameter-business collaboration feature matrix, it is necessary to perform feature adjustment and optimization on the business data association feature vector and parameter identifier vector based on this matrix to obtain the optimized business data association feature vector and the optimized parameter identifier vector. First, the business data association feature vector is processed by performing a linear transformation. The linear transformation is achieved through a preset transformation matrix, mapping the business data association feature vector to a new feature space, thereby obtaining the first query feature vector and the first value feature vector. Simultaneously, the knowledge optimization parameter-business collaboration feature matrix is ​​used as the key matrix.

[0078] The first query feature vector, the first value feature vector, and the key matrix are input into a feature adjustment module based on a Transformer architecture. The Transformer architecture includes components such as a self-attention mechanism and a feedforward neural network. In the feature adjustment module, the first query feature vector undergoes self-attention calculation with the key matrix to determine the attention weight for each feature position, thus focusing on the parts closely related to the features in the key matrix. Then, the first value feature vector is weighted and summed according to the attention weights to obtain the adjusted feature representation. Further processing by the feedforward neural network performs nonlinear transformations and dimensional adjustments on the features, ultimately yielding the optimized business data association feature vector.

[0079] The parameter identifier vector is processed. A linear transformation is performed on the parameter identifier vector, using another preset transformation matrix to convert it into a second query feature vector and a second value feature vector. Similarly, using the knowledge optimization parameter-business collaboration feature matrix as the key matrix, the second query feature vector, the second value feature vector, and the key matrix are input into the feature adjustment module based on the Transformer structure. In this module, following the same process as processing the business data associated feature vector, steps such as self-attention calculation, weighted summation, and feedforward neural network processing are performed to obtain the optimized parameter identifier vector.

[0080] The optimized business data association feature vector and the optimized parameter identifier vector are divided element-wise at corresponding positions to obtain the parameter-business collaborative feature vector. This element-wise division operation divides the corresponding elements of the two vectors to obtain a new vector element. This operation highlights the feature differences and correlations between the two vectors, allowing the parameter-business collaborative feature vector to more comprehensively reflect the collaborative relationship between business data features and access parameter features. Throughout the process, each step depends on the result of the previous step. Through the collaborative work of components such as the feature collaboration interaction network, the attention processing unit, and the feature adjustment module of the Transformer structure, multi-dimensional feature collaboration processing of the business data association feature vector and the parameter identifier vector is achieved, ultimately resulting in a parameter-business collaborative feature vector that can effectively support the access of large models and servers in actual business scenarios.

[0081] Example 3:

[0082] In this embodiment, when completing the server access between a large model and a real business scenario based on the parameter-business collaboration feature vector, the parameter-business collaboration feature vector needs to be input into the access verification module based on the decision-maker to obtain the access verification result. The specific implementation method is as follows: The parameter-business collaboration feature vector is subjected to dimension alignment processing to ensure that its dimensions match the preset standard feature dimensions. When performing dimension alignment processing, the preset standard feature dimension parameters are first extracted. These parameters are preset according to the requirements of the large model and the actual business scenario, specifying the expected number of feature dimensions and the meaning of each dimension.

[0083] The calculation parameter is the deviation between the current dimension parameter of the business collaboration feature vector and the standard feature dimension parameter. The current dimension parameter includes information such as the number of dimensions in the feature vector and the value range of each dimension. By comparing the current dimension parameter with the standard feature dimension parameter, the differences between the two are determined, such as whether the number of dimensions is consistent and whether the value range of each dimension is within the allowable error range.

[0084] Based on the calculated deviation, the dimensions of the parameter-business collaboration feature vector are adjusted. If the number of dimensions of the current dimension parameter is greater than that of the standard feature dimension parameter, it may be necessary to remove some unimportant dimensions through feature selection or dimensionality reduction algorithms; if the number of dimensions is less than that of the standard feature dimension parameter, it is necessary to supplement the missing dimensions through feature generation or interpolation. For the deviation of the value range of each dimension, normalization or standardization methods can be used to adjust it so that the adjusted feature vector dimensions are consistent with the standard feature dimension parameters, thereby obtaining an aligned feature vector.

[0085] After dimensional alignment is completed, the aligned feature vectors are input into the base validation layer of the decision maker for initial validation. The base validation layer of the decision maker has a series of initial validation rules. These rules mainly validate the basic characteristics of the feature vectors, such as whether there are obvious outliers and whether the distribution of feature values ​​conforms to basic statistical laws. The base validation layer checks each input aligned feature vector and generates initial validation markers according to the preset rules. These initial validation markers record the validation results of the aligned feature vectors in the base validation layer, such as which features meet the requirements and which features have potential problems.

[0086] The initial verification markers are input into the deep verification layer of the decision maker for secondary verification. The deep verification layer analyzes and verifies the initial verification markers from a more in-depth perspective, containing more complex verification rules and algorithms. The deep verification layer not only considers the basic features of the feature vectors but also evaluates the business logic and parameter matching reflected by the feature vectors, taking into account the specific needs of the large model and actual business scenarios. For example, it checks whether the collaborative relationship between business data association features and access parameter features meets the requirements of the business scenario, and whether the parameter-business collaborative feature vectors accurately reflect the interaction requirements between the large model and the actual business scenario.

[0087] The deep verification layer performs detailed analysis and complex calculations on the initial verification markers to obtain the final access verification result. This result indicates whether the large model has successfully connected with the server in the actual business scenario. If the verification result shows compliance with all verification rules and requirements, the large model has successfully connected with the server in the actual business scenario; otherwise, the connection has failed.

[0088] Throughout the implementation process, each step is closely linked. Dimension alignment is to ensure that the format of the feature vectors meets the requirements of the verification module. The initial verification of the basic verification layer is to perform preliminary screening of the basic features of the feature vectors. The secondary verification of the deep verification layer is to conduct in-depth analysis from the perspective of business logic and actual needs, and finally obtain accurate access verification results.

[0089] Example 4:

[0090] In this embodiment, when inputting each standard business data unit in the standard business data unit set into the feature distribution measurement module to obtain the business data distribution measurement coefficient set, specific processing operations need to be performed for each standard business data unit. Taking user shopping behavior data from an e-commerce platform as an example, assuming that the standard business data unit set contains shopping record data of multiple users, each standard business data unit corresponds to a user's shopping record, which includes features such as product category, purchase quantity, consumption amount, and shopping time.

[0091] For one of the standard business data units, namely a user's shopping record data, numerical standardization is performed. Taking the consumption amount feature as an example, the value range of this feature in the standard business data unit may be 0 to 10,000 yuan. To map it to a specific range, a common standardization method, such as min-max standardization, is used. Specifically, the minimum value of this feature is determined to be 0, and the maximum value is 10,000. The user's consumption amount is subtracted from the minimum value of 0, and then divided by the difference between the maximum and minimum values, 10,000, to obtain the standardized consumption amount value. Assuming the user's consumption amount is 5,000 yuan, the standardized value is (5,000-0) / (10,000-0) = 0.5. Similarly, standardization is performed on other features such as product category, purchase quantity, and shopping time. For non-numerical features such as product category, it may be necessary to encode them first, such as encoding different product categories with different values, before standardization. If the purchase quantity feature ranges from 1 to 100, it is also processed using the min-max standardization method. Shopping time characteristics may need to be converted into numerical data first, such as the number of days from a certain baseline time, and then standardized. Through this process, each feature value of the entire standard business data unit is mapped to a specific interval, such as [0,1], to obtain the standardized business data unit.

[0092] After processing the standardized business data units, calculate the sum of the squares of all feature values ​​in that unit. Continuing with the user's shopping record example, assume the standardized spending amount is 0.5, the standardized product category code is 0.3, the standardized purchase quantity is 0.6, and the standardized shopping time is 0.4. Then the sum of the squares of all feature values ​​is 0.5. 2 +0.3 2 +0.6 2 +0.4 2 =0.25+0.09+0.36+0.16=0.86.

[0093] Calculate the feature range value for the standardized business data unit. The feature range value is the difference between the maximum and minimum values ​​of each standardized feature. For example, in this standardized business data unit, the maximum standardized feature value is 0.6 (purchase quantity feature), and the minimum value is 0.3 (product category feature). Therefore, the feature range value is 0.6 - 0.3 = 0.3.

[0094] Dividing the sum of the squares of all eigenvalues ​​by the eigenvalue range value yields the business data distribution metric coefficient. That is, 0.86 / 0.3 ≈ 2.87. Therefore, the business data distribution metric coefficient corresponding to this standard business data unit is approximately 2.87.

[0095] The same method is used to process each standard business data unit in the standard business data unit set. For example, in another user's shopping record, the purchase amount is 8000 yuan, which is standardized to 0.8; the product category code is standardized to 0.7; the purchase quantity is standardized to 0.9; and the shopping time is standardized to 0.5. The sum of the squares of all feature values ​​is 0.8. 2 +0.7 2 +0.9 2 +0.5 2 =0.64+0.49+0.81+0.25=2.19. Assuming the maximum value of the characteristic of the business data unit in this specification is 0.9, the minimum value is 0.5, the characteristic range is 0.9-0.5=0.4, and the business data distribution metric coefficient is 2.19 / 0.4=5.475.

[0096] For example, in the third user's shopping record, the purchase amount is 2000 yuan, which is standardized to 0.2; the product category code is standardized to 0.4; the purchase quantity is standardized to 0.3; and the shopping time is standardized to 0.2. The sum of the squares of all feature values ​​is 0.2. 2 +0.4 2 +0.3 2 +0.2 2 =0.04+0.16+0.09+0.04=0.33. The maximum value of the feature is 0.4, the minimum value is 0.2, the feature range is 0.4-0.2=0.2, and the business data distribution metric coefficient is 0.33 / 0.2=1.65.

[0097] After processing all standard business data units in the standard business data unit set according to the above steps, the business data distribution metric coefficients obtained from each unit are collected to form the business data distribution metric coefficient set. In this e-commerce platform example, the business data distribution metric coefficient set may contain multiple values, such as 2.87, 5.475, 1.65, etc., each value corresponding to the business data distribution metric coefficient of a user's shopping record data.

[0098] This process enables the input of each unit in the standard business data unit set into the feature distribution measurement module to obtain a set of business data distribution measurement coefficients. For specific business data features, this process accurately obtains the business data distribution measurement coefficients for each standard business data unit through standardization, calculation of quadratic sums, and feature range values. This provides the foundational data for subsequent feature weight selection and the calculation of the business data association benchmark representation vector. The entire process, combined with specific business scenario examples, details the operation methods and calculation procedures for each step, ensuring the operability and understandability of the implementation.

[0099] Example 5:

[0100] This embodiment requires a detailed explanation of the operation process when analyzing the feature distribution trend of business data related to the benchmark point representation vector based on the standard business data unit set. Taking a loan application business of a financial institution as an example, assume that the standard business data unit set contains information data of multiple loan applicants, and each standard business data unit corresponds to the information of one applicant, including features such as income status, credit score, loan amount, and loan term.

[0101] After obtaining the set of business data distribution measurement coefficients, they are input into the feature weight filtering control unit. For example, this set contains business data distribution measurement coefficients for three standard business data units: 3.2 for applicant A, 5.8 for applicant B, and 2.1 for applicant C. The feature weight filtering control unit internally pre-sets weight filtering rules based on business logic. For instance, it determines the weight factor based on the magnitude of the business data distribution measurement coefficients; the larger the coefficient, the greater the impact of the feature distribution of that data unit on the overall picture, and the higher the corresponding weight factor should be.

[0102] The feature weight screening control unit evaluates these three coefficients. For applicant A's coefficient of 3.2, according to the rules, it may be converted into a corresponding business data distribution weight factor of 0.3; applicant B's coefficient of 5.8 is relatively large, so it is converted into a weight factor of 0.5; applicant C's coefficient of 2.1 is relatively small, so it is converted into a weight factor of 0.2. This results in the set of business data distribution weight factors, namely [0.3, 0.5, 0.2].

[0103] Next, based on the set of business data distribution weight factors, a weighted sum of the standard business data unit sets is calculated to obtain the business data association benchmark representation vector. Each standard business data unit contains multiple features. Taking income status, credit score, loan amount, and loan term as an example, let's assume:

[0104] Applicant A's standardized data are: Income 0.6, Credit Score 0.7, Loan Amount 0.4, Loan Term 0.5;

[0105] Applicant B's standardized data are: Income 0.8, Credit Score 0.9, Loan Amount 0.6, Loan Term 0.7;

[0106] The standardized data for applicant C are as follows: income status 0.3, credit score 0.4, loan amount 0.2, and loan term 0.3.

[0107] When calculating the weighted sum, the weights are calculated separately for each feature:

[0108] The weighted sum of income is: 0.6×0.3+0.8×0.5+0.3×0.2=0.18+0.4+0.06=0.64;

[0109] The weighted sum of the credit scores is: 0.7×0.3+0.9×0.5+0.4×0.2=0.21+0.45+0.08=0.74;

[0110] The weighted sum of the loan amounts is: 0.4×0.3+0.6×0.5+0.2×0.2=0.12+0.3+0.04=0.46;

[0111] The weighted sum of the loan terms is: 0.5×0.3+0.7×0.5+0.3×0.2=0.15+0.35+0.06=0.56.

[0112] These weighted sums are combined to obtain the business data correlation benchmark vector [0.64, 0.74, 0.46, 0.56]. This vector reflects the overall characteristic distribution of the standard business data unit set in the loan application business, such as the comprehensive characteristic value of income status being 0.64 and the comprehensive characteristic value of credit score being 0.74.

[0113] Taking another medical diagnostic business scenario as an example, the standard business data unit set contains the patient's diagnostic data, and each unit has features such as body temperature, blood pressure, heart rate, and blood oxygen saturation. Assume that the business data distribution metric coefficient set is [4.5, 3.7, 6.2], corresponding to the business data distribution metric coefficients of three patients. The feature weight filtering and control unit converts it into a weight factor set [0.25, 0.2, 0.55] according to the rules.

[0114] The standardized feature values ​​for each standard business data unit are as follows:

[0115] Patient A: Temperature 0.7°C, Blood Pressure 0.6°C, Heart Rate 0.8°C, Blood Oxygen Saturation 0.9°C;

[0116] Patient B: Temperature 0.5°C, Blood Pressure 0.7°C, Heart Rate 0.6°C, Blood Oxygen Saturation 0.8°C;

[0117] Patient C: Temperature 0.9, Blood pressure 0.8, Heart rate 0.9, Blood oxygen saturation 0.95.

[0118] Calculate the weighted sum of all features:

[0119] Body temperature: 0.7×0.25+0.5×0.2+0.9×0.55=0.175+0.1+0.495=0.77;

[0120] Blood pressure: 0.6×0.25+0.7×0.2+0.8×0.55=0.15+0.14+0.44=0.73;

[0121] Heart rate: 0.8×0.25+0.6×0.2+0.9×0.55=0.2+0.12+0.495=0.815;

[0122] Blood oxygen saturation: 0.9×0.25+0.8×0.2+0.95×0.55=0.225+0.16+0.5225=0.9075.

[0123] The obtained business data association benchmark vector is [0.77, 0.73, 0.815, 0.9075]. This vector integrates the diagnostic data feature distribution of the three patients and reflects the overall feature trend. For example, the comprehensive value of body temperature is 0.77, which is close to the standardized value of the normal body temperature range. The comprehensive value of blood oxygen saturation is relatively high, reflecting the good condition of the overall patient group in this feature.

[0124] In practice, the rules for feature weight selection and control units can be more complex, requiring consideration of the actual meaning and importance of business data. For example, in financial lending, credit scoring features may be assigned higher weights, while in medical diagnosis, blood oxygen saturation might be a more critical feature and would be given priority during weight selection. By processing the business data distribution measurement coefficients through the feature weight selection and control unit, and combining them with the rules of specific business scenarios, a reasonable set of business data distribution weight factors is generated. Then, through weighted summation, a business data association benchmark vector is obtained. This vector effectively aggregates the feature information of the standard business data unit set, providing crucial foundational data for subsequent calculations of the association direction of standard business data units relative to this vector and for integration processing. This ensures that the overall feature distribution of business data can be accurately grasped during the server access process between the large model and the actual business scenario, supporting the smooth progress of subsequent steps.

[0125] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0126] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A server access method for large models and real-world business scenarios, characterized in that, include: Configure access parameters, which include protocol type identifier, interface call frequency, and data transmission format; Under the constraints of the access parameters, obtain the business interaction data between the target business scenario and the large model system to obtain a business interaction data set; The access parameters are subjected to feature extraction processing to obtain parameter identifier vectors; The business interaction data set is integrated based on feature association to obtain a business data association feature vector; The business data association feature vector and the parameter identifier vector are subjected to multi-dimensional feature collaborative processing to obtain the parameter-business collaborative feature vector; Based on the aforementioned parameter-business collaboration feature vector, server access between the large model and actual business scenarios is completed.

2. The server access method for large models and actual business scenarios according to claim 1, characterized in that, The business interaction data set is integrated based on feature association to obtain a business data association feature vector, including: Each business interaction data in the business interaction data set is processed by a format conversion unit to obtain a standard business data unit set. Based on the characteristic distribution pattern of the standard business data unit set, analyze the business data association reference point representation vector; Calculate the associated motion direction of each standard business data unit in the set of standard business data units relative to the business data association reference point representation vector to obtain the set of business data association directions; Based on the set of business data association directions, the set of standard business data units is integrated and processed towards the business data association reference point representation vector to obtain the business data association feature vector.

3. The server access method for large models and actual business scenarios according to claim 2, characterized in that, Based on the characteristic distribution pattern of the aforementioned standard business data unit set, the business data association reference point representation vector is analyzed, including: Each standard business data unit in the set of standard business data units is input into the feature distribution measurement module to obtain a set of business data distribution measurement coefficients. The set of business data distribution measurement coefficients is input into the feature weight filtering and control unit to obtain the set of business data distribution weight factors; Based on the set of business data distribution weight factors, the weighted sum of the set of standard business data units is calculated to obtain the business data association reference point representation vector.

4. The server access method for large models and actual business scenarios according to claim 3, characterized in that, Each standard business data unit in the set of standard business data units is input into the feature distribution measurement module to obtain a set of business data distribution measurement coefficients, including: The standard business data unit is subjected to numerical standardization to obtain a standardized business data unit; The sum of the squares of each feature value in the standardized business data unit is calculated and divided by the feature range value of the standardized business data unit to obtain the business data distribution metric coefficient.

5. The server access method for large models and actual business scenarios according to claim 4, characterized in that, The direction of motion associated with the business data is the inverse cosine similarity value between the standard business data unit and the business data associated reference point representation vector.

6. The server access method for large models and actual business scenarios according to claim 5, characterized in that, The business data association feature vector and the parameter identifier vector are subjected to multi-dimensional feature collaborative processing to obtain a parameter-business collaborative feature vector, including: Based on external business knowledge, the business data association feature vector and the parameter identification vector are adjusted by feature interaction based on an attention mechanism to obtain the knowledge optimization parameter-business collaboration feature matrix. Based on the knowledge optimization parameter-business collaboration feature matrix, the business data association feature vector and the parameter identifier vector are respectively subjected to feature adjustment and optimization to obtain the optimized business data association feature vector and the optimized parameter identifier vector; The parameter-business collaboration feature vector is obtained by dividing the positional points between the optimized business data association feature vector and the optimized parameter identifier vector.

7. The server access method for large models and actual business scenarios according to claim 6, characterized in that, Based on external business knowledge, the business data association feature vector and the parameter identifier vector are adjusted using an attention mechanism to obtain a knowledge optimization parameter-business collaborative feature matrix, including: The business data associated feature vector and the parameter identifier vector are input into the feature collaborative interaction network to obtain the parameter-business collaborative feature matrix; The parameter-business collaboration feature matrix is ​​input into the attention processing unit based on external business knowledge to obtain the knowledge optimization parameter-business collaboration feature matrix.

8. The server access method for large models and actual business scenarios according to claim 7, characterized in that, Based on the knowledge optimization parameter-business collaboration feature matrix, the business data association feature vector and the parameter identifier vector are respectively subjected to feature adjustment and optimization to obtain the optimized business data association feature vector and the optimized parameter identifier vector, including: The business data association feature vector is linearly transformed to obtain a first query feature vector and a first value feature vector. The knowledge optimization parameter-business collaboration feature matrix is ​​used as the key matrix. The first query feature vector, the first value feature vector and the key matrix are input into the feature adjustment module based on the Transformer structure to obtain the optimized business data association feature vector. The parameter identifier vector is linearly transformed to obtain a second query feature vector and a second value feature vector. The knowledge optimization parameter-business collaboration feature matrix is ​​used as the key matrix. The second query feature vector, the second value feature vector, and the key matrix are input into the feature adjustment module based on the Transformer structure to obtain the optimized parameter identifier vector.

9. The server access method for large models and actual business scenarios according to claim 8, characterized in that, Based on the aforementioned parameter-business collaboration feature vector, server integration between the large model and actual business scenarios is completed, including: The parameter-business collaboration feature vector is input into the access verification module based on the decision maker to obtain the access verification result, which is used to indicate whether the large model and the server of the actual business scenario have completed the access.

10. The server access method for large models and actual business scenarios according to claim 9, characterized in that, The parameter-service collaboration feature vector is input into the access verification module based on the determiner to obtain the access verification result, including: The parameter-business collaboration feature vector is subjected to dimension alignment processing to obtain an aligned feature vector; The aligned feature vector is input into the base verification layer of the determiner for initial verification to obtain an initial verification mark; The initial verification mark is input into the deep verification layer of the decision-maker for secondary verification to obtain the access verification result; The step of performing dimension alignment processing on the parameter-business collaboration feature vector to obtain an aligned feature vector includes: Extract the preset standard feature dimension parameters; Calculate the deviation between the current dimension parameter of the parameter-business collaboration feature vector and the standard feature dimension parameter; The alignment feature vector is obtained by adjusting the dimension of the parameter-business collaboration feature vector based on the deviation.

Images