Method and system for intelligent transaction of public accumulation fund business

Abstract

The invention relates to a method and a system for intelligent transaction of a public accumulation fund service, and relates to the technical field of processing of the public accumulation fund service. The method mainly comprises the following steps: in response to an accumulation fund region transfer request initiated by a user, obtaining user basic data, service data, operation data, network data and equipment data of the user; determining a feature vector set and hash summary information according to the user basic data, the service data, the operation data, the network data and the equipment data of the user; determining an original risk score corresponding to the accumulation fund region transfer request according to the feature vector set; determining a final risk score corresponding to the accumulation fund region transfer request through the original risk score, the user basic data, the network data and the equipment data; and executing a service operation corresponding to the accumulation fund region transfer request according to a decision instruction corresponding to the final risk score, and storing the Hash summary information and the decision instruction in a block chain.

Description

Technical Field

[0001] This invention relates to the field of processing technology for housing provident fund business, and more specifically to a method and system for intelligent processing of housing provident fund business. Background Technology

[0002] Housing provident fund refers to the long-term housing savings deposited by state organs, state-owned enterprises, urban collective enterprises, foreign-invested enterprises, urban private enterprises and other urban enterprises, public institutions, private non-enterprise units, social organizations and their employees.

[0003] With the continuous development of the social economy, the mobility of employees is becoming increasingly greater, often moving from one region to another. At the same time, it is necessary to transfer housing provident funds from the original region to the new region. In the process of automated transfer, how to accurately verify the employee's identity and intention, and ensure the safety of funds during the transfer process, is an important technical consideration. Summary of the Invention

[0004] The present invention aims to provide a method and system for intelligent processing of housing provident fund business, so as to solve the shortcomings of the existing technology. The technical problem to be solved by the present invention is achieved through the following technical solution.

[0005] This invention provides a method for intelligent processing of housing provident fund business, the method comprising:

[0006] In response to a user's request to transfer their housing provident fund to another region, the system obtains the user's basic data, business data, operation data, network data, and device data.

[0007] The feature vector set and hash digest information are determined based on the user's basic data, business data, operation data, network data, and device data.

[0008] The original risk score corresponding to the housing provident fund regional transfer request is determined based on the set of feature vectors.

[0009] The final risk score corresponding to the housing provident fund regional transfer request is determined by the original risk score, the user's basic data, the network data, and the device data.

[0010] The business operation corresponding to the housing provident fund regional transfer request is executed according to the ruling instruction corresponding to the final risk score, and the hash digest information and the ruling instruction are stored in the blockchain.

[0011] In an optional embodiment, determining the feature vector set and hash digest information based on the user's basic data, business data, operation data, network data, and device data includes:

[0012] The user's basic data, business data, operation data, network data, and device data are preprocessed, and the preprocessing includes at least data cleaning and normalization.

[0013] The preprocessed user basic data, business data, operation data, network data, and device data are each converted into feature vectors; and all feature vectors are combined to obtain the feature vector set.

[0014] The user's basic data, business data, operation data, network data, and device data, which have been preprocessed, are hashed using a cryptographic hash function to obtain hash digest information.

[0015] In an optional embodiment, determining the original risk score corresponding to the housing provident fund regional transfer request based on the feature vector set includes:

[0016] Extract the main feature vector and the meta-context vector from the feature vector set, wherein the meta-context vector is a vector determined based on business data; and the main feature vector is a vector in the feature vector set other than the meta-context vector.

[0017] The dynamic modulation signal is determined based on the meta-context vector, and the abstract features are determined based on the principal feature vector.

[0018] The original risk score corresponding to the housing provident fund regional transfer request is determined by the dynamic modulation signal and the defined abstract features.

[0019] In an optional embodiment, determining the dynamic modulation signal based on the meta-context vector includes:

[0020] The meta-context vector is input into the input layer of the housing provident fund risk model, and the input layer inputs the meta-context vector into the hidden layer of the housing provident fund risk model.

[0021] The activation value of the hidden layer is obtained by performing a linear transformation on the meta-context vector through the hidden layer in the aforementioned housing provident fund risk model and then following a nonlinear activation function.

[0022] The dynamic modulation signal is determined by the activation value of the hidden layer, and the dynamic modulation signal includes a scaling factor and an offset.

[0023] In an optional embodiment, determining the abstract features based on the principal feature vector includes:

[0024] The meta-context vector is input into the input layer of the housing provident fund risk model, and the input layer inputs the meta-context vector into the fully connected layer of the housing provident fund risk model.

[0025] The abstract features are obtained by performing multi-layer transformations on the meta-context vector through the fully connected layer in the housing provident fund risk model.

[0026] In an optional embodiment, determining the original risk score corresponding to the housing provident fund regional transfer request through the dynamic modulation signal and the determined abstract features includes:

[0027] The abstract feature is dynamically calibrated using the dynamic modulation signal to obtain the calibrated abstract feature;

[0028] The original risk score corresponding to the housing provident fund regional transfer request is obtained by calculating the calibration abstract features using an activation function.

[0029] In an optional embodiment, determining the final risk score corresponding to the housing provident fund regional transfer request using the original risk score, the user's basic data, the network data, and the device data includes:

[0030] The user's basic data, the network data, and the device data are matched with the hard rules in the preset hard rule table.

[0031] If any hard rule is successfully matched, the final risk score corresponding to the housing provident fund regional transfer request is determined according to the matched hard rule;

[0032] If no hard rule is successfully matched, the rule adjustment factor is determined based on the user's basic data, the network data, and the device data.

[0033] The final risk score corresponding to the housing provident fund regional transfer request is determined by the rule adjustment factor and the original risk score.

[0034] In an optional embodiment, determining the rule adjustment factor based on the user's basic data, the network data, and the device data includes:

[0035] Each piece of data in the user basic data, network data, and device data is matched with the elastic rules in the preset elastic rule table to determine the adjustment contribution value corresponding to each piece of data.

[0036] The rule adjustment factor is obtained by summing up all the adjustment contribution values.

[0037] In an optional embodiment, determining the final risk score corresponding to the housing provident fund regional transfer request using the rule adjustment factor and the original risk score includes:

[0038] The temporary score is obtained by adding the rule adjustment factor and the original risk score;

[0039] According to the formula Determine the final risk score corresponding to the housing provident fund regional transfer request;

[0040] in, The final risk score, The temporary score is denoted as .

[0041] This invention provides a system for intelligent processing of housing provident fund business, the system comprising:

[0042] The acquisition module is used to respond to a user's request for a transfer of housing provident fund regions and acquire the user's basic data, business data, operation data, network data, and device data.

[0043] The first determining module is used to determine a set of feature vectors and hash digest information based on the user's basic data, business data, operation data, network data, and device data.

[0044] The second determining module is used to determine the original risk score corresponding to the housing provident fund regional transfer request based on the feature vector set;

[0045] The third determining module is used to determine the final risk score corresponding to the housing provident fund regional transfer request through the original risk score, the user's basic data, the network data, and the device data;

[0046] The execution module is used to execute the business operation corresponding to the housing provident fund regional transfer request according to the ruling instruction corresponding to the final risk score, and store the hash digest information and the ruling instruction in the blockchain.

[0047] The embodiments of the present invention have the following advantages:

[0048] This invention provides a method and system for intelligent processing of housing provident fund business. In response to a user's request for a housing provident fund regional transfer, the system acquires the user's basic data, business data, operation data, network data, and device data. Then, based on the user's basic data, business data, operation data, network data, and device data, it determines a set of feature vectors and hash digest information. Next, it determines the original risk score corresponding to the housing provident fund regional transfer request based on the feature vector set, and then determines the final risk score corresponding to the housing provident fund regional transfer request using the original risk score, the user's basic data, the network data, and the device data. Finally, it executes the business operation corresponding to the housing provident fund regional transfer request according to the ruling instruction corresponding to the final risk score, and stores the hash digest information and the ruling instruction in the blockchain. Compared to existing technologies that require manual verification of housing provident fund regional transfer requests, this application, upon receiving a housing provident fund regional transfer request initiated by a user, determines the final risk score corresponding to the housing provident fund regional transfer request based on the original risk score, user basic data, network data, and device data. Then, it executes the business operation corresponding to the housing provident fund regional transfer request according to the ruling instruction corresponding to the final risk score. Thus, this application can improve the security of housing provident fund regional transfers. Attached Figure Description

[0049] Figure 1 This is a flowchart of a method for intelligent processing of housing provident fund business provided by an embodiment of the present invention;

[0050] Figure 2 This is a schematic diagram of the structure of a system for intelligent processing of housing provident fund business provided by an embodiment of the present invention. Detailed Implementation

[0051] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0052] like Figure 1 As shown in this embodiment, a method for intelligent processing of housing provident fund business is provided, which involves performing the following steps:

[0053] S101, in response to a user-initiated request to transfer housing provident fund regions, obtain the user's basic data, business data, operation data, network data, and device data.

[0054] In this embodiment, the transfer of housing provident fund to a different region refers to the act of an employee transferring funds and contribution information from their housing provident fund account in their original place of work to the housing provident fund center in their new place of work due to job relocation or household registration change.

[0055] The user's basic data may include user ID, name, age, years of contribution, contribution base, account balance, etc.; business data may include user ID, historical contribution records, historical withdrawal times, loan records, recent transfer times, etc.; operation data may include the time of this application, filling speed, number of modifications, etc.; network data may include the geographical location of the IP address, IP reputation, etc.; device data may include the device's unique identifier, APP version, etc. This embodiment does not specifically limit the above data.

[0056] S102, determine the feature vector set and hash digest information based on the user's basic data, business data, operation data, network data and device data.

[0057] In one optional embodiment provided in this application, determining the feature vector set and hash digest information based on the user's basic data, business data, operation data, network data, and device data includes: preprocessing the user's basic data, business data, operation data, network data, and device data, wherein the preprocessing includes at least data cleaning and normalization; converting the preprocessed user's basic data, business data, operation data, network data, and device data into feature vectors respectively; combining all feature vectors to obtain the feature vector set; and using an encrypted hash function to perform hash calculation on the preprocessed user's basic data, business data, operation data, network data, and device data to obtain hash digest information. Specifically, this embodiment can use an encrypted hash function (such as SHA-256) to perform hash calculation on the preprocessed user's basic data, business data, operation data, network data, and device data to obtain hash digest information.

[0058] S103, determine the original risk score corresponding to the housing provident fund regional transfer request based on the feature vector set.

[0059] The original risk score is used to assess the risk of a housing provident fund regional transfer request. The higher the original risk score, the greater the risk of the housing provident fund regional transfer; the lower the original risk score, the lower the risk of the housing provident fund regional transfer.

[0060] In one optional embodiment provided in this application, determining the original risk score corresponding to the housing provident fund regional transfer request based on the feature vector set includes:

[0061] S1031, extract the main feature vector and the meta-context vector from the feature vector set.

[0062] The meta-context vector C is a vector determined based on business data, with a dimension of d_c (lower). It describes the attributes of the current business and serves as the key to the modulator. The main feature vector F is the vector in the feature vector set excluding the meta-context vector, with a dimension of d_f, containing standardized features from all data in the feature vector set.

[0063] S1032, determine the dynamic modulation signal based on the meta-context vector, and determine the abstract features based on the principal feature vector.

[0064] Specifically, determining the dynamic modulation signal based on the meta-context vector includes: inputting the meta-context vector into the input layer of the provident fund risk model, wherein the input layer inputs the meta-context vector into the hidden layer of the provident fund risk model; performing a linear transformation on the meta-context vector through the hidden layer in the provident fund risk model and then applying a nonlinear activation function to obtain the hidden layer activation value; and determining the dynamic modulation signal based on the hidden layer activation value, wherein the dynamic modulation signal includes a scaling factor and an offset.

[0065] In this implementation, the process by which the hidden layer determines the dynamically modulated signal is as follows: ;

[0066] in, The activation value of the hidden layer. For activation function, and For training parameters, This is the meta-context vector;

[0067] The dynamic modulation signal is then determined using the activation values ​​of the hidden layer. ; ;

[0068] in, As a scaling factor, the Softplus function is used to ensure the output is positive, and incrementing by 1 initializes it to be close to 1, facilitating stable training. and For training parameters; This is the offset, used for translation modulation of the feature. and These are the training parameters.

[0069] Specifically, determining the abstract features based on the principal feature vector includes: inputting the meta-context vector into the input layer of the housing provident fund risk model, wherein the input layer inputs the meta-context vector into the fully connected layer of the housing provident fund risk model; and performing multi-level transformations on the meta-context vector through the fully connected layer in the housing provident fund risk model to obtain the abstract features.

[0070] The fully connected layer is used to extract complex, abstract features from the original high-dimensional features. The extraction process of these abstract features is shown in the following formula: ;

[0071] in, Principal feature vectors and These are the trainable weights and biases for each layer. These are abstract features obtained through multiple layers of computation.

[0072] S1033, determine the original risk score corresponding to the housing provident fund regional transfer request through the dynamic modulation signal and the determined abstract features.

[0073] Specifically, determining the original risk score corresponding to the housing provident fund regional transfer request through the dynamic modulation signal and the determined abstract feature includes: dynamically calibrating the abstract feature through the dynamic modulation signal to obtain a calibrated abstract feature; and calculating the calibrated abstract feature through an activation function to obtain the original risk score corresponding to the housing provident fund regional transfer request.

[0074] In this implementation, it can be done through the formula Calculate the calibration abstract features; where, This represents the Hadamard product (element-by-element multiplication).

[0075] Then according to the formula Calculate the original risk score corresponding to the housing provident fund regional transfer request, where The Sigmoid activation function compresses the output to the (0, 1) interval, which is the original risk score.

[0076] S104, determine the final risk score corresponding to the housing provident fund regional transfer request based on the original risk score, the user's basic data, the network data, and the device data.

[0077] In one optional embodiment provided in this application, a basic risk score can be determined first based on user basic data, network data, and device data. Then, a final risk score can be determined based on the original risk score and the basic risk score. For example, the original risk score and the basic risk score can be weighted, or the maximum value between the original risk score and the basic risk score can be taken as the final risk score corresponding to the housing provident fund regional transfer request.

[0078] This embodiment can determine the basic risk score using either a basic risk prediction model or a pre-set mapping table. The risk prediction model is a pre-trained neural network model. In this embodiment, user basic data, network data, and device data are first converted into data feature vectors. These feature data vectors are then input into the risk prediction model to obtain the basic risk score. Alternatively, this embodiment can match user basic data, network data, and device data with each data item in the pre-set mapping table, and then sum the matching results and the scores of the corresponding matched items to obtain the basic risk score.

[0079] In another optional embodiment provided in this application, determining the final risk score corresponding to the housing provident fund regional transfer request through the original risk score, the user's basic data, the network data, and the device data includes:

[0080] S1041, Match the user basic data, the network data, and the device data with the hard rules in the preset hard rule table.

[0081] In this embodiment, the hard rules in the pre-set hard rule table are the highest priority and most enforceable rules. Based on deterministic logic, once triggered, they immediately execute a pre-defined, irreversible final action (such as direct rejection or approval). For example, the hard rules in the pre-set hard rule table may include list-based rules, identity-based rules, channel-based rules, etc. If the current device fingerprint exists in the global fraud device blacklist, the housing provident fund regional transfer request is rejected; if the applicant's ID number is already listed as a dishonest person subject to enforcement, the housing provident fund regional transfer request is rejected; if the request originates from a known malicious IP segment or proxy server, the housing provident fund regional transfer request is rejected.

[0082] Specifically, this embodiment can check the hard rules in the preset hard rule table in parallel for user basic data, network data, and device data. These hard rules have the highest priority, and once triggered, subsequent model score adjustments are skipped, and the final risk score is given directly.

[0083] S1042, if any hard rule is successfully matched, the final risk score corresponding to the housing provident fund regional transfer request is determined according to the matched hard rule.

[0084] In this embodiment, if any hard rule is successfully matched, a final risk score is directly generated. For example, if a hard rule determines that a device is on a blacklist or that an IP address in the network data is a malicious node, a veto is directly applied (the veto flag is set to 1), thus setting the final risk score corresponding to the housing provident fund regional transfer request to 1.

[0085] S1043, if no hard rule is successfully matched, then the rule adjustment factor is determined based on the user basic data, the network data, and the device data.

[0086] In this embodiment, determining the rule adjustment factor based on the user basic data, the network data, and the device data includes: matching each piece of data in the user basic data, network data, and device data with the elastic rules in the preset elastic rule table to determine the adjustment contribution value corresponding to each piece of data; and summing all the adjustment contribution values ​​to obtain the rule adjustment factor.

[0087] The pre-set flexible rules in the flexible rule table are used to fine-tune and calibrate the final risk score. Based on probabilistic logic, they do not directly make a final decision upon triggering; instead, they accumulate all adjustment contributions to obtain a rule adjustment factor. Finally, the final risk score corresponding to the housing provident fund regional transfer request is determined by the rule adjustment factor and the original risk score. In this embodiment, based on the adjustment direction of the flexible rules, they are divided into two categories: risk-up rules and risk-down rules.

[0088] The risk adjustment rules, when triggered, add a positive value to the total risk score, indicating an increased risk. For example, if the current operation environment is more than 1000 kilometers away from the city of the last login, the rule adjustment factor Δ = +0.15; if the current device is detected as an emulator or jailbroken, the rule adjustment factor Δ = +0.20; if the request occurs between 2 AM and 5 AM, the rule adjustment factor Δ = +0.10. The risk reduction rules, when triggered, deduct a negative value from the total risk score, indicating a decreased risk. For example, if the user account level has no overdue records, the rule adjustment factor Δ = -0.10; if the user has logged in more than 200 times in the past year, the rule adjustment factor Δ = -0.05 (active users are more trustworthy); if the operation originates from a user's frequently used device (login within the last 90 days), the rule adjustment factor Δ = -0.08.

[0089] In this embodiment, a rule adjustment factor Δ is calculated for the adjustment contribution value S_raw based on a series of flexible rules. Each rule contributes an adjustment contribution value, and its weight is set by expert experience.

[0090] Specifically, it can be calculated using a formula. Adjust the contribution value. Here, Δ is the rule adjustment factor, which is the total adjustment value for S_raw, where S_raw is the value adjusted according to... The calculated adjustment contribution value of the i-th elastic rule. w_i is the preset weight parameter of the i-th elastic rule (can be positive or negative), set by risk control experts according to the importance of the rule. r_i is the triggering result of the i-th elastic rule, 1 for triggering and 0 for not triggering.

[0091] S1044, the final risk score corresponding to the housing provident fund regional transfer request is determined by the rule adjustment factor and the original risk score.

[0092] Specifically, determining the final risk score corresponding to the housing provident fund regional transfer request using the rule adjustment factor and the original risk score includes: adding the rule adjustment factor and the original risk score to obtain a provisional score; and then, according to the formula... Determine the final risk score corresponding to the housing provident fund regional transfer request; wherein, The final risk score, The temporary score is denoted as .

[0093] In this embodiment, the formula can be used. Obtain the temporary score, then apply the formula. Determine the final risk score corresponding to the housing provident fund regional transfer request.

[0094] S105, execute the business operation corresponding to the provident fund regional transfer request according to the ruling instruction corresponding to the final risk score, and store the hash digest information and the ruling instruction in the blockchain.

[0095] In this embodiment, after obtaining the final risk score Subsequently, the corresponding business operation for the housing provident fund regional transfer request can be determined based on the following score range:

[0096] if =0.0, then the ruling is "automatic pass", which is usually caused by the hard rule of "passing with one vote".

[0097] if = 1.0, then the ruling is "rejection", usually caused by a hard rule of "veto" or extremely high overall risk.

[0098] if If θ_low < θ_low, the decision instruction is "auto-pass," where θ_low is a low threshold, such as 0.2. Low-risk requests require no intervention.

[0099] If θ_low <= If the threshold is less than θ_high, the ruling instruction is "enhanced verification," where θ_high is a high threshold, such as 0.75. This indicates a medium-risk request requiring additional verification (such as SMS verification code or facial recognition).

[0100] if If the value is greater than or equal to θ_high, the ruling instruction will be "manual review". High-risk requests will be subject to final decision by risk control experts.

[0101] It should be noted that this embodiment stores the hash digest and adjudication instructions on the blockchain, primarily to establish an immutable, independently verifiable, and timestamped trust anchor. Because the hash digest is unique and irreversible within the original data, once stored on the blockchain, any alteration to the original data will result in a mismatch between its hash value and the hash value recorded on the blockchain, thus immediately exposing the tampering.

[0102] This embodiment provides a method for intelligent processing of housing provident fund business. In response to a user-initiated housing provident fund regional transfer request, the method acquires the user's basic data, business data, operation data, network data, and device data. Then, based on the user's basic data, business data, operation data, network data, and device data, it determines a set of feature vectors and hash digest information. Next, it determines the original risk score corresponding to the housing provident fund regional transfer request based on the feature vector set, and then determines the final risk score corresponding to the housing provident fund regional transfer request using the original risk score, the user's basic data, the network data, and the device data. Finally, it executes the business operation corresponding to the housing provident fund regional transfer request according to the ruling instruction corresponding to the final risk score, and stores the hash digest information and the ruling instruction in the blockchain. Compared to the existing technology of manually verifying and processing housing provident fund regional transfer requests, this application, after receiving a user-initiated housing provident fund regional transfer request, determines the final risk score corresponding to the housing provident fund regional transfer request based on the original risk score, user's basic data, network data, and device data, and then executes the business operation corresponding to the housing provident fund regional transfer request according to the ruling instruction corresponding to the final risk score. Therefore, this application can improve the security of housing provident fund regional transfers.

[0103] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0104] In one embodiment, a system for intelligent processing of housing provident fund business is provided. For example... Figure 2As shown, the detailed descriptions of each functional module of the system for intelligent processing of housing provident fund business are as follows:

[0105] The acquisition module 21 is used to acquire the user's basic user data, business data, operation data, network data, and device data in response to the user's request for a transfer of housing provident fund regions.

[0106] The first determining module 22 is used to determine a set of feature vectors and hash digest information based on the user's basic data, business data, operation data, network data and device data;

[0107] The second determining module 23 is used to determine the original risk score corresponding to the housing provident fund regional transfer request based on the feature vector set;

[0108] The third determining module 24 is used to determine the final risk score corresponding to the housing provident fund regional transfer request through the original risk score, the user's basic data, the network data, and the device data;

[0109] The execution module 25 is used to execute the business operation corresponding to the housing provident fund regional transfer request according to the adjudication instruction corresponding to the final risk score, and store the hash digest information and the adjudication instruction in the blockchain.

[0110] In an optional embodiment, the first determining module 22 is specifically used for:

[0111] The user's basic data, business data, operation data, network data, and device data are preprocessed, and the preprocessing includes at least data cleaning and normalization.

[0112] The preprocessed user basic data, business data, operation data, network data, and device data are each converted into feature vectors; and all feature vectors are combined to obtain the feature vector set.

[0113] The user's basic data, business data, operation data, network data, and device data, which have been preprocessed, are hashed using a cryptographic hash function to obtain hash digest information.

[0114] In an optional embodiment, the second determining module 23 is specifically used for:

[0115] Extract the main feature vector and the meta-context vector from the feature vector set, wherein the meta-context vector is a vector determined based on business data; and the main feature vector is a vector in the feature vector set other than the meta-context vector.

[0116] The dynamic modulation signal is determined based on the meta-context vector, and the abstract features are determined based on the principal feature vector.

[0117] The original risk score corresponding to the housing provident fund regional transfer request is determined by the dynamic modulation signal and the defined abstract features.

[0118] In an optional embodiment, the second determining module 23 is specifically used for:

[0119] The meta-context vector is input into the input layer of the housing provident fund risk model, and the input layer inputs the meta-context vector into the hidden layer of the housing provident fund risk model.

[0120] The activation value of the hidden layer is obtained by performing a linear transformation on the meta-context vector through the hidden layer in the aforementioned housing provident fund risk model and then following a nonlinear activation function.

[0121] The dynamic modulation signal is determined by the activation value of the hidden layer, and the dynamic modulation signal includes a scaling factor and an offset.

[0122] In an optional embodiment, the second determining module 23 is specifically used for:

[0123] The meta-context vector is input into the input layer of the housing provident fund risk model, and the input layer inputs the meta-context vector into the fully connected layer of the housing provident fund risk model.

[0124] The abstract features are obtained by performing multi-layer transformations on the meta-context vector through the fully connected layer in the housing provident fund risk model.

[0125] In an optional embodiment, the second determining module 23 is specifically used for:

[0126] The abstract feature is dynamically calibrated using the dynamic modulation signal to obtain the calibrated abstract feature;

[0127] The original risk score corresponding to the housing provident fund regional transfer request is obtained by calculating the calibration abstract features using an activation function.

[0128] In an optional embodiment, the third determining module 24 is specifically used for:

[0129] The user's basic data, the network data, and the device data are matched with the hard rules in the preset hard rule table.

[0130] If any hard rule is successfully matched, the final risk score corresponding to the housing provident fund regional transfer request is determined according to the matched hard rule;

[0131] If no hard rule is successfully matched, the rule adjustment factor is determined based on the user's basic data, the network data, and the device data.

[0132] The final risk score corresponding to the housing provident fund regional transfer request is determined by the rule adjustment factor and the original risk score.

[0133] In an optional embodiment, the third determining module 24 is specifically used for:

[0134] Each piece of data in the user basic data, network data, and device data is matched with the elastic rules in the preset elastic rule table to determine the adjustment contribution value corresponding to each piece of data.

[0135] The rule adjustment factor is obtained by summing up all the adjustment contribution values.

[0136] In an optional embodiment, the third determining module 24 is specifically used for:

[0137] The temporary score is obtained by adding the rule adjustment factor and the original risk score;

[0138] According to the formula Determine the final risk score corresponding to the housing provident fund regional transfer request;

[0139] in, The final risk score, The temporary score is denoted as .

[0140] It should be noted that the above detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0141] Specific limitations regarding the system used for intelligent processing of housing provident fund business can be found in the limitations of the methods used for intelligent processing of housing provident fund business mentioned above, and will not be repeated here. Each module in the aforementioned equipment can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0142] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above.

[0143] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for intelligent processing of housing provident fund business, characterized in that, The method includes: In response to a user's request to transfer their housing provident fund to another region, the system obtains the user's basic data, business data, operation data, network data, and device data. The feature vector set and hash digest information are determined based on the user's basic data, business data, operation data, network data, and device data. The original risk score corresponding to the housing provident fund regional transfer request is determined based on the set of feature vectors. The final risk score corresponding to the housing provident fund regional transfer request is determined by the original risk score, the user's basic data, the network data, and the device data. The business operation corresponding to the housing provident fund regional transfer request is executed according to the ruling instruction corresponding to the final risk score, and the hash digest information and the ruling instruction are stored in the blockchain.

2. The method according to claim 1, characterized in that, The step of determining the feature vector set and hash digest information based on the user's basic data, business data, operation data, network data, and device data includes: The user's basic data, business data, operation data, network data, and device data are preprocessed, and the preprocessing includes at least data cleaning and normalization. The preprocessed user basic data, business data, operation data, network data, and device data are each converted into feature vectors; and all feature vectors are combined to obtain the feature vector set. The user's basic data, business data, operation data, network data, and device data, which have been preprocessed, are hashed using a cryptographic hash function to obtain hash digest information.

3. The method according to claim 1, characterized in that, Determining the original risk score corresponding to the housing provident fund regional transfer request based on the feature vector set includes: Extract the main feature vector and the meta-context vector from the feature vector set, wherein the meta-context vector is a vector determined based on business data; and the main feature vector is a vector in the feature vector set other than the meta-context vector. The dynamic modulation signal is determined based on the meta-context vector, and the abstract features are determined based on the principal feature vector. The original risk score corresponding to the housing provident fund regional transfer request is determined by the dynamic modulation signal and the defined abstract features.

4. The method according to claim 3, characterized in that, Determining the dynamic modulation signal based on the meta-context vector includes: The meta-context vector is input into the input layer of the housing provident fund risk model, and the input layer inputs the meta-context vector into the hidden layer of the housing provident fund risk model. The activation value of the hidden layer is obtained by performing a linear transformation on the meta-context vector through the hidden layer in the aforementioned housing provident fund risk model and then following a nonlinear activation function. The dynamic modulation signal is determined by the activation value of the hidden layer, and the dynamic modulation signal includes a scaling factor and an offset.

5. The method according to claim 3, characterized in that, The step of determining the abstract features based on the principal feature vector includes: The meta-context vector is input into the input layer of the housing provident fund risk model, and the input layer inputs the meta-context vector into the fully connected layer of the housing provident fund risk model. The abstract features are obtained by performing multi-layer transformations on the meta-context vector through the fully connected layer in the housing provident fund risk model.

6. The method according to claim 3, characterized in that, The step of determining the original risk score corresponding to the housing provident fund regional transfer request through the dynamic modulation signal and the determined abstract features includes: The abstract feature is dynamically calibrated using the dynamic modulation signal to obtain the calibrated abstract feature; The original risk score corresponding to the housing provident fund regional transfer request is obtained by calculating the calibration abstract features using an activation function.

7. The method according to any one of claims 1-6, characterized in that, The process of determining the final risk score corresponding to the housing provident fund regional transfer request using the original risk score, the user's basic data, the network data, and the device data includes: The user's basic data, the network data, and the device data are matched with the hard rules in the preset hard rule table. If any hard rule is successfully matched, the final risk score corresponding to the housing provident fund regional transfer request is determined according to the matched hard rule; If no hard rule is successfully matched, the rule adjustment factor is determined based on the user's basic data, the network data, and the device data. The final risk score corresponding to the housing provident fund regional transfer request is determined by the rule adjustment factor and the original risk score.

8. The method according to claim 7, characterized in that, The step of determining the rule adjustment factor based on the user's basic data, the network data, and the device data includes: Each piece of data in the user basic data, network data, and device data is matched with the elastic rules in the preset elastic rule table to determine the adjustment contribution value corresponding to each piece of data. The rule adjustment factor is obtained by summing up all the adjustment contribution values.

9. The method according to claim 7, characterized in that, The process of determining the final risk score corresponding to the housing provident fund regional transfer request through the rule adjustment factor and the original risk score includes: The temporary score is obtained by adding the rule adjustment factor and the original risk score; According to the formula Determine the final risk score corresponding to the housing provident fund regional transfer request; in, The final risk score, The temporary score is denoted as .

10. A system for intelligent processing of housing provident fund business, characterized in that, The system includes: The acquisition module is used to respond to a user's request for a transfer of housing provident fund regions and acquire the user's basic data, business data, operation data, network data, and device data. The first determining module is used to determine a set of feature vectors and hash digest information based on the user's basic data, business data, operation data, network data, and device data. The second determining module is used to determine the original risk score corresponding to the housing provident fund regional transfer request based on the feature vector set; The third determining module is used to determine the final risk score corresponding to the housing provident fund regional transfer request through the original risk score, the user's basic data, the network data, and the device data; The execution module is used to execute the business operation corresponding to the housing provident fund regional transfer request according to the adjudication instruction corresponding to the final risk score, and store the hash digest information and the adjudication instruction in the blockchain.

Images