Artificial intelligence-based data prediction processing method, apparatus, device, and medium

Abstract

The application belongs to the technical field of artificial intelligence and relates to a data prediction processing method based on artificial intelligence, which comprises the following steps: generating a target inner product based on feature data, a weight coefficient and an inner product; generating a first-order partial derivative and a second-order partial derivative of the target inner product; constructing a diagonal matrix based on the second-order partial derivative, constructing first initial data based on the diagonal matrix and the feature data, and constructing second initial data based on the diagonal matrix, the target inner product and the first-order partial derivative; encrypting the diagonal matrix, the first initial data and the second initial data to obtain encrypted data; calculating an inverse matrix and dividing the encrypted matrix to obtain a matrix; dividing the encrypted matrix to obtain a division matrix based on the encrypted matrix, and generating a first model weight coefficient based on the division matrix; generating a target model weight coefficient based on the first model weight coefficient and a second model weight coefficient; and performing prediction processing on business data based on the target prediction model. The application effectively improves the training efficiency and model effect of the target prediction model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence development and the field of financial technology, and in particular relates to a data prediction processing method and device based on artificial intelligence, a computer device and a storage medium. BACKGROUND

[0002] In the field of data analysis, Generalized Linear Models (GLMs) have attracted much attention due to their flexibility and wide range of applications as an extension of linear regression models. GLMs not only encompass traditional linear regression but also include various variants such as Poisson regression, Tweedie regression, etc. These models can handle various types of data distribution and response variables, making them powerful tools for solving complex problems.

[0003] In the context of car insurance business, the application of GLMs is particularly prominent. For example, Poisson regression models can be used to predict the frequency of users' claims, which is crucial for insurance companies to assess risks and develop premium strategies. Tweedie regression models, on the other hand, are suitable for predicting pure premiums, as they can effectively handle the zero inflation and overdispersion phenomena in insurance claim data, improving the accuracy of predictions. The application of these models not only enhances the operational efficiency of insurance companies but also provides consumers with more fair and reasonable insurance services.

[0004] However, in practical applications, the training process of GLMs faces many challenges. Since label data and feature data are often scattered across different institutions or companies, data integration and transmission become a difficult problem. Moreover, existing training methods for GLMs mostly rely on gradient descent algorithms to update model parameters. Although these methods are theoretically feasible, they have problems such as long training time, low efficiency, slow convergence speed, etc. in practical applications, and result in low accuracy in subsequent prediction processing using GLMs. SUMMARY

[0005] The purpose of the embodiments of the present application is to propose a data prediction processing method and device based on artificial intelligence, a computer device and a storage medium, to solve the technical problem that the existing training method for GLMs mostly relies on gradient descent algorithms to update model parameters, which has problems such as long training time, low efficiency, slow convergence speed, etc. in practical applications, and results in low accuracy in subsequent prediction processing using GLMs.

[0006] To solve the above technical problems, the embodiments of the present application provide a data prediction processing method based on artificial intelligence, which adopts the following technical solutions:

[0007] generate a corresponding target inner product based on preset feature data, weight coefficients, and an inner product received from a non-labeled party;

[0008] generate first-order and second-order partial derivatives corresponding to the target inner product based on a loss function of a preset generalized linear model;

[0009] construct a corresponding diagonal matrix based on the second-order partial derivative, and construct first initial data based on the diagonal matrix and the feature data, and construct second initial data based on the diagonal matrix, the target inner product, and the first-order partial derivative;

[0010] respectively encrypt the diagonal matrix, the first initial data, and the second initial data to obtain corresponding encrypted data;

[0011] perform inverse matrix and partition processing based on a plurality of encrypted matrices received from the non-labeled party to obtain corresponding first, second, third, and fourth matrices; wherein the encrypted matrices are generated based on the encrypted data; the third and fourth matrices are generated based on the first matrix;

[0012] perform matrix partitioning based on a ciphertext matrix received from the non-labeled party to obtain a fifth and sixth matrix, and generate a first model weight coefficient based on the third and fifth matrices; wherein the ciphertext matrix is generated based on the second matrix and the second initial data;

[0013] When a preset iteration termination condition is met, generate a target model weight coefficient based on the first model weight coefficient and a second model weight coefficient received from the non-labeled party; wherein the second model weight coefficient is generated based on the fourth and sixth matrices;

[0014] perform prediction processing on the to-be-processed business data based on a target prediction model corresponding to the target model weight coefficient.

[0015] Further, the step of generating a corresponding target inner product based on preset feature data, weight coefficients, and an inner product received from a non-labeled party, specifically includes:

[0016] calculate the product of the feature data and the weight coefficients to obtain a corresponding specified inner product;

[0017] receive an inner product sent by the non-labeled party;

[0018] calculate the sum of the specified inner product and the inner product;

[0019] take the sum as the target inner product.

[0020] Further, the step of respectively encrypting the diagonal matrix, the first initial data and the second initial data to obtain corresponding encrypted data specifically comprises:

[0021] Obtaining a preset homomorphic encryption strategy;

[0022] Based on the homomorphic encryption strategy, the diagonal matrix, the first initial data and the second initial data are respectively encrypted to obtain corresponding first encrypted data, second encrypted data and third encrypted data;

[0023] Integrating the first encrypted data, the second encrypted data and the third encrypted data to obtain corresponding integrated data;

[0024] The integrated data is used as the encrypted data.

[0025] Further, the step of calculating the inverse matrix and dividing based on the received plurality of encrypted matrices sent by the non-tag party to obtain corresponding first matrix, second matrix, third matrix and fourth matrix specifically comprises:

[0026] Based on the diagonal matrix and the feature data, a first specified matrix is calculated;

[0027] Decrypting the plurality of encrypted matrices to obtain corresponding first decrypted matrix;

[0028] Based on the first specified matrix and the first decrypted matrix, an integrated matrix is generated, and the inverse matrix of the integrated matrix is calculated;

[0029] Based on the preset first feature dimension, the inverse matrix is divided longitudinally to obtain the first matrix and the second matrix;

[0030] Based on the first matrix, the feature data and the second initial data, a second specified matrix is constructed;

[0031] Based on the first feature dimension, the second specified matrix is divided horizontally to obtain the third matrix and the fourth matrix.

[0032] Further, the step of dividing the matrix based on the received ciphertext matrix sent by the non-tag party to obtain the fifth matrix and the sixth matrix specifically comprises:

[0033] Decrypting the ciphertext matrix to obtain a corresponding second decrypted matrix;

[0034] Obtaining a preset second feature dimension;

[0035] The second decryption matrix is divided horizontally based on the second feature dimension, to obtain the fifth matrix and the sixth matrix.

[0036] Further, when the preset iteration termination condition is met, the step of generating the target model weight coefficient based on the first model weight coefficient and the received second model weight coefficient sent by the non-labeled party, specifically includes:

[0037] When the preset iteration termination condition is met, the second model weight coefficient sent by the non-labeled party is received;

[0038] A preset integration strategy is obtained;

[0039] The first model weight coefficient and the second model weight coefficient are integrated based on the integration strategy to obtain a corresponding integration coefficient;

[0040] The integration coefficient is used as the target model weight coefficient.

[0041] Further, before the step of predicting the business data to be processed based on the target prediction model corresponding to the target model weight coefficient, further includes:

[0042] A specified storage strategy corresponding to the target prediction model is determined;

[0043] A specified storage medium corresponding to the specified storage strategy is called;

[0044] The target prediction model is stored based on the specified storage medium.

[0045] To solve the above technical problems, the embodiment of the application also provides a data prediction processing device based on artificial intelligence, which adopts the following technical scheme:

[0046] A first generation module is configured to generate a corresponding target inner product based on preset feature data, weight coefficients, and a received inner product sent by a non-labeled party;

[0047] A second generation module is configured to generate a first-order partial derivative and a second-order partial derivative corresponding to the target inner product based on a loss function of a preset generalized linear model;

[0048] A construction module is configured to construct a corresponding diagonal matrix based on the second-order partial derivative, construct first initial data based on the diagonal matrix and the feature data, and construct second initial data based on the diagonal matrix, the target inner product, and the first-order partial derivative;

[0049] An encryption module is configured to respectively encrypt the diagonal matrix, the first initial data, and the second initial data to obtain corresponding encrypted data.

[0050] The partitioning module is used to calculate the inverse matrix and partition the received multiple encryption matrices sent by the non-label party to obtain corresponding first, second, third, and fourth matrices; wherein the encryption matrix is ​​generated based on the encrypted data; and the third and fourth matrices are generated based on the first matrix.

[0051] The third generation module is used to perform matrix partitioning based on the received ciphertext matrix sent by the non-label party to obtain a fifth matrix and a sixth matrix, and to generate a first model weight coefficient based on the third matrix and the fifth matrix; wherein, the ciphertext matrix is ​​generated based on the second matrix and the second initial data;

[0052] The fourth generation module is used to generate target model weight coefficients based on the first model weight coefficients and the second model weight coefficients received from the non-label party when a preset iteration termination condition is met; wherein the second model weight coefficients are generated based on the fourth matrix and the sixth matrix.

[0053] The processing module is used to perform predictive processing on the business data to be processed based on the target prediction model corresponding to the weight coefficients of the target model.

[0054] To address the aforementioned technical problems, this application also provides a computer device that employs the following technical solution:

[0055] The target inner product is generated based on the preset feature data, weight coefficients, and the inner product received from the non-label party.

[0056] The loss function based on the preset generalized linear model generates the first-order and second-order partial derivatives corresponding to the inner product of the target.

[0057] Construct a corresponding diagonal matrix based on the second-order partial derivative, construct first initial data based on the diagonal matrix and the feature data, and construct second initial data based on the diagonal matrix, the target inner product, and the first-order partial derivative;

[0058] The diagonal matrix, the first initial data, and the second initial data are encrypted respectively to obtain the corresponding encrypted data.

[0059] Based on the received multiple encryption matrices sent by the non-label party, the inverse matrix is ​​calculated and partitioned to obtain the corresponding first matrix, second matrix, third matrix, and fourth matrix; wherein, the encryption matrix is ​​generated based on the encryption data; the third matrix and the fourth matrix are generated based on the first matrix;

[0060] The received ciphertext matrix sent by the non-label party is used to partition the matrix to obtain a fifth matrix and a sixth matrix, and a first model weight coefficient is generated based on the third matrix and the fifth matrix; wherein, the ciphertext matrix is ​​generated based on the second matrix and the second initial data;

[0061] When the preset iteration termination condition is met, the target model weight coefficient is generated based on the first model weight coefficient and the second model weight coefficient received by the non-label party; wherein, the second model weight coefficient is generated based on the fourth matrix and the sixth matrix;

[0062] The business data to be processed is predicted based on the target prediction model corresponding to the weight coefficients of the target model.

[0063] To address the aforementioned technical problems, this application also provides a computer-readable storage medium, employing the technical solution described below:

[0064] The target inner product is generated based on the preset feature data, weight coefficients, and the inner product received from the non-label party.

[0065] The loss function based on the preset generalized linear model generates the first-order and second-order partial derivatives corresponding to the inner product of the target.

[0066] Construct a corresponding diagonal matrix based on the second-order partial derivative, construct first initial data based on the diagonal matrix and the feature data, and construct second initial data based on the diagonal matrix, the target inner product, and the first-order partial derivative;

[0067] The diagonal matrix, the first initial data, and the second initial data are encrypted respectively to obtain the corresponding encrypted data.

[0068] Based on the received multiple encryption matrices sent by the non-label party, the inverse matrix is ​​calculated and partitioned to obtain the corresponding first matrix, second matrix, third matrix, and fourth matrix; wherein, the encryption matrix is ​​generated based on the encryption data; the third matrix and the fourth matrix are generated based on the first matrix;

[0069] The received ciphertext matrix sent by the non-label party is used to partition the matrix to obtain a fifth matrix and a sixth matrix, and a first model weight coefficient is generated based on the third matrix and the fifth matrix; wherein, the ciphertext matrix is ​​generated based on the second matrix and the second initial data;

[0070] When the preset iteration termination condition is met, the target model weight coefficient is generated based on the first model weight coefficient and the second model weight coefficient received by the non-label party; wherein, the second model weight coefficient is generated based on the fourth matrix and the sixth matrix;

[0071] The business data to be processed is predicted based on the target prediction model corresponding to the weight coefficients of the target model.

[0072] Compared with the prior art, the embodiments of this application have the following main advantages:

[0073] This application first generates a target inner product based on preset feature data, weight coefficients, and the inner product received from the non-labeled party; then, it generates first-order and second-order partial derivatives corresponding to the target inner product based on the loss function of a preset generalized linear model; and constructs a corresponding diagonal matrix based on the second-order partial derivative, and constructs first initial data based on the diagonal matrix and the feature data, and constructs second initial data based on the diagonal matrix, the target inner product, and the first-order partial derivative; then, it encrypts the diagonal matrix, the first initial data, and the second initial data respectively to obtain corresponding encrypted data; subsequently, it generates a target inner product based on the received inner product sent by the non-labeled party. The method involves calculating the inverse matrix and partitioning multiple encrypted matrices sent by the non-labeled party to obtain corresponding first, second, third, and fourth matrices. Further, based on the received ciphertext matrix sent by the non-labeled party, a fifth and sixth matrix are obtained through matrix partitioning. First model weight coefficients are generated based on the third and fifth matrices. When a preset iteration termination condition is met, target model weight coefficients are generated based on the first model weight coefficients and the received second model weight coefficients sent by the non-labeled party. Finally, prediction processing is performed on the business data to be processed based on the target prediction model corresponding to the target model weight coefficients. This application utilizes a generalized linear model training method based on longitudinal federated learning to construct the target prediction model. Compared to gradient descent, this method allows for faster model convergence with fewer iterations, is less sensitive to the initial values ​​of model parameters, and is less prone to saddle points, effectively improving the training efficiency of the target prediction model and achieving better model performance after training. Furthermore, regarding security during the computation process, encryption and decryption of intermediate results prevent the leakage or inference of the original data and model parameters of each participant, effectively protecting the entire target prediction model modeling process. This enables the subsequent use of the target prediction model to accurately predict and process business data to obtain corresponding prediction results, thereby improving the prediction effect of business data prediction and processing and increasing the accuracy of the generated prediction results. Attached Figure Description

[0074] To more clearly illustrate the solutions in this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0075] Figure 1 This is an exemplary system architecture diagram to which this application can be applied;

[0076] Figure 2 A flowchart of an embodiment of the artificial intelligence-based data prediction processing method according to this application;

[0077] Figure 3 This is a schematic diagram of a structure of an embodiment of the artificial intelligence-based data prediction processing apparatus according to this application;

[0078] Figure 4 This is a schematic diagram of the structure of one embodiment of the computer device according to this application. Detailed Implementation

[0079] Unless otherwise defined, 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; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings of this application, are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings of this application are used to distinguish different objects, not to describe a particular order.

[0080] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0081] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0082] like Figure 1As shown, system architecture 100 may include terminal device 101, network 102, and server 103. Terminal device 101 may be a laptop 1011, tablet 1012, or mobile phone 1013. Network 102 is used as a medium to provide a communication link between terminal device 101 and server 103. Network 102 may include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.

[0083] Users can use terminal device 101 to interact with server 103 via network 102 to receive or send messages, etc. Various communication client applications can be installed on terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social media platform software, etc.

[0084] Terminal device 101 can be various electronic devices with a display screen and support web browsing. In addition to laptops 1011, tablets 1012, or mobile phones 1013, terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 player (Moving Picture Experts Group Audio Layer IV), a laptop computer, and a desktop computer, etc.

[0085] Server 103 can be a server that provides various services, such as a backend server that provides support for the pages displayed on terminal device 101.

[0086] It should be noted that the data prediction processing method based on artificial intelligence provided in the embodiments of this application is generally executed by a server / terminal device, and correspondingly, the data prediction processing device based on artificial intelligence is generally set in the server / terminal device.

[0087] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0088] Continue to refer to Figure 2This document illustrates a flowchart of an embodiment of the AI-based data prediction processing method according to this application. The order of steps in the flowchart can be changed, and some steps can be omitted, depending on different requirements. The AI-based data prediction processing method provided in this application can be applied to any scenario requiring data prediction, and thus can be applied to products in these scenarios, such as data prediction processing in the financial insurance field. The AI-based data prediction processing method includes the following steps:

[0089] Step S201: Generate the corresponding target inner product based on the preset feature data, weight coefficients and the inner product received from the non-label party.

[0090] In this embodiment, the artificial intelligence-based data prediction processing method runs on an electronic device (e.g., Figure 1 The server / terminal device shown can acquire feature data and weighting coefficients via wired or wireless connection. It should be noted that the aforementioned wireless connection methods may include, but are not limited to, 3G / 4G / 5G connections, WiFi connections, Bluetooth connections, WiMAX connections, Zigbee connections, UWB (ultra-wideband) connections, and other currently known or future-developed wireless connection methods. Specifically, this application involves possessing tag data and feature data X. A and weighting coefficient w A The labeled party. The non-labeled party B possesses feature data X. B and weighting coefficient w B The specific implementation process of generating the corresponding target inner product based on the preset feature data, weight coefficients, and the inner product received from the non-label party will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0091] In this embodiment, for example, in the business scenario of product push in financial insurance, the business tracking data may include transaction data, payment data, business data, etc.

[0092] Step S202: Generate the first-order and second-order partial derivatives corresponding to the inner product of the target based on the loss function of the preset generalized linear model.

[0093] In this embodiment, this application can be applied to predicting the frequency of user accidents or the net premium in auto insurance business scenarios. The aforementioned generalized linear model may include a Poisson regression model for predicting the frequency of user accidents in auto insurance business scenarios, or a Tweedie regression model for predicting net premiums, etc. Specifically, the first-order partial derivative g and the second-order partial derivative h of the loss function L with respect to the target inner product η can be calculated. The loss function L and its first-order and second-order partial derivatives g and h can be generated using different formulas according to the specific generalized linear model.

[0094] Step S203: Construct a corresponding diagonal matrix based on the second-order partial derivative, construct first initial data based on the diagonal matrix and the feature data, and construct second initial data based on the diagonal matrix, the target inner product, and the first-order partial derivative.

[0095] In this embodiment, the aforementioned diagonal matrix includes: W = diag(h), where W is a diagonal matrix with h as its diagonal elements, and h is a second-order partial derivative. The aforementioned first initial data includes WX A Where W is a diagonal matrix, X A These are characteristic data. The second initial data mentioned above includes Wz, where z = η - W. T g,WX A Where η is the target inner product, W is a diagonal matrix, and g is a first-order partial derivative.

[0096] Step S204: Encrypt the diagonal matrix, the first initial data, and the second initial data respectively to obtain the corresponding encrypted data.

[0097] In this embodiment, the specific implementation process of encrypting the diagonal matrix, the first initial data, and the second initial data to obtain the corresponding encrypted data will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here. Additionally, after obtaining the encrypted data Enc(W), Enc(WX)... A After Enc(Wz), the encrypted data will be sent to the non-label party B.

[0098] Step S205: Based on the received multiple encryption matrices sent by the non-label party, calculate the inverse matrix and perform partitioning to obtain the corresponding first matrix, second matrix, third matrix and fourth matrix; wherein, the encryption matrix is ​​generated based on the encryption data; the third matrix and the fourth matrix are generated based on the first matrix.

[0099] In this embodiment, the non-label party B, based on the received encrypted data Enc(W), Enc(Wx)... A),Enc(Wz), combined with internal feature data X B According to the following calculation formula The three encrypted matrices Enc(D) are calculated. AB ), Enc(D BA ), Enc(D BB The first, second, third, and fourth matrices are obtained by calculating the inverse matrix and partitioning the multiple encrypted matrices received from the non-label party. This process will be further detailed in subsequent embodiments and will not be elaborated upon here. Furthermore, the second and fourth matrices are sent to the non-label party B.

[0100] Step S206: Based on the received ciphertext matrix sent by the non-label party, perform matrix partitioning to obtain a fifth matrix and a sixth matrix, and generate a first model weight coefficient based on the third matrix and the fifth matrix; wherein, the ciphertext matrix is ​​generated based on the second matrix and the second initial data.

[0101] In this embodiment, the specific implementation process of dividing the received ciphertext matrix sent by the non-label party into a fifth and a sixth matrix, and generating the first model weight coefficients based on the third and fifth matrices, will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated upon here. Furthermore, the aforementioned first model weight coefficients include: β A =M A1 +M A2 Among them, β A M represents the weight coefficients of the first model. A1 M is the third matrix. A2 This is the fifth matrix.

[0102] Step S207: When the preset iteration termination condition is met, the target model weight coefficient is generated based on the first model weight coefficient and the received second model weight coefficient sent by the non-label party; wherein, the second model weight coefficient is generated based on the fourth matrix and the sixth matrix.

[0103] In this embodiment, the unlabeled party B updates its second model weight coefficient β by calculating and updating the unlabeled party B. B =M B1 +M B2 , where β B M represents the weight coefficients of the second model. B1 M is the fourth matrix. B2This is the sixth matrix. The specific implementation process of generating the target model weight coefficients based on the first model weight coefficients and the received second model weight coefficients sent by the non-label party when the preset iteration termination condition is met will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated upon here. Additionally, the loop steps S201-S207 or the end of the model training process can be selected based on the loss function value or the total number of iterations.

[0104] Step S208: Perform prediction processing on the business data to be processed based on the target prediction model corresponding to the weight coefficients of the target model.

[0105] In this embodiment, after obtaining the final target model weight coefficients, these weight coefficients can be used to represent the trained model; that is, the target prediction model can be represented as y = β. T X+∈, where y is the target variable and X is the feature vector (in this embodiment, it is composed of feature data X). A and feature data X B Composition), β is the weight vector (composed of β) A and β B (Integrated), where ∈ represents the error term. In the financial sector, the aforementioned business data can be user data used to predict user accident frequency in auto insurance, or user data used for pure premium prediction. Specifically, by inputting the aforementioned business data to be processed into the target prediction model, the target prediction model performs prediction processing on the business data, thereby outputting the prediction result corresponding to the business data, such as user accident frequency result or premium prediction result.

[0106] This application first generates a target inner product based on preset feature data, weight coefficients, and the inner product received from the non-labeled party; then, it generates first-order and second-order partial derivatives corresponding to the target inner product based on the loss function of a preset generalized linear model; and constructs a corresponding diagonal matrix based on the second-order partial derivative, and constructs first initial data based on the diagonal matrix and the feature data, and constructs second initial data based on the diagonal matrix, the target inner product, and the first-order partial derivative; then, it encrypts the diagonal matrix, the first initial data, and the second initial data respectively to obtain corresponding encrypted data; subsequently, it generates a target inner product based on the received inner product sent by the non-labeled party. The method involves calculating the inverse matrix and partitioning multiple encrypted matrices sent by the non-labeled party to obtain corresponding first, second, third, and fourth matrices. Further, based on the received ciphertext matrix sent by the non-labeled party, a fifth and sixth matrix are obtained through matrix partitioning. First model weight coefficients are generated based on the third and fifth matrices. When a preset iteration termination condition is met, target model weight coefficients are generated based on the first model weight coefficients and the received second model weight coefficients sent by the non-labeled party. Finally, prediction processing is performed on the business data to be processed based on the target prediction model corresponding to the target model weight coefficients. This application utilizes a generalized linear model training method based on longitudinal federated learning to construct the target prediction model. Compared to gradient descent, this method allows for faster model convergence with fewer iterations, is less sensitive to the initial values ​​of model parameters, and is less prone to saddle points, effectively improving the training efficiency of the target prediction model and achieving better model performance after training. Furthermore, regarding security during the computation process, encryption and decryption of intermediate results prevent the leakage or inference of the original data and model parameters of each participant, effectively protecting the entire target prediction model modeling process. This enables the subsequent use of the target prediction model to accurately predict and process business data to obtain corresponding prediction results, thereby improving the prediction effect of business data prediction and processing and increasing the accuracy of the generated prediction results.

[0107] In some alternative implementations, step S201 includes the following steps:

[0108] Calculate the product of the feature data and the weight coefficient to obtain the corresponding specified inner product.

[0109] In this embodiment, tag provider A possesses tag data y and feature data X. A and weighting coefficient w A The unlabeled party B possesses feature data X. B and weighting coefficient w B The specified inner product η A =X A w A.

[0110] Receive the inner product sent by the non-label party.

[0111] In this embodiment, the non-labeled party B calculates X. B With w B The product of these two products yields the inner product η. B .

[0112] Calculate the sum of the specified inner product and the inner product.

[0113] In this embodiment, η can be calculated. A +η B The sum of η is used to obtain the target inner product.

[0114] The sum is used as the target inner product.

[0115] This application obtains a specified inner product by calculating the product of the feature data and the weight coefficient; then it receives the inner product sent by the non-labeled party; subsequently, it calculates the sum of the specified inner product and the received inner product; and finally, it uses the sum as the target inner product. This application obtains the specified inner product by calculating the product of the feature data and the weight coefficient, and then calculates the sum between the specified inner product and the received inner product sent by the non-labeled party, thereby enabling fast and accurate calculation of the corresponding target inner product and ensuring the accuracy of the obtained target inner product data.

[0116] In some optional implementations of this embodiment, step S204 includes the following steps:

[0117] Obtain the preset homomorphic encryption strategy.

[0118] In this embodiment, the selection of homomorphic encryption strategy is not specifically limited. For example, Paillier encryption, fully homomorphic encryption, and other strategies can be used.

[0119] Based on the homomorphic encryption strategy, the diagonal matrix, the first initial data, and the second initial data are encrypted respectively to obtain the corresponding first encrypted data, second encrypted data, and third encrypted data.

[0120] In this embodiment, the diagonal matrix, the first initial data, and the second initial data can be encrypted according to the additive homomorphic property of the above homomorphic encryption strategy to obtain the corresponding first encrypted data, second encrypted data, and third encrypted data.

[0121] The first encrypted data, the second encrypted data, and the third encrypted data are integrated to obtain the corresponding integrated data.

[0122] In this embodiment, the integrated data includes the first encrypted data, the second encrypted data, and the third encrypted data.

[0123] The integrated data is used as the encrypted data.

[0124] This application obtains a preset homomorphic encryption strategy; then, based on the homomorphic encryption strategy, it encrypts the diagonal matrix, the first initial data, and the second initial data respectively to obtain corresponding first encrypted data, second encrypted data, and third encrypted data; subsequently, it integrates the first encrypted data, the second encrypted data, and the third encrypted data to obtain corresponding integrated data; finally, it uses the integrated data as the encrypted data. This application, by using a homomorphic encryption strategy, can achieve homomorphic encryption processing of the diagonal matrix, the first initial data, and the second initial data respectively, improving the data accuracy and data security of the generated encrypted data. Furthermore, the homomorphic encryption processing can effectively prevent the original data and model parameters of each participant from being leaked or inferred, effectively protecting the security of the model construction process of the target prediction model.

[0125] In some alternative implementations, step S205 includes the following steps:

[0126] A first specified matrix is ​​calculated based on the diagonal matrix and the feature data.

[0127] In this embodiment, it can be based on the formula The first specified matrix is ​​calculated. Where, D AA Let W be the first specified matrix, and X be a diagonal matrix. A For feature data, In this context, T represents the transpose of the feature data.

[0128] The multiple encryption matrices are decrypted to obtain the corresponding first decryption matrix.

[0129] In this embodiment, the encryption matrix includes three encrypted matrices Enc(D) AB ), Enc(D BA ), Enc(D BB Multiple first decryption matrices can be obtained by decrypting the encryption matrix.

[0130] An integrated matrix is ​​generated based on the first specified matrix and the first decryption matrix, and the inverse matrix of the integrated matrix is ​​calculated.

[0131] In this embodiment, H can be obtained by dividing the first specified matrix and the first decryption matrix into blocks to form a large integrated matrix and then obtaining the corresponding inverse matrix.inv .

[0132] The inverse matrix is ​​vertically divided based on a preset first feature dimension to obtain the first matrix and the second matrix.

[0133] In this embodiment, the aforementioned first feature dimension specifically refers to the dimension of the features of each participating party. H can be... inv Based on the dimensions of each participant's characteristics, it is vertically divided into H... A and H B Two matrices, namely the first matrix and the second matrix.

[0134] A second specified matrix is ​​constructed based on the first matrix, the feature data, and the second initial data.

[0135] In this embodiment, it can be calculated This yields the second specified matrix mentioned above. Where H... A Let X be the first matrix. A is the feature data, and Wz is the second initial data.

[0136] The second specified matrix is ​​horizontally divided based on the first feature dimension to obtain the third matrix and the fourth matrix.

[0137] In this embodiment, the second specified matrix can be used. Horizontally divided into M based on the characteristics of each participant. A1 and M B1 Two matrices are obtained, namely the third matrix and the fourth matrix. Subsequently, matrix H... B and M B1 Send to non-label party B.

[0138] This application calculates a first specified matrix based on the diagonal matrix and the feature data; then decrypts the multiple encrypted matrices to obtain a corresponding first decrypted matrix; subsequently, it generates an integrated matrix based on the first specified matrix and the first decrypted matrix, and calculates the inverse matrix of the integrated matrix; next, it vertically divides the inverse matrix based on a preset first feature dimension to obtain the first matrix and the second matrix; further, it constructs a second specified matrix based on the first matrix, the feature data, and the second initial data; finally, it horizontally divides the second specified matrix based on the first feature dimension to obtain the third matrix and the fourth matrix. Thus, by using the first feature dimension to perform vertical and horizontal division processing on the inverse matrix and the second specified matrix constructed from multiple encrypted matrices sent by a non-tagged party, this application can quickly and accurately obtain the corresponding first, second, third, and fourth matrices, improving the processing efficiency and accuracy of matrix division.

[0139] In some alternative implementations, step S206 includes the following steps:

[0140] The ciphertext matrix is ​​decrypted to obtain the corresponding second decryption matrix.

[0141] In this embodiment, the aforementioned ciphertext matrix is ​​specifically as follows: H B Let X be the second matrix. B The features are within the unlabeled square, and Wz is the second initial data.

[0142] Obtain the preset second feature dimension.

[0143] In this embodiment, the second feature dimension specifically refers to the dimension of the features of each participating party.

[0144] The second decryption matrix is ​​horizontally divided based on the second feature dimension to obtain the fifth matrix and the sixth matrix.

[0145] In this embodiment, the second decryption matrix is ​​horizontally divided into M parts according to the feature dimensions of each participant. A2 and M B2 Two matrices are used to obtain the fifth matrix and the sixth matrix.

[0146] This application decrypts the ciphertext matrix to obtain a corresponding second decryption matrix; then, it obtains a preset second feature dimension; subsequently, it horizontally divides the second decryption matrix based on the second feature dimension to obtain the fifth matrix and the sixth matrix. By decrypting the ciphertext matrix to obtain the corresponding second decryption matrix, and then horizontally dividing the second decryption matrix based on the second feature dimension, this application can quickly and accurately obtain the corresponding fifth and sixth matrices, effectively improving the processing efficiency and accuracy of matrix partitioning.

[0147] In some optional implementations of this embodiment, step S207 includes the following steps:

[0148] When the preset iteration termination condition is met, the second model weight coefficient sent by the non-label party is received.

[0149] In this embodiment, the aforementioned iteration termination condition refers to the loss function value falling below a certain numerical threshold or the total number of iterations reaching a preset maximum number of iterations. The numerical threshold and the maximum number of iterations are not specifically limited and can be set according to actual business needs. Furthermore, the weight coefficient of the second model can be obtained by adding the fourth and sixth matrices, i.e., β. B =MB1 +M B2 Among them, β B M represents the weight coefficients of the second model. B1 M is the fourth matrix. B2 This is the sixth matrix.

[0150] Obtain the preset integration strategy.

[0151] In this embodiment, the strategy content of the above integration strategy may include appropriate splicing or weighting based on the individual characteristics and dimensions of each participant's data.

[0152] Based on the integration strategy, the weight coefficients of the first model and the weight coefficients of the second model are integrated to obtain the corresponding integration coefficients.

[0153] In this embodiment, the first model weight coefficient and the second model weight coefficient can be integrated according to the strategy content of the above integration strategy to obtain the corresponding integration coefficient, which is then used as the corresponding target model weight coefficient.

[0154] The integration coefficients are used as the weight coefficients of the target model.

[0155] When the preset iteration termination condition is met, this application receives the second model weight coefficient sent by the non-label party; then obtains the preset integration strategy; then integrates the first model weight coefficient and the second model weight coefficient based on the integration strategy to obtain the corresponding integration coefficient; subsequently, the integration coefficient is used as the target model weight coefficient.

[0156] In some optional implementations of this embodiment, before step S208, the electronic device may further perform the following steps:

[0157] Determine the specified storage strategy corresponding to the target prediction model.

[0158] In this embodiment, the selection of the specified storage strategy is not specifically limited and can be determined according to the actual storage business needs. For example, the specified storage strategy can be any one of blockchain storage, cloud server storage, local database storage, etc.

[0159] Invoke the specified storage medium corresponding to the specified storage policy.

[0160] In this embodiment, the target storage medium may include blockchain, cloud server, local database, etc.

[0161] The target prediction model is stored and processed based on the specified storage medium.

[0162] In this embodiment, after determining the designated storage medium corresponding to the specified storage strategy, the storage processing of the target prediction model is completed by storing the target prediction model in the designated storage medium.

[0163] This application determines a specified storage strategy corresponding to the target prediction model; then calls a specified storage medium corresponding to the specified storage strategy; and subsequently performs storage processing on the target prediction model based on the specified storage medium. By determining a specified storage strategy corresponding to the target prediction model and then calling a specified storage medium corresponding to the specified storage strategy to perform storage processing on the target prediction model, this application can effectively ensure the data security of the target prediction and improve the storage intelligence of the target prediction model.

[0164] In some alternative implementations, the user information obtained is subject to user consent and complies with relevant laws and policies.

[0165] This application proposes a training method for generalized linear models based on the application of iterative reweighted least squares in generalized linear models. It decomposes the formula and performs encrypted computation on the data, and applies it to two-party longitudinal federated learning to protect data privacy.

[0166] Furthermore, any software tools or components not belonging to our company that appear in the embodiments of this application are merely illustrative examples and do not represent actual use.

[0167] 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.

[0168] It should be emphasized that, in order to further ensure the privacy and security of the aforementioned target model weight coefficients, the aforementioned target model weight coefficients can also be stored in a blockchain node.

[0169] The blockchain referred to in this application is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked together using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include an underlying blockchain platform, a platform product service layer, and an application service layer.

[0170] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0171] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0172] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing related hardware with computer-readable instructions. These computer-readable instructions can be stored in a computer-readable storage medium. When executed, the program can include the processes of the embodiments of the above methods. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, or read-only memory (ROM), or random access memory (RAM).

[0173] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0174] Further reference Figure 3 As a response to the above Figure 2 To implement the method shown, this application provides an embodiment of a data prediction processing device based on artificial intelligence, which is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0175] like Figure 3As shown, the artificial intelligence-based data prediction processing device 300 described in this embodiment includes: a first generation module 301, a second generation module 302, a construction module 303, an encryption module 304, a partitioning module 305, a third generation module 306, a fourth generation module 307, and a processing module 308. Wherein:

[0176] The first generation module 301 is used to generate a corresponding target inner product based on preset feature data, weight coefficients and the inner product received from the non-label party;

[0177] The second generation module 302 is used to generate the first-order partial derivative and the second-order partial derivative corresponding to the inner product of the target based on the loss function of the preset generalized linear model.

[0178] The construction module 303 is used to construct a corresponding diagonal matrix based on the second-order partial derivative, construct first initial data based on the diagonal matrix and the feature data, and construct second initial data based on the diagonal matrix, the target inner product, and the first-order partial derivative.

[0179] Encryption module 304 is used to encrypt the diagonal matrix, the first initial data and the second initial data respectively to obtain the corresponding encrypted data;

[0180] The partitioning module 305 is used to calculate the inverse matrix and partition the received multiple encryption matrices sent by the non-label party to obtain corresponding first matrix, second matrix, third matrix and fourth matrix; wherein, the encryption matrix is ​​generated based on the encryption data; the third matrix and the fourth matrix are generated based on the first matrix;

[0181] The third generation module 306 is used to perform matrix partitioning based on the received ciphertext matrix sent by the non-label party to obtain a fifth matrix and a sixth matrix, and to generate a first model weight coefficient based on the third matrix and the fifth matrix; wherein, the ciphertext matrix is ​​generated based on the second matrix and the second initial data;

[0182] The fourth generation module 307 is used to generate target model weight coefficients based on the first model weight coefficients and the second model weight coefficients received from the non-label party when a preset iteration termination condition is met; wherein the second model weight coefficients are generated based on the fourth matrix and the sixth matrix.

[0183] The processing module 308 is used to perform prediction processing on the business data to be processed based on the target prediction model corresponding to the weight coefficients of the target model.

[0184] In some optional implementations of this embodiment, the first generation module 301 includes:

[0185] The first calculation submodule is used to calculate the product of the feature data and the weight coefficient to obtain the corresponding specified inner product;

[0186] The first receiving submodule is used to receive the inner product sent by the non-label party;

[0187] The second calculation submodule is used to calculate the sum of the specified inner product and the inner product;

[0188] The first determining submodule is used to use the sum as the target inner product.

[0189] In some optional implementations of this embodiment, the encryption module 304 includes:

[0190] The first acquisition submodule is used to acquire the preset homomorphic encryption strategy;

[0191] The encryption submodule is used to encrypt the diagonal matrix, the first initial data and the second initial data respectively based on the homomorphic encryption strategy to obtain the corresponding first encrypted data, second encrypted data and third encrypted data.

[0192] The first integration submodule is used to integrate the first encrypted data, the second encrypted data, and the third encrypted data to obtain the corresponding integrated data.

[0193] The second determining submodule is used to use the integrated data as the encrypted data.

[0194] In some optional implementations of this embodiment, the partitioning module 305 includes:

[0195] The third calculation submodule is used to calculate the first specified matrix based on the diagonal matrix and the feature data;

[0196] The first decryption submodule is used to decrypt the plurality of encryption matrices to obtain the corresponding first decryption matrix;

[0197] A generation submodule is used to generate an integrated matrix based on the first specified matrix and the first decryption matrix, and to calculate the inverse matrix of the integrated matrix;

[0198] The first partitioning submodule is used to vertically partition the inverse matrix based on a preset first feature dimension to obtain the first matrix and the second matrix.

[0199] A construction submodule is used to construct a second specified matrix based on the first matrix, the feature data, and the second initial data;

[0200] The second partitioning submodule is used to horizontally partition the second specified matrix based on the first feature dimension to obtain the third matrix and the fourth matrix.

[0201] In some optional implementations of this embodiment, the third generation module 306 includes:

[0202] The second decryption submodule is used to decrypt the ciphertext matrix to obtain the corresponding second decryption matrix;

[0203] The second acquisition submodule is used to acquire a preset second feature dimension;

[0204] The third partitioning submodule is used to horizontally partition the second decryption matrix based on the second feature dimension to obtain the fifth matrix and the sixth matrix.

[0205] In some optional implementations of this embodiment, the fourth generation module 307 includes:

[0206] The second receiving submodule is used to receive the second model weight coefficient sent by the non-label party when the preset iteration termination condition is met.

[0207] The third acquisition submodule is used to acquire the preset integration strategy;

[0208] The second integration submodule is used to integrate the weight coefficients of the first model and the weight coefficients of the second model based on the integration strategy to obtain the corresponding integration coefficients.

[0209] The third determining submodule is used to use the integration coefficients as the weight coefficients of the target model.

[0210] In some optional implementations of this embodiment, the artificial intelligence-based data prediction processing device further includes:

[0211] The determination module is used to determine the specified storage strategy corresponding to the target prediction model;

[0212] The calling module is used to call the specified storage medium corresponding to the specified storage strategy;

[0213] The storage module is used to store the target prediction model based on the specified storage medium.

[0214] To address the aforementioned technical problems, embodiments of this application also provide a computer device. Please refer to [link / reference needed]. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0215] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected via a system bus. It should be noted that only the computer device 4 with components 41-43 is shown in the figure; however, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described here is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes, but is not limited to, microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0216] The computer device can be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer device can interact with the user via a keyboard, mouse, remote control, touchpad, or voice control.

[0217] The memory 41 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 41 may be an internal storage unit of the computer device 4, such as the hard disk or memory of the computer device 4. In other embodiments, the memory 41 may also be an external storage device of the computer device 4, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 4. Of course, the memory 41 may also include both the internal storage unit and its external storage device of the computer device 4. In this embodiment, the memory 41 is typically used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions for data prediction processing methods based on artificial intelligence. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or will be output.

[0218] In some embodiments, the processor 42 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip. The processor 42 is typically used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to execute computer-readable instructions stored in the memory 41 or to process data, for example, to execute computer-readable instructions of the artificial intelligence-based data prediction processing method.

[0219] The network interface 43 may include a wireless network interface or a wired network interface, which is typically used to establish communication connections between the computer device 4 and other electronic devices.

[0220] This application also provides another embodiment, namely, providing a computer-readable storage medium storing computer-readable instructions that can be executed by at least one processor to cause the at least one processor to perform the steps of the artificial intelligence-based data prediction processing method described above.

[0221] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0222] Obviously, the embodiments described above are only some embodiments of this application, not all embodiments. The accompanying drawings show preferred embodiments of this application, but do not limit the patent scope of this application. This application can be implemented in many different forms; rather, the purpose of providing these embodiments is to provide a more thorough and comprehensive understanding of the disclosure of this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this application's specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the scope of patent protection of this application.

Claims

1. A data prediction and processing method based on artificial intelligence, characterized in that, Includes the following steps: The target inner product is generated based on the preset feature data, weight coefficients, and the inner product received from the non-label party. The loss function based on the preset generalized linear model generates the first-order and second-order partial derivatives corresponding to the inner product of the target. Construct a corresponding diagonal matrix based on the second-order partial derivative, construct first initial data based on the diagonal matrix and the feature data, and construct second initial data based on the diagonal matrix, the target inner product, and the first-order partial derivative; The diagonal matrix, the first initial data, and the second initial data are encrypted respectively to obtain the corresponding encrypted data. Based on the received multiple encryption matrices sent by the non-label party, the inverse matrix is ​​calculated and partitioned to obtain the corresponding first matrix, second matrix, third matrix, and fourth matrix; wherein, the encryption matrix is ​​generated based on the encryption data; the third matrix and the fourth matrix are generated based on the first matrix; The received ciphertext matrix sent by the non-label party is used to partition the matrix to obtain a fifth matrix and a sixth matrix, and a first model weight coefficient is generated based on the third matrix and the fifth matrix; wherein, the ciphertext matrix is ​​generated based on the second matrix and the second initial data; When the preset iteration termination condition is met, the target model weight coefficient is generated based on the first model weight coefficient and the second model weight coefficient received by the non-label party; wherein, the second model weight coefficient is generated based on the fourth matrix and the sixth matrix; The business data to be processed is predicted based on the target prediction model corresponding to the weight coefficients of the target model.

2. The data prediction and processing method based on artificial intelligence according to claim 1, characterized in that, The step of generating the corresponding target inner product based on preset feature data, weight coefficients, and the inner product received from the non-label party specifically includes: Calculate the product of the feature data and the weight coefficient to obtain the corresponding specified inner product; Receive the inner product sent by the non-label party; Calculate the sum of the specified inner product and the inner product; The sum is used as the target inner product.

3. The data prediction and processing method based on artificial intelligence according to claim 1, characterized in that, The step of encrypting the diagonal matrix, the first initial data, and the second initial data respectively to obtain the corresponding encrypted data specifically includes: Obtain the preset homomorphic encryption strategy; Based on the homomorphic encryption strategy, the diagonal matrix, the first initial data and the second initial data are encrypted respectively to obtain the corresponding first encrypted data, second encrypted data and third encrypted data; The first encrypted data, the second encrypted data, and the third encrypted data are integrated to obtain the corresponding integrated data. The integrated data is used as the encrypted data.

4. The data prediction and processing method based on artificial intelligence according to claim 1, characterized in that, The step of calculating the inverse matrix and partitioning the multiple encryption matrices received from the non-tagged party to obtain the corresponding first, second, third, and fourth matrices specifically includes: A first specified matrix is ​​calculated based on the diagonal matrix and the feature data; The plurality of encryption matrices are decrypted to obtain the corresponding first decryption matrix; An integrated matrix is ​​generated based on the first specified matrix and the first decryption matrix, and the inverse matrix of the integrated matrix is ​​calculated. The inverse matrix is ​​vertically divided based on a preset first feature dimension to obtain the first matrix and the second matrix; Construct a second specified matrix based on the first matrix, the feature data, and the second initial data; The second specified matrix is ​​horizontally divided based on the first feature dimension to obtain the third matrix and the fourth matrix.

5. The data prediction and processing method based on artificial intelligence according to claim 1, characterized in that, The step of dividing the received ciphertext matrix sent by the non-label party into a fifth and a sixth matrix specifically includes: The ciphertext matrix is ​​decrypted to obtain the corresponding second decryption matrix; Obtain the preset second feature dimension; The second decryption matrix is ​​horizontally divided based on the second feature dimension to obtain the fifth matrix and the sixth matrix.

6. The data prediction and processing method based on artificial intelligence according to claim 1, characterized in that, The step of generating target model weight coefficients based on the first model weight coefficients and the received second model weight coefficients sent by the non-label party when the preset iteration termination condition is met specifically includes: When the preset iteration termination condition is met, the second model weight coefficient sent by the non-label party is received; Obtain the preset integration strategy; Based on the integration strategy, the weight coefficients of the first model and the weight coefficients of the second model are integrated to obtain the corresponding integration coefficients; The integration coefficients are used as the weight coefficients of the target model.

7. The data prediction and processing method based on artificial intelligence according to claim 1, characterized in that, Before the step of performing prediction processing on the business data to be processed based on the target prediction model corresponding to the weight coefficients of the target model, the method further includes: Determine the specified storage strategy corresponding to the target prediction model; Invoke the specified storage medium corresponding to the specified storage policy; The target prediction model is stored and processed based on the specified storage medium.

8. A data prediction and processing device based on artificial intelligence, characterized in that, include: The first generation module is used to generate the corresponding target inner product based on the preset feature data, weight coefficients and the inner product received from the non-label party; The second generation module is used to generate the first-order and second-order partial derivatives of the inner product of the target based on the loss function of the preset generalized linear model. The construction module is used to construct a corresponding diagonal matrix based on the second-order partial derivative, construct first initial data based on the diagonal matrix and the feature data, and construct second initial data based on the diagonal matrix, the target inner product, and the first-order partial derivative. An encryption module is used to encrypt the diagonal matrix, the first initial data, and the second initial data respectively to obtain the corresponding encrypted data. The partitioning module is used to calculate the inverse matrix and partition the received multiple encryption matrices sent by the non-label party to obtain corresponding first, second, third, and fourth matrices; wherein the encryption matrix is ​​generated based on the encrypted data; and the third and fourth matrices are generated based on the first matrix. The third generation module is used to perform matrix partitioning based on the received ciphertext matrix sent by the non-label party to obtain a fifth matrix and a sixth matrix, and to generate a first model weight coefficient based on the third matrix and the fifth matrix; wherein, the ciphertext matrix is ​​generated based on the second matrix and the second initial data; The fourth generation module is used to generate target model weight coefficients based on the first model weight coefficients and the second model weight coefficients received from the non-label party when a preset iteration termination condition is met; wherein the second model weight coefficients are generated based on the fourth matrix and the sixth matrix. The processing module is used to perform predictive processing on the business data to be processed based on the target prediction model corresponding to the weight coefficients of the target model.

9. A computer device, characterized in that, The method includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of the data prediction processing method based on artificial intelligence as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the data prediction processing method based on artificial intelligence as described in any one of claims 1 to 7.

Images