Data processing method and device based on artificial intelligence, computer equipment and medium
By filtering out the pre-fine-tuning datasets with the most similar distribution to the private datasets of local client devices, pre-fine-tuning and federated fine-tuning of the federated learning model, the applicability and prediction accuracy of the federated learning model caused by data heterogeneity among financial institutions are solved. The problem of low performance achieves the effect of improving the applicability of the model and prediction accuracy.
Patent Information
- Application Number
- CN202411458551.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-18
- Publication Date
- 2025-02-14
AI Technical Summary
When multiple financial institutions jointly participate in the optimization process of the federated learning model, data heterogeneity increases the complexity of model training, resulting in low applicability and prediction accuracy of federated learning models in different financial institutions.
By obtaining the pre-acquisitioned pre-fine-tuned data set and the private data sets of each local client device, the preset target distance algorithm is used to calculate the distance between the data sets, and the target pre-fine-tuned data set with the smallest distance is selected to the federated learning model Pre-fine-tuning and federal fine-tuning are performed to obtain target federal learning models suitable for different financial institutions.
This method can accurately select the pre-fine-tuned dataset that is most similar to the private dataset distribution of the local client device, ensuring that the federated learning model is optimized for data with common characteristics during the fine-tuning process, and improving the target federated learning model on different clients The applicability and prediction accuracy in the process improve the overall effectiveness of federated learning.
Smart Images
Figure CN119443309A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of artificial intelligence development technology and financial technology, and in particular to artificial intelligence-based data processing methods, devices, computer equipment and storage media. Background Art
[0002] In the current financial technology field, federated learning, as an emerging distributed machine learning framework, is gradually becoming an important means for multiple financial institutions to share data value and collaboratively optimize model performance. In particular, for key business areas such as credit scoring models, federated learning effectively resolves the contradiction between data privacy protection and model effect improvement by allowing all participants to jointly train models without directly transmitting original data.
[0003] However, in practical applications, when multiple financial institutions jointly participate in the optimization process of federated learning models, they face a series of challenges. Specifically, due to differences in customer groups, business models, and market environments among different financial institutions, the customer data they hold often show significant heterogeneity in multiple dimensions such as credit history, consumer behavior, and financial status. This data heterogeneity not only increases the complexity of model training, but also makes it difficult to effectively update model parameters and evenly improve model performance when fine-tuning large federated learning models, resulting in low applicability and prediction accuracy of the constructed federated learning models in different financial institutions. Summary of the invention
[0004] The purpose of the embodiments of the present application is to propose a data processing method, apparatus, computer equipment and storage medium based on artificial intelligence to solve the technical problem that the applicability and prediction accuracy of the constructed federated learning model in the optimization process of the existing multiple financial institutions participating in the federated learning model are low in different financial institutions.
[0005] In order to solve the above technical problems, the embodiment of the present application provides a data processing method based on artificial intelligence, which adopts the following technical solutions:
[0006] Acquire a pre-collected pre-fine-tuning data set corresponding to the federated learning model; wherein the number of the pre-fine-tuning data sets includes multiple;
[0007] Query the private data sets of each local client device;
[0008] Calculating the distance between the pre-fine-tuning dataset and the private dataset of each of the local client devices based on a preset target distance algorithm;
[0009] Filter out a target pre-fine-tuning dataset with the smallest distance from all the pre-fine-tuning datasets;
[0010] Pre-fine-tune the federated learning model based on the target pre-fine-tune data set to obtain a pre-fine-tune federated learning model;
[0011] Federated fine-tuning the pre-fine-tuned federated learning model based on the private data sets of each of the local client devices to obtain a corresponding target federated learning model;
[0012] The business data to be processed is predicted based on the target federated learning model.
[0013] Furthermore, the step of calculating the distance between the pre-fine-tuning data set and the private data set of each of the local client devices based on a preset target distance algorithm specifically includes:
[0014] Performing vector mapping on the pre-fine-tuning data set to obtain a corresponding first vector;
[0015] Calculate a first mean and a first covariance matrix corresponding to the pre-fine-tuning dataset based on the first vector;
[0016] Receiving a second mean value corresponding to the private data set returned by each of the local client devices; wherein the second mean value is generated after processing the private data set based on a preset differential privacy noise;
[0017] Aggregate all the second mean values to obtain corresponding first aggregate data, and send the first aggregate data to each of the local client devices respectively;
[0018] Receive a second covariance matrix returned by each of the local client devices; wherein the second covariance matrix is generated based on the second mean and the first aggregated data;
[0019] Performing aggregation processing on all the second covariance matrices to obtain corresponding second aggregated data;
[0020] Based on the target distance algorithm, distance calculation is performed on the first mean, the first aggregated data, the first covariance matrix, and the second aggregated data to obtain the distance between the pre-fine-tuning data set and the private data sets of each of the local client devices.
[0021] Furthermore, the step of performing vector mapping on the pre-fine-tuning data set to obtain a corresponding first vector specifically includes:
[0022] Call the preset embedded function;
[0023] Mapping the pre-fine-tuning data set into a preset vector space based on the embedding function to obtain a corresponding mapping vector;
[0024] The mapped vector is used as the first vector.
[0025] Furthermore, the step of aggregating all the second means to obtain corresponding first aggregated data specifically includes:
[0026] Obtaining a preset homomorphic encryption strategy; wherein the number of the homomorphic encryption strategies includes multiple;
[0027] Filter out a target homomorphic encryption strategy from all the homomorphic encryption strategies;
[0028] Aggregating all the second means based on the target homomorphic encryption strategy to obtain corresponding mean data;
[0029] The mean data is used as the first aggregated data.
[0030] Furthermore, the step of performing predictive processing on the business data to be processed based on the target federated learning model specifically includes:
[0031] Obtain the business data to be processed input by the user;
[0032] Inputting the business data into the target federated learning model;
[0033] Performing prediction processing on the business data based on the target federated learning model to obtain corresponding prediction results;
[0034] The prediction result is returned to the user.
[0035] Furthermore, the step of obtaining a pre-collected pre-fine-tuning data set corresponding to the federated learning model specifically includes:
[0036] Collecting initial pre-fine-tuning data sets corresponding to the federated learning model from various preset channels;
[0037] Integrating all the initial pre-fine-tuning data sets to obtain an integrated initial pre-fine-tuning data set;
[0038] The integrated initial pre-fine-tuning dataset is used as the pre-fine-tuning dataset.
[0039] Furthermore, after the step of performing federated fine-tuning on the pre-fine-tuned federated learning model based on the private data sets of each of the local client devices to obtain a corresponding target federated learning model, the method further includes:
[0040] Determining a target storage mode corresponding to the target federated learning model;
[0041] Calling a target storage medium corresponding to the target storage mode;
[0042] The target federated learning model is stored based on the target storage medium.
[0043] In order to solve the above technical problems, the embodiment of the present application further provides a data processing device based on artificial intelligence, which adopts the following technical solution:
[0044] An acquisition module, used to acquire a pre-collected pre-fine-tuning data set corresponding to the federated learning model; wherein the number of the pre-fine-tuning data sets includes multiple;
[0045] A query module, used to query the private data set of each local client device;
[0046] A calculation module, configured to calculate the distance between the pre-fine-tuning data set and the private data set of each of the local client devices based on a preset target distance algorithm;
[0047] A screening module, used to screen out a target pre-fine-tuning dataset with the smallest distance from all the pre-fine-tuning datasets;
[0048] A first fine-tuning module, configured to pre-fine-tune the federated learning model based on the target pre-fine-tuning dataset to obtain a pre-fine-tuned federated learning model;
[0049] A second fine-tuning module, configured to perform a federal fine-tuning on the pre-fine-tuned federated learning model based on a private data set of each of the local client devices to obtain a corresponding target federated learning model;
[0050] A processing module is used to perform predictive processing on the business data to be processed based on the target federated learning model.
[0051] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:
[0052] Acquire a pre-collected pre-fine-tuning data set corresponding to the federated learning model; wherein the number of the pre-fine-tuning data sets includes multiple;
[0053] Query the private data sets of each local client device;
[0054] Calculating the distance between the pre-fine-tuning dataset and the private dataset of each of the local client devices based on a preset target distance algorithm;
[0055] Filter out a target pre-fine-tuning dataset with the smallest distance from all the pre-fine-tuning datasets;
[0056] Pre-fine-tune the federated learning model based on the target pre-fine-tune data set to obtain a pre-fine-tune federated learning model;
[0057] Federated fine-tuning the pre-fine-tuned federated learning model based on the private data sets of each of the local client devices to obtain a corresponding target federated learning model;
[0058] The business data to be processed is predicted based on the target federated learning model.
[0059] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:
[0060] Acquire a pre-collected pre-fine-tuning data set corresponding to the federated learning model; wherein the number of the pre-fine-tuning data sets includes multiple;
[0061] Query the private data sets of each local client device;
[0062] Calculating the distance between the pre-fine-tuning dataset and the private dataset of each of the local client devices based on a preset target distance algorithm;
[0063] Filter out a target pre-fine-tuning dataset with the smallest distance from all the pre-fine-tuning datasets;
[0064] Pre-fine-tune the federated learning model based on the target pre-fine-tune data set to obtain a pre-fine-tune federated learning model;
[0065] Federated fine-tuning the pre-fine-tuned federated learning model based on the private data sets of each of the local client devices to obtain a corresponding target federated learning model;
[0066] The business data to be processed is predicted based on the target federated learning model.
[0067] Compared with the prior art, the embodiments of the present application have the following beneficial effects:
[0068] The present application first obtains a pre-collected pre-fine-tuning data set corresponding to a federated learning model; wherein the number of the pre-fine-tuning data sets includes multiple; then queries the private data set of each local client device; then calculates the distance between the pre-fine-tuning data set and the private data set of each local client device based on a preset target distance algorithm; and screens out a target pre-fine-tuning data set with the smallest distance from all the pre-fine-tuning data sets; subsequently pre-fine-tunes the federated learning model based on the target pre-fine-tuning data set to obtain a pre-fine-tuned federated learning model; further, the pre-fine-tuned federated learning model is federated fine-tuned based on the private data sets of each local client device to obtain a corresponding target federated learning model; and finally, predicts the business data to be processed based on the target federated learning model. The present application obtains a pre-collected pre-fine-tuning data set corresponding to the federated learning model, queries the private data set of each local client device, and then calculates the distance between the pre-fine-tuning data set and the private data set of each local client device by using a preset target distance algorithm, and selects the target pre-fine-tuning data set with the smallest distance from all the pre-fine-tuning data sets, so that the target pre-fine-tuning data set with the most similar distribution to the private data set of the local client device can be accurately selected from all the pre-fine-tuning data sets, thereby ensuring that the federated learning model can be optimized for data with common characteristics during the fine-tuning process, effectively improving the applicability and prediction accuracy of the generated target federated learning model in different clients, thereby improving the overall effect of federated learning. By using the target federated learning model in the future, it is possible to accurately predict and process the business data and obtain the corresponding prediction results, thereby improving the prediction effect of the prediction processing of the business data and the accuracy of the generated prediction results. BRIEF DESCRIPTION OF THE DRAWINGS
[0069] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0070] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;
[0071] Figure 2 A flowchart of an embodiment of an artificial intelligence-based data processing method according to the present application;
[0072] Figure 3 is a schematic structural diagram of an embodiment of an artificial intelligence-based data processing device according to the present application;
[0073] Figure 4 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION
[0074] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.
[0075] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.
[0076] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.
[0077] like Figure 1 As shown, the system architecture 100 may include a terminal device 101, a network 102 and a server 103. The terminal device 101 may be a laptop 1011, a tablet computer 1012 or a mobile phone 1013. The network 102 is used to provide a medium for a communication link between the terminal device 101 and the server 103. The network 102 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.
[0078] The user can use the terminal device 101 to interact with the server 103 through the network 102 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.
[0079] The terminal device 101 can be any electronic device with a display screen and supporting web browsing. In addition to the laptop computer 1011, tablet computer 1012 or mobile phone 1013, the terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 (Moving Picture Experts Group Audio Layer IV), a laptop computer, a desktop computer, etc.
[0080] The server 103 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal device 101 .
[0081] It should be noted that the artificial intelligence-based data processing method provided in the embodiments of the present application is generally executed by a server / terminal device, and accordingly, the artificial intelligence-based data processing device is generally set in the server / terminal device.
[0082] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.
[0083] Continue to refer Figure 2 , showing a flow chart of an embodiment of an artificial intelligence-based data processing method according to the present application. According to different requirements, the order of the steps in the flow chart can be changed, and some steps can be omitted. The artificial intelligence-based data processing method provided in the embodiment of the present application can be applied to any scenario that requires data processing, and the artificial intelligence-based data processing method can be applied to products in these scenarios, for example, predictive processing of business data in the financial insurance field. The artificial intelligence-based data processing method comprises the following steps:
[0084] Step S201, obtaining a pre-collected pre-fine-tuning data set corresponding to a federated learning model; wherein the number of the pre-fine-tuning data sets includes multiple.
[0085] In this embodiment, the data processing method based on artificial intelligence is executed on the electronic device (e.g. Figure 1The server / terminal device shown in the figure) can obtain the pre-fine-tuning data set through a wired connection or a wireless connection. It should be pointed out that the above-mentioned wireless connection method may include but is not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future. The executor of this application is specifically the Chinese and English server. In the business scenario of credit scoring in the financial field, the above-mentioned target federated learning model can specifically be a credit scoring model constructed by federated learning. Among them, the specific implementation process of obtaining the pre-collected pre-fine-tuning data set corresponding to the federated learning model will be further described in detail in the subsequent specific embodiments of this application, and will not be elaborated on here.
[0086] Step S202: query the private data set of each local client device.
[0087] In this embodiment, the private data set of each local client device can be queried by adopting a privacy query method.
[0088] Step S203 : calculating the distance between the pre-fine-tuning dataset and the private dataset of each of the local client devices based on a preset target distance algorithm.
[0089] In this embodiment, the specific implementation process of calculating the distance between the pre-fine-tuning data set and the private data set of each local client device based on the preset target distance algorithm will be further described in detail in subsequent specific embodiments of the present application and will not be elaborated on here.
[0090] Step S204: Filter out a target pre-fine-tuning dataset with the smallest distance from all the pre-fine-tuning datasets.
[0091] In this embodiment, the distances of all pre-fine-tuning data sets may be numerically compared to select a target pre-fine-tuning data set with the smallest distance from all pre-fine-tuning data sets as the optimal choice.
[0092] Step S205 , pre-fine-tune the federated learning model based on the target pre-fine-tune data set to obtain a pre-fine-tune federated learning model.
[0093] In this embodiment, the federated learning model can be pre-fine-tuned by loading the selected target pre-fine-tuning dataset and then using the target pre-fine-tuning dataset to initialize the model parameters, thereby obtaining the pre-fine-tuned federated learning model, and saving the model parameters of the pre-fine-tuned federated learning model.
[0094] Step S206, performing federal fine-tuning on the pre-fine-tuned federated learning model based on the private data sets of each of the local client devices to obtain a corresponding target federated learning model.
[0095] In this embodiment, the global model parameters (i.e., the model parameters of the pre-fine-tuned federated learning model) are initialized in the central server, and the global model parameters are sent to all local client devices. After receiving the global model parameters, each local client device trains the pre-fine-tuned federated learning model using a local private data set to obtain a local update. The local update is sent to the central server. The central server receives the local updates sent by all local clients. The local updates are aggregated and the global model parameters are updated. The above process is then repeated until the predetermined training round is reached or the model performance meets the requirements, and the corresponding target federated learning model is finally obtained.
[0096] Step S207: performing prediction processing on the business data to be processed based on the target federated learning model.
[0097] In this embodiment, the specific implementation process of the above-mentioned predictive processing of the business data to be processed based on the target federated learning model will be further described in detail in the subsequent specific embodiments of this application and will not be elaborated in detail here.
[0098] The present application first obtains a pre-collected pre-fine-tuning data set corresponding to a federated learning model; wherein the number of the pre-fine-tuning data sets includes multiple; then queries the private data set of each local client device; then calculates the distance between the pre-fine-tuning data set and the private data set of each local client device based on a preset target distance algorithm; and screens out a target pre-fine-tuning data set with the smallest distance from all the pre-fine-tuning data sets; subsequently pre-fine-tunes the federated learning model based on the target pre-fine-tuning data set to obtain a pre-fine-tuned federated learning model; further, the pre-fine-tuned federated learning model is federated fine-tuned based on the private data sets of each local client device to obtain a corresponding target federated learning model; and finally, predicts the business data to be processed based on the target federated learning model. The present application obtains a pre-collected pre-fine-tuning data set corresponding to the federated learning model, queries the private data set of each local client device, and then calculates the distance between the pre-fine-tuning data set and the private data set of each local client device by using a preset target distance algorithm, and selects the target pre-fine-tuning data set with the smallest distance from all the pre-fine-tuning data sets, so that the target pre-fine-tuning data set with the most similar distribution to the private data set of the local client device can be accurately selected from all the pre-fine-tuning data sets, thereby ensuring that the federated learning model can be optimized for data with common characteristics during the fine-tuning process, effectively improving the applicability and prediction accuracy of the generated target federated learning model in different clients, thereby improving the overall effect of federated learning. By using the target federated learning model in the future, it is possible to accurately predict and process the business data and obtain the corresponding prediction results, thereby improving the prediction effect of the prediction processing of the business data and the accuracy of the generated prediction results.
[0099] In some optional implementations, step S203 includes the following steps:
[0100] Perform vector mapping on the pre-fine-tuning data set to obtain a corresponding first vector.
[0101] In this embodiment, the embedding function f can be used to perform vector mapping on the pre-fine-tuning data set, thereby obtaining the corresponding first vector. Specifically, assuming that X1 represents the pre-fine-tuning data set, X2 represents the private data set on the local client device, the embedding function f represents the use of the embedding layer to map the data set to the space of d-dimensional vectors, Tr represents the calculation of the trace of the matrix, then E1=f(X1), and E1 is the first vector.
[0102] A first mean and a first covariance matrix corresponding to the pre-fine-tuning dataset are calculated based on the first vector.
[0103] In this embodiment, the first mean m1 and the first covariance matrix C1 of the pre-fine-tuning data set X1 are calculated by using the following formulas, ie, m1=mean(E1), C1=Cov(E1).
[0104] Receive a second mean value corresponding to the private data set returned by each of the local client devices, wherein the second mean value is generated by processing the private data set based on a preset differential privacy noise.
[0105] In this embodiment, the embedding function may be sent to each of the local client devices. Each local client device calculates the second initial mean E2 corresponding to the private data set X2 by using the embedding function f, adds differential privacy noise to E2, and trims E2 to obtain the final second mean χ(E2), which is then sent to the central server;
[0106] Aggregate all the second mean values to obtain corresponding first aggregate data, and send the first aggregate data to each of the local client devices respectively.
[0107] In this embodiment, all received χ(E2) can be securely aggregated to obtain first aggregate data pmean2, and sent to each local client device. The specific implementation process of aggregating all the second means to obtain the corresponding first aggregate data will be further described in detail in the subsequent specific embodiments of this application, and will not be elaborated on here.
[0108] Receive a second covariance matrix returned by each of the local client devices, wherein the second covariance matrix is generated based on the second mean and the first aggregated data.
[0109] In this embodiment, each local client device calculates its second covariance matrix according to χ(E2) minus pmean2 and the differential privacy budget, that is, C2=Cov(χ(E2)-pmean2), and sends it to the central server.
[0110] Aggregation processing is performed on all the second covariance matrices to obtain corresponding second aggregated data.
[0111] In this embodiment, the second aggregated data pcov2 can be obtained by securely aggregating all received second covariance matrices C2.
[0112] Based on the target distance algorithm, distance calculation is performed on the first mean, the first aggregated data, the first covariance matrix, and the second aggregated data to obtain the distance between the pre-fine-tuning data set and the private data sets of each of the local client devices.
[0113] In this embodiment, the above-mentioned target distance algorithm is specifically a method for calculating the federated private data distance (Federated Private Frechet Distance, FreD). It is used to evaluate the similarity between the pre-fine-tuning dataset and the federated dataset. FreD uses distributed differential privacy mean and covariance estimators to calculate and compare the Frechet distance between embedded data generated by large language models. This calculation method can not only ensure privacy protection, but also provide accurate pre-fine-tuning dataset selection, so that the federated learning model can better match the private datasets of each client. For example, when multiple financial institutions jointly optimize the credit scoring model, since the customer data of different financial institutions may have significant differences in credit history, consumption behavior, etc., it is difficult to achieve effective update of model parameters and balanced improvement of model performance when fine-tuning the federated learning model. Based on the use of the target distance algorithm, the present application screens out target pre-fine-tuning datasets with similar distributions while protecting data privacy, ensuring that the federated learning model can be optimized for data with common characteristics during the fine-tuning process, thereby improving the applicability and prediction accuracy of the target federated learning model in different financial institutions.
[0114] Specifically, the target distance algorithm includes: Among them, Fred is the federal Fréchet distance, m1 is the first mean, pmean2 is the first aggregate data, Tr represents the trace of the calculation matrix, C1 is the first covariance matrix, and pcov2 is the second aggregate data.
[0115] The present application obtains a corresponding first vector by performing vector mapping on the pre-fine-tuning data set; then calculates a first mean and a first covariance matrix corresponding to the pre-fine-tuning data set based on the first vector; then receives a second mean corresponding to the private data set returned by each of the local client devices; and aggregates all the second means to obtain corresponding first aggregated data, and sends the first aggregated data to each of the local client devices; subsequently receives a second covariance matrix returned by each of the local client devices; wherein the second covariance matrix is generated based on the second mean and the first aggregated data; and aggregates all the second covariance matrices to obtain corresponding second aggregated data; finally, based on the target distance algorithm, performs distance calculation on the first mean, the first aggregated data, the first covariance matrix and the second aggregated data to obtain the distance between the pre-fine-tuning data set and the private data sets of each of the local client devices. By using the target distance algorithm as the information metric for selecting the fine-tuning dataset, the present application can quickly and accurately calculate the distance between the pre-fine-tuning dataset and the private datasets of each of the local client devices, thereby ensuring the accuracy of the obtained distance data, and facilitating the subsequent selection of the target pre-fine-tuning dataset that is most suitable for the local client device in a privacy-preserving manner, thereby helping to improve the performance and effect of the federated learning model.
[0116] In some optional implementations of this embodiment, performing vector mapping on the pre-fine-tuning data set to obtain a corresponding first vector includes the following steps:
[0117] Calls a preset embedded function.
[0118] In this embodiment, the above embedding function may specifically adopt an embedding layer of a pre-trained neural network model (such as BERT, GPT, etc.).
[0119] The pre-fine-tuning data set is mapped into a preset vector space based on the embedding function to obtain a corresponding mapping vector.
[0120] In this embodiment, the pre-fine-tuning data set may be mapped to a d-dimensional vector space by using the above embedding function, thereby obtaining a corresponding mapping vector.
[0121] The mapped vector is used as the first vector.
[0122] The present application calls a preset embedding function; then maps the pre-fine-tuning data set to a preset vector space based on the embedding function to obtain a corresponding mapping vector; and subsequently uses the mapping vector as the first vector. The present application maps the pre-fine-tuning data set to a preset vector space based on the use of an embedding function, thereby achieving fast and accurate completion of the vector mapping process for the pre-fine-tuning data set, thereby improving the generation efficiency of the generated first vector.
[0123] In some optional implementations, the aggregating all the second means to obtain corresponding first aggregated data includes the following steps:
[0124] Obtaining a preset homomorphic encryption strategy; wherein the number of the homomorphic encryption strategies includes multiple;
[0125] In this embodiment, the above homomorphic encryption strategy may at least include Paillier encryption, fully homomorphic encryption and other strategies.
[0126] A target homomorphic encryption strategy is selected from all the homomorphic encryption strategies.
[0127] In this embodiment, there is no specific limitation on the method for selecting the target homomorphic encryption strategy, and one strategy can be randomly selected from all the homomorphic encryption strategies to serve as the target homomorphic encryption strategy.
[0128] Based on the target homomorphic encryption strategy, all the second means are aggregated to obtain corresponding mean data.
[0129] In this embodiment, all second means can be aggregated according to the additive homomorphic property of the above-mentioned target homomorphic encryption strategy to obtain the encrypted global mean, that is, the above-mentioned mean data.
[0130] The mean data is used as the first aggregated data.
[0131] The present application obtains a preset homomorphic encryption strategy; wherein the number of the homomorphic encryption strategies includes multiple; then a target homomorphic encryption strategy is screened out from all the homomorphic encryption strategies; then all the second means are aggregated based on the target homomorphic encryption strategy to obtain corresponding mean data; and the mean data is subsequently used as the first aggregate data. The present application screens out a target homomorphic encryption strategy from multiple homomorphic encryption strategies, and then uses the target homomorphic encryption strategy to aggregate all the second means, so that the corresponding first aggregate data can be generated quickly and accurately, thereby improving the data accuracy and data security of the first aggregate data.
[0132] In some optional implementations, step S207 includes the following steps:
[0133] Get the business data to be processed entered by the user.
[0134] In this embodiment, in the business scenario of credit scoring in the financial field, the above-mentioned business data may be customer data corresponding to the user, for example, may include credit history, consumption behavior and other data.
[0135] The business data is input into the target federated learning model.
[0136] In this embodiment, the target federated learning model may specifically be a credit scoring model constructed through federated learning.
[0137] The business data is predicted and processed based on the target federated learning model to obtain corresponding prediction results.
[0138] In this embodiment, the business data is predicted by using the target federated learning model, thereby outputting a prediction result corresponding to the business data, namely, credit score data.
[0139] The prediction result is returned to the user.
[0140] In this embodiment, the prediction result can be returned to the user by sending a text message, sending an email, displaying an interface, etc.
[0141] The present application obtains the business data to be processed input by the user; then inputs the business data into the target federated learning model; then predicts and processes the business data based on the target federated learning model to obtain corresponding prediction results; and subsequently returns the prediction results to the user. The present application obtains the business data to be processed input by the user, and then inputs the business data into the target federated learning model, so as to accurately predict and process the business data based on the use of the target federated learning model to obtain corresponding prediction results, thereby improving the prediction effect of the prediction processing of the business data and improving the accuracy of the generated prediction results.
[0142] In some optional implementations of this embodiment, step S201 includes the following steps:
[0143] An initial pre-fine-tuning dataset corresponding to the federated learning model is collected from each preset channel.
[0144] In this embodiment, the preset channels may include at least public data sets, historical data or other sources. A large number of possible pre-fine-tuning data sets matching the federated learning model may be collected by collecting a large number of possible pre-fine-tuning data sets from public data sets, historical data or other sources.
[0145] All the initial pre-fine-tuning data sets are integrated to obtain an integrated initial pre-fine-tuning data set.
[0146] In this embodiment, corresponding integrated data, ie, the integrated initial pre-fine-tuning data set, is obtained by integrating all the initial pre-fine-tuning data sets.
[0147] The integrated initial pre-fine-tuning dataset is used as the pre-fine-tuning dataset.
[0148] The present application collects the initial pre-fine-tuning data set corresponding to the federated learning model from various preset channels respectively; then integrates all the initial pre-fine-tuning data sets to obtain an integrated initial pre-fine-tuning data set; and subsequently uses the integrated initial pre-fine-tuning data set as the pre-fine-tuning data set. The present application collects the initial pre-fine-tuning data set corresponding to the federated learning model from various preset channels respectively, and then integrates all the initial pre-fine-tuning data sets to obtain a pre-fine-tuning data set corresponding to the federated learning model in a fast and intelligent manner, thereby improving the construction efficiency of the pre-fine-tuning data set and ensuring the data diversity and richness of the obtained pre-fine-tuning data set.
[0149] In some optional implementations of this embodiment, after step S206, the electronic device may further perform the following steps:
[0150] A target storage mode corresponding to the target federated learning model is determined.
[0151] In this embodiment, there is no specific limitation on the selection of the above-mentioned target storage method, which can be determined according to the actual storage business needs. Exemplarily, the target storage method can adopt any one of blockchain storage, network disk storage, cloud server storage, local database storage, etc. Among them, blockchain storage is preferably used as the above-mentioned target storage method. By using the storage method of blockchain storage to store and manage the target federated learning model, the security and non-tamperability of the target federated learning model can be effectively guaranteed.
[0152] The target storage medium corresponding to the target storage mode is called.
[0153] In this embodiment, the above-mentioned target storage medium may include blockchain, network disk, cloud server, local database, etc.
[0154] The target federated learning model is stored based on the target storage medium.
[0155] In this embodiment, after determining the target storage medium corresponding to the target storage method, the target federated learning model is further stored in the above-mentioned target storage medium to complete the storage processing of the target federated learning model.
[0156] The present application determines the target storage method corresponding to the target federated learning model; then calls the target storage medium corresponding to the target storage method; and subsequently stores and processes the target federated learning model based on the target storage medium. The present application obtains the target storage method that matches the target federated learning model, and then calls the target storage medium corresponding to the target storage method to store and process the target federated learning model, which can effectively ensure the data security and non-tamperability of the target federated learning model and improve the storage intelligence of the target federated learning model.
[0157] In some optional implementations, the user information obtained is subject to the user's consent and complies with relevant laws and policies.
[0158] In addition, the non-Company software tools or components that appear in the embodiments of the present application are merely examples and do not represent actual use.
[0159] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean 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 embodiment of the present invention.
[0160] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned target federated learning model, the above-mentioned target federated learning model can also be stored in a node of a blockchain.
[0161] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.
[0162] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.
[0163] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.
[0164] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).
[0165] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.
[0166] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of a data processing device based on artificial intelligence. Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.
[0167] like Figure 3 As shown, the artificial intelligence-based data processing device 300 described in this embodiment includes: an acquisition module 301, a query module 302, a calculation module 303, a screening module 304, a first fine-tuning module 305, a second fine-tuning module 306 and a processing module 307. Among them:
[0168] An acquisition module 301 is used to acquire a pre-collected pre-fine-tuning data set corresponding to a federated learning model; wherein the number of the pre-fine-tuning data sets includes multiple;
[0169] A query module 302, for querying a private data set of each local client device;
[0170] A calculation module 303, configured to calculate the distance between the pre-fine-tuning data set and the private data set of each of the local client devices based on a preset target distance algorithm;
[0171] A screening module 304 is used to screen out a target pre-fine-tuning data set with the smallest distance from all the pre-fine-tuning data sets;
[0172] A first fine-tuning module 305, configured to pre-fine-tune the federated learning model based on the target pre-fine-tuning dataset to obtain a pre-fine-tuned federated learning model;
[0173] A second fine-tuning module 306 is used to perform federated fine-tuning on the pre-fine-tuned federated learning model based on the private data sets of each of the local client devices to obtain a corresponding target federated learning model;
[0174] The processing module 307 is used to perform prediction processing on the business data to be processed based on the target federated learning model.
[0175] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based data processing method in the aforementioned implementation mode, and will not be repeated here.
[0176] In some optional implementations of this embodiment, the calculation module 303 includes:
[0177] A mapping submodule, used to perform vector mapping on the pre-fine-tuning data set to obtain a corresponding first vector;
[0178] A first calculation submodule, configured to calculate a first mean and a first covariance matrix corresponding to the pre-fine-tuning data set based on the first vector;
[0179] A first receiving submodule, configured to receive a second mean value corresponding to a private data set returned by each of the local client devices; wherein the second mean value is generated by processing the private data set based on a preset differential privacy noise;
[0180] A first processing submodule, configured to aggregate all the second mean values to obtain corresponding first aggregate data, and send the first aggregate data to each of the local client devices respectively;
[0181] A second receiving submodule, configured to receive a second covariance matrix returned by each of the local client devices; wherein the second covariance matrix is generated based on the second mean and the first aggregated data;
[0182] A second processing submodule, configured to perform aggregation processing on all the second covariance matrices to obtain corresponding second aggregated data;
[0183] The second calculation submodule is used to perform distance calculation on the first mean, the first aggregated data, the first covariance matrix and the second aggregated data based on the target distance algorithm to obtain the distance between the pre-fine-tuning data set and the private data set of each of the local client devices.
[0184] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based data processing method in the aforementioned implementation mode, and will not be repeated here.
[0185] In some optional implementations of this embodiment, the mapping submodule includes:
[0186] A calling unit, used for calling a preset embedded function;
[0187] A mapping unit, configured to map the pre-fine-tuning data set into a preset vector space based on the embedding function to obtain a corresponding mapping vector;
[0188] A first determining unit is configured to use the mapping vector as the first vector.
[0189] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based data processing method in the aforementioned implementation mode, and will not be repeated here.
[0190] In some optional implementations of this embodiment, the first processing submodule includes:
[0191] An acquisition unit, used to acquire a preset homomorphic encryption strategy; wherein the number of the homomorphic encryption strategies includes multiple;
[0192] A screening unit, used to screen out a target homomorphic encryption strategy from all the homomorphic encryption strategies;
[0193] an aggregation unit, configured to aggregate all the second means based on the target homomorphic encryption strategy to obtain corresponding mean data;
[0194] The second determining unit is configured to use the mean data as the first aggregated data.
[0195] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based data processing method in the aforementioned implementation mode, and will not be repeated here.
[0196] In some optional implementations of this embodiment, the processing module 307 includes:
[0197] The acquisition submodule is used to obtain the business data to be processed input by the user;
[0198] An input submodule, used to input the business data into the target federated learning model;
[0199] A prediction submodule, used to perform prediction processing on the business data based on the target federated learning model to obtain corresponding prediction results;
[0200] The return submodule is used to return the prediction result to the user.
[0201] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based data processing method in the aforementioned implementation mode, and will not be repeated here.
[0202] In some optional implementations of this embodiment, the acquisition module 301 includes:
[0203] A collection submodule, used to collect initial pre-fine-tuning data sets corresponding to the federated learning model from various preset channels;
[0204] An integration submodule, used for integrating all the initial pre-fine-tuning data sets to obtain an integrated initial pre-fine-tuning data set;
[0205] A determination submodule is configured to use the integrated initial pre-fine-tuning dataset as the pre-fine-tuning dataset.
[0206] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based data processing method in the aforementioned implementation mode, and will not be repeated here.
[0207] In some optional implementations of this embodiment, the artificial intelligence-based data processing device further includes:
[0208] A determination module, used to determine a target storage mode corresponding to the target federated learning model;
[0209] A calling module, used for calling a target storage medium corresponding to the target storage mode;
[0210] A storage module is used to store the target federated learning model based on the target storage medium.
[0211] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based data processing method in the aforementioned implementation mode, and will not be repeated here.
[0212] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.
[0213] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 with components 41-43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform 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 (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.
[0214] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.
[0215] The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), 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, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk equipped on the computer device 4, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions of data processing methods based on artificial intelligence, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.
[0216] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run computer-readable instructions stored in the memory 41 or process data, such as computer-readable instructions for running the artificial intelligence-based data processing method.
[0217] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.
[0218] Compared with the prior art, the embodiments of the present application have the following beneficial effects:
[0219] In the embodiment of the present application, by obtaining a pre-collected pre-fine-tuning data set corresponding to the federated learning model, and querying the private data set of each local client device, and then calculating the distance between the pre-fine-tuning data set and the private data set of each local client device by using a preset target distance algorithm, and filtering out the target pre-fine-tuning data set with the smallest distance from all the pre-fine-tuning data sets, it is possible to accurately select the target pre-fine-tuning data set that is most similar to the distribution of the private data set of the local client device from all the pre-fine-tuning data sets, thereby ensuring that the federated learning model can be optimized for data with common characteristics during the fine-tuning process, effectively improving the applicability and prediction accuracy of the generated target federated learning model in different clients, thereby improving the overall effect of federated learning. By subsequently using the target federated learning model, it is possible to accurately predict and process the business data and obtain the corresponding prediction results, thereby improving the prediction effect of the prediction processing of the business data and the accuracy of the generated prediction results.
[0220] The present application also provides another embodiment, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the artificial intelligence-based data processing method as described above.
[0221] Compared with the prior art, the embodiments of the present application have the following beneficial effects:
[0222] In the embodiment of the present application, by obtaining a pre-collected pre-fine-tuning data set corresponding to the federated learning model, and querying the private data set of each local client device, and then calculating the distance between the pre-fine-tuning data set and the private data set of each local client device by using a preset target distance algorithm, and filtering out the target pre-fine-tuning data set with the smallest distance from all the pre-fine-tuning data sets, it is possible to accurately select the target pre-fine-tuning data set that is most similar to the distribution of the private data set of the local client device from all the pre-fine-tuning data sets, thereby ensuring that the federated learning model can be optimized for data with common characteristics during the fine-tuning process, effectively improving the applicability and prediction accuracy of the generated target federated learning model in different clients, thereby improving the overall effect of federated learning. By subsequently using the target federated learning model, it is possible to accurately predict and process the business data and obtain the corresponding prediction results, thereby improving the prediction effect of the prediction processing of the business data and the accuracy of the generated prediction results.
[0223] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.
[0224] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.
Claims
1. A data processing method based on artificial intelligence, characterized in that: The steps include: Acquire a pre-collected pre-fine-tuning data set corresponding to the federated learning model; wherein the number of the pre-fine-tuning data sets includes multiple; Query the private data sets of each local client device; Calculating the distance between the pre-fine-tuning dataset and the private dataset of each of the local client devices based on a preset target distance algorithm; Filter out a target pre-fine-tuning dataset with the smallest distance from all the pre-fine-tuning datasets; Pre-fine-tune the federated learning model based on the target pre-fine-tune data set to obtain a pre-fine-tune federated learning model; Federated fine-tuning the pre-fine-tuned federated learning model based on the private data sets of each of the local client devices to obtain a corresponding target federated learning model; The business data to be processed is predicted based on the target federated learning model.
2. The data processing method based on artificial intelligence according to claim 1, characterized in that: The step of calculating the distance between the pre-fine-tuning data set and the private data sets of each of the local client devices based on a preset target distance algorithm specifically includes: Performing vector mapping on the pre-fine-tuning data set to obtain a corresponding first vector; Calculate a first mean and a first covariance matrix corresponding to the pre-fine-tuning dataset based on the first vector; Receiving a second mean value corresponding to the private data set returned by each of the local client devices; wherein the second mean value is generated after processing the private data set based on a preset differential privacy noise; Aggregate all the second mean values to obtain corresponding first aggregate data, and send the first aggregate data to each of the local client devices respectively; Receive a second covariance matrix returned by each of the local client devices; wherein the second covariance matrix is generated based on the second mean and the first aggregated data; Performing aggregation processing on all the second covariance matrices to obtain corresponding second aggregated data; Based on the target distance algorithm, distance calculation is performed on the first mean, the first aggregated data, the first covariance matrix, and the second aggregated data to obtain the distance between the pre-fine-tuning data set and the private data sets of each of the local client devices.
3. The data processing method based on artificial intelligence according to claim 2, characterized in that: The step of performing vector mapping on the pre-fine-tuning data set to obtain a corresponding first vector specifically includes: Call the preset embedded function; Mapping the pre-fine-tuning data set into a preset vector space based on the embedding function to obtain a corresponding mapping vector; The mapped vector is used as the first vector.
4. The data processing method based on artificial intelligence according to claim 2, characterized in that: The step of aggregating all the second means to obtain corresponding first aggregated data specifically includes: Obtaining a preset homomorphic encryption strategy; wherein the number of the homomorphic encryption strategies includes multiple; Filter out a target homomorphic encryption strategy from all the homomorphic encryption strategies; Aggregating all the second means based on the target homomorphic encryption strategy to obtain corresponding mean data; The mean data is used as the first aggregated data.
5. The data processing method based on artificial intelligence according to claim 1, characterized in that: The step of performing prediction processing on the business data to be processed based on the target federated learning model specifically includes: Obtain the business data to be processed input by the user; Inputting the business data into the target federated learning model; Performing prediction processing on the business data based on the target federated learning model to obtain corresponding prediction results; The prediction result is returned to the user.
6. The data processing method based on artificial intelligence according to claim 1, characterized in that: The step of obtaining a pre-collected pre-fine-tuning data set corresponding to the federated learning model specifically includes: Collecting initial pre-fine-tuning data sets corresponding to the federated learning model from various preset channels; Integrating all the initial pre-fine-tuning data sets to obtain an integrated initial pre-fine-tuning data set; The integrated initial pre-fine-tuning dataset is used as the pre-fine-tuning dataset.
7. The data processing method based on artificial intelligence according to claim 1, characterized in that: After the step of performing federated fine-tuning on the pre-fine-tuned federated learning model based on the private data sets of each of the local client devices to obtain a corresponding target federated learning model, the method further includes: Determining a target storage mode corresponding to the target federated learning model; Calling a target storage medium corresponding to the target storage mode; The target federated learning model is stored based on the target storage medium.
8. A data processing device based on artificial intelligence, characterized in that: include: An acquisition module, used to acquire a pre-collected pre-fine-tuning data set corresponding to the federated learning model; wherein the number of the pre-fine-tuning data sets includes multiple; A query module, used to query the private data set of each local client device; A calculation module, configured to calculate the distance between the pre-fine-tuning data set and the private data set of each of the local client devices based on a preset target distance algorithm; A screening module, used to screen out a target pre-fine-tuning dataset with the smallest distance from all the pre-fine-tuning datasets; A first fine-tuning module, configured to pre-fine-tune the federated learning model based on the target pre-fine-tuning dataset to obtain a pre-fine-tuned federated learning model; A second fine-tuning module, configured to perform a federal fine-tuning on the pre-fine-tuned federated learning model based on a private data set of each of the local client devices to obtain a corresponding target federated learning model; A processing module is used to perform predictive processing on the business data to be processed based on the target federated learning model.
9. A computer device, characterized in that: It comprises a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the artificial intelligence-based data processing method as described in any one of claims 1 to 7 are implemented.
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 artificial intelligence-based data processing method according to any one of claims 1 to 7.