A financial product recommendation method, device and equipment and storage medium

Abstract

The application discloses a financial product recommendation method and device, equipment and storage medium, applied to the field of block chain, comprising: each distributed host respectively carries out initial training on the initial model according to the local data set to obtain the local initial parameter, and sends to the corresponding connected block node, wherein the initial model comprises a multi-feature input model;Each block node calculates the uploaded parameter according to the received initial parameter and sends it to the corresponding connected distributed host;Each distributed host corrects and trains the initial model after local initial training according to the uploaded parameter to obtain the local product recommendation model;According to the searched target product recommendation model, the recommended financial product suitable for the user is predicted. Through the architecture of the decentralized block chain federated learning, the safety and privacy of the data are ensured, and the multi-feature input model is adopted, so that the trained model can accurately predict the financial product more suitable for the user's preference.

Description

Technical Field

[0001] This invention relates to the field of blockchain technology, and in particular to a method, apparatus, device, and storage medium for recommending financial products. Background Technology

[0002] With the development of internet technology, online banking and mobile banking applications are becoming increasingly feature-rich, leading many customers to choose to manage their finances remotely from home. However, the current abundance and wide selection of banking products, coupled with the lack of dedicated account managers to recommend products online, can easily overwhelm customers, leaving them unsure where to begin. With the rapid development and widespread application of deep learning technology, methods using deep learning for recommendation systems have emerged.

[0003] However, currently, when using deep learning for financial product recommendations, a large amount of customer data is required for training, and the customer data is transmitted to a central server. Since the banking and financial industry is itself a sector frequently attacked by hackers, the data transmission process poses a significant threat to the security and privacy of customer data. Furthermore, existing financial product recommendation methods, due to the limitations of deep learning models, cannot accurately provide the best product recommendations to match user preferences. Summary of the Invention

[0004] This invention provides a method for recommending financial products, so as to achieve accurate financial product recommendations while ensuring data security.

[0005] According to a first aspect of the present invention, a method for recommending financial products is provided, applied to a blockchain federated learning architecture, the architecture comprising a blockchain layer and a node layer, wherein the node layer is deployed with at least two distributed hosts, and the blockchain layer is deployed with block nodes respectively connected to each of the distributed hosts, the method comprising:

[0006] Each of the distributed hosts generates a local dataset based on user information collected locally, performs initial training on a local initial model based on the local dataset to obtain local initial parameters, and sends the local initial parameters to the corresponding connected block node. The initial model includes a multi-feature input model.

[0007] Each block node calculates and obtains the upload parameters based on the initial parameters directly received and the initial parameters sent by the remaining block nodes, and sends the upload parameters to the corresponding connected distributed host;

[0008] Each of the distributed hosts performs calibration training on the initial model after local initial training according to the uploaded parameters to obtain a local product recommendation model;

[0009] When a product recommendation request is received through the blockchain layer, a target product recommendation model is obtained from the node layer based on the user identifier in the product recommendation request, and the target product recommendation model is used to predict recommended financial products that are suitable for the user.

[0010] According to another aspect of the present invention, a financial product recommendation device is provided. The device includes: a dataset acquisition module, configured to generate a local dataset by each distributed host based on user information collected locally, perform initial training on a local initial model based on the local dataset to obtain local initial parameters, and send the local initial parameters to the corresponding connected block node, wherein the initial model includes a multi-feature input model.

[0011] The uploading parameter acquisition module is used for each block node to calculate and obtain uploading parameters based on the directly received initial parameters and the initial parameters sent by the remaining block nodes, and to send the uploading parameters to the corresponding connected distributed host.

[0012] The product recommendation model acquisition module is used by each of the distributed hosts to perform correction training on the initial model after local initial training according to the uploaded parameters to obtain a local product recommendation model.

[0013] The financial product recommendation module is used to, when a product recommendation request is received through the blockchain layer, search the node layer to obtain a target product recommendation model based on the user identifier in the product recommendation request, and use the target product recommendation model to predict recommended financial products that are suitable for the user.

[0014] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0015] At least one processor; and

[0016] A memory communicatively connected to the at least one processor; wherein,

[0017] The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the method described in any embodiment of the present invention.

[0018] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the method described in any embodiment of the present invention.

[0019] According to another aspect of the present invention, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method described in any embodiment of the present invention.

[0020] The beneficial technical effect of this invention lies in the use of a decentralized blockchain federated learning architecture to train the financial product recommendation model, thereby avoiding malicious attacks when data is transmitted to a central server for training, and ensuring data security and privacy. Furthermore, the distributed host employs a multi-feature input model during local training, enabling the trained model to accurately predict financial products that better match user preferences.

[0021] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart of a financial product recommendation method provided in Embodiment 1 of the present invention;

[0024] Figure 2 This is a schematic diagram of an architecture for blockchain federated learning provided in Embodiment 1 of the present invention;

[0025] Figure 3 This is a flowchart of a financial product recommendation method provided in Embodiment 2 of the present invention;

[0026] Figure 4 This is a schematic diagram of a financial product recommendation device according to Embodiment 3 of the present invention;

[0027] Figure 5 This is a schematic diagram of the structure of the electronic device provided in Embodiment 4 of the present invention. Detailed Implementation

[0028] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0029] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices. Additionally, the information collected in this embodiment is information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure, and application of related data all comply with the relevant laws, regulations, and standards of the relevant countries and regions, necessary confidentiality measures have been taken, public order and good morals have not been violated, and corresponding operation interfaces are provided for users to choose to authorize or refuse.

[0030] Example 1

[0031] Figure 1 This is a flowchart illustrating a method for recommending financial products according to Embodiment 1 of the present invention. This embodiment is applicable to situations where accurate financial product recommendations are made to users. This method can be executed by a financial product recommendation device, which can be implemented in hardware and / or software. Figure 1 As shown, the method includes:

[0032] In step S101, each distributed host generates a local dataset based on the user information collected locally, performs initial training on the local initial model based on the local dataset to obtain local initial parameters, and sends the local initial parameters to the corresponding connected block node.

[0033] Among them, such as Figure 2The diagram illustrates the architecture of the blockchain federated learning described in this embodiment. As shown, the architecture includes a blockchain layer and a node layer. The node layer comprises multiple independent distributed hosts, each storing a local model. The blockchain layer includes multiple block nodes, each containing a compute node and a block representative. Each distributed host in the node layer is virtually connected to a block representative within a block node, and block representatives can transmit information through broadcast within the blockchain layer. This embodiment does not limit the number of distributed hosts in the node layer or the number of block nodes in the blockchain layer; as long as the number of block nodes corresponds to the number of distributed hosts, it falls within the scope of protection of this application.

[0034] Optionally, each distributed host generates a local dataset based on the user information collected locally, including: each distributed host collects user registration information and user purchase information from its local database; preprocesses the user registration information to obtain user behavior feature attributes, and obtains user tags based on the user purchase information; and labels each user's user tags onto the user behavior feature attributes to generate the dataset.

[0035] Specifically, as a decentralized, immutable, and shared distributed ledger and database, blockchain grants equal rights to all nodes in the network, mitigating the potential data leaks inherent in centralized systems and enhancing privacy and data security. As an open system, blockchain allows all data owners to join and update model parameters independently. Furthermore, the immutability and traceability of information within the blockchain mitigate the risk of malicious data breaches. Therefore, in this implementation, leveraging the decentralized nature of blockchain, each distributed host only needs to collect user information from its local database and generate a local dataset. Based on this local dataset, a local model is trained, and training parameters are shared at the blockchain layer. The shared parameters are then used to further train the local model until the required training accuracy or iteration is achieved. The dataset typically consists of two parts: feature attributes and labels. In this implementation, user registration information and information on user purchases of financial products are collected from a local database. Based on the user registration information, preprocessing is performed to obtain user behavioral feature attributes. User registration information may include the user's name, gender, region, occupation, age, and personal assets. User profiling is then performed based on this information. User behavioral features are composed of many words and phrases, which can be processed through natural language processing to form semantically rich term units with the smallest part-of-speech tag. After segmenting and tagging each behavioral feature, missing terms are padded with zeros, and default terms are set to their default values ​​to ensure that each user's term has a value. Normalization, outlier removal, and noise removal are then performed on each term to obtain the final user behavioral feature attributes. The information on the financial products purchased by the user is then used as labels to annotate the user behavioral feature attributes to obtain the final dataset. Since the local model stored on each distributed host is an untrained initial model at the beginning, each distributed host trains its local initial model based on its local dataset to obtain local initial parameters.

[0036] It should be noted that the initial model in this embodiment includes a multi-feature input model, such as a multi-input neural network. The multi-feature input model includes a multi-feature input module, a feature extraction module, and a fusion module. The multi-feature input model is used to input user behavior feature attributes, i.e., user profiles, to complete the collection of input information. The subsequent feature extraction module can be a three-layer convolutional neural network (CNN) used to extract features from different input information. The CNN layers are Conv1, Conv2, Conv3, Conv4, Conv5, Conv6, Conv7, Conv8, and Conv9, respectively. Each convolutional module is used to extract spatial features from one feature. Each CNN layer also includes a zero-padding layer, a dropout layer, an activation function, and a maxpooling layer. The zero-padding layer serves to pad the model. The dropout layer prevents overfitting, the ReLU activation function is used, and the maxpooling layer reduces network complexity. Next, the extracted features are fused through a concatenation layer. The Long Short-Term Memory (LSTM) layer primarily extracts the temporal characteristics of the fused first extracted feature to obtain the second extracted feature. This allows for more comprehensive feature extraction from both temporal and spatial perspectives. Finally, a fully connected (Dense) layer maps the extracted fused features to a sparse space. The softmax function maps the output of the fully connected layer to the (0,1) interval, selecting the modulation scheme with the highest probability as the first recommended prediction. Of course, this embodiment only uses the structure of the multi-feature input model described above as an example to illustrate the initial training process, and does not limit the specific structure of the multi-feature input model. As long as training based on multiple user features can be achieved, it falls within the scope of protection of this application, and this embodiment does not limit it.

[0037] In step S102, each block node calculates and obtains the upload parameters based on the initial parameters directly received and the initial parameters sent by the remaining block nodes, and sends the upload parameters to the corresponding connected distributed host.

[0038] Optionally, before each block node calculates and obtains the upload parameters based on the initial parameters directly received and the initial parameters sent by the remaining block nodes, the process further includes: each block node directly receiving the initial parameters sent by the corresponding connected distributed host through a block representative, sending the directly received initial parameters to the corresponding connected computing node, and broadcasting the initial parameters at the blockchain layer; each block node receiving the initial parameters sent by the block representatives of the remaining block nodes via broadcast through a block representative, and sending the broadcast-received initial parameters to the corresponding connected computing node through a block representative.

[0039] Optionally, each block node calculates and obtains the upload parameters based on the directly received initial parameters and the initial parameters sent by the remaining block nodes, and sends the upload parameters to the corresponding connected distributed host. This includes: the compute nodes in each block node obtaining the weights of each block node; the compute nodes in each block node performing weighted aggregation of the directly received initial parameters and the broadcast received initial parameters based on the weights to obtain the upload parameters; and the compute nodes in each block node sending the upload parameters to the corresponding connected block representative, who then sends them to the distributed host connected to the block node.

[0040] Specifically, in this embodiment, each distributed host sends the initial parameters obtained through training on its local dataset to the corresponding connected block node. Since each block node includes a compute node and a block representative, the initial parameters are sent to the block representative within the block node. Furthermore, each block representative also receives initial parameters broadcast by block representatives from other block nodes. For example, when the node layer includes local host A and local host B, and the corresponding blockchain layer includes block node A and block node B, the block representative receives not only the initial parameter X sent by local host A, but also the initial parameter Y broadcast by block representative B at the blockchain layer. Block representative A then sends both initial parameters X and Y to compute node A. Since different weights are pre-configured for each block representative on the compute node (e.g., for compute node A, the weight of block representative A is 'a', and the weight of block representative B is 'b', with 'a' being greater than 'b', i.e., 'a' is greater than 'b'), the weight of block representative A is greater than 'b'. Directly connected blocks are assigned higher weights, and vice versa. Computation node A will perform weighted aggregation of initial parameters X and Y based on the acquired weights, i.e., X*a + Y*b = Z, and use parameter Z as the final uploaded parameter. Computation node A will transmit the uploaded parameter Z to block representative A, and block representative A will send the uploaded parameter Z to distributed host A. Of course, this embodiment only illustrates the case of two block nodes in the blockchain layer, and does not limit the specific number of block nodes contained in the blockchain layer.

[0041] It should be noted that the parameters sent by each block representative to the distributed host connected to the object are different. For example, although a compute node receives the initial parameter Y sent by block representative B and the initial parameter X sent by block representative A through broadcast at the blockchain layer, for compute node B, the weight of block representative B is a and the weight of block representative A is b. Compute node B will perform weighted aggregation of the initial parameters X and Y according to the obtained weights, that is, X*b+Y*a=W. Therefore, compute node B will transmit the obtained upload parameter W to block representative B, and block representative B will send the upload parameter W to distributed host B. Thus, it can be seen that the values ​​of the upload parameters received by different distributed hosts from the directly connected block representatives are different. Of course, this embodiment is only an example and does not limit the specific values ​​of the upload parameters obtained by each distributed host.

[0042] Optionally, the method also includes: the block representative in each block node updating its own block according to the received uploaded parameters; and the block representative in each block node broadcasting the received uploaded parameters in the blockchain layer so that the remaining block nodes update their own blocks.

[0043] Specifically, in this embodiment, each block representative not only sends the received upload parameters to the directly connected distributed hosts, but also broadcasts the upload parameters received from the computing nodes to the remaining block representatives in the blockchain layer, and receives upload parameters broadcast by other block representatives in the blockchain layer. This allows them to update their own blocks and facilitates other block representatives to update their own blocks. For example, when block representative A receives initial parameter X from local host A, and block representative B receives initial parameter Y from local host B, when... When block representative A receives the uploaded parameter Z from computing node A, and block representative B receives the uploaded parameter W from computing node B, block representative A will update its own initial parameter X to Z, and block representative B will update its own initial parameter Y to W. At the same time, block representative A will also record the newly updated uploaded parameter W from block representative B, and block representative B will also record the uploaded parameter Z updated by block representative A. Of course, this embodiment is only an example and does not limit the specific parameter values ​​updated by each block representative in the blockchain layer.

[0044] In step S103, each distributed host performs calibration training on the initial model after local initial training based on the uploaded parameters to obtain a local product recommendation model.

[0045] Specifically, in this embodiment, after each distributed host in the node layer obtains the uploaded parameters issued by the block representative of the corresponding connection in the blockchain layer, it will perform correction training on the initial model after initial training based on the obtained uploaded parameters until the termination condition is reached or the model reaches the preset accuracy. Therefore, each distributed host only needs to use the data in its local database and can obtain the parameters obtained by training based on the data in other distributed hosts through blockchain federated learning in order to train the local model.

[0046] Federated learning is an emerging artificial intelligence technology that enables efficient machine learning among multiple participants or computing nodes, while ensuring information security during big data exchange, protecting terminal data and personal data privacy, and guaranteeing legal compliance. Unlike distributed machine learning, federated learning requires a decentralized distributed system to ensure user privacy and security, enabling effective machine learning while maintaining data security and exchange / training efficiency. Blockchain, as a decentralized, encrypted, and tamper-proof distributed shared database, can provide data confidentiality for federated learning data exchange to protect user privacy, ensure data security among participants, and guarantee data consistency for model training among multiple participants.

[0047] It should be noted that this embodiment trains the model based on the aforementioned multi-input neural network and blockchain federated learning method. By using the blockchain federated learning framework, it weakens the attack of malicious nodes while proposing a decentralized framework. This framework can more accurately recommend desired financial products to customers without uploading data to a central server, thus meeting users' requirements for data security and privacy and solving the problem of malicious attacks on central servers encountered in existing deep learning-based financial product recommendation scenarios.

[0048] It is worth mentioning that although the initial model is the same in each distributed host in this embodiment, the values ​​of the parameters sent from the blockchain layer are different for each distributed host. Therefore, the final product recommendation model obtained by calibration training with different sent parameters is different. This embodiment does not specifically limit the product recommendation model on each distributed host.

[0049] Step S104: When a product recommendation request is received through the blockchain layer, the target product recommendation model is obtained by searching the node layer according to the user identifier in the product recommendation request, and the target product recommendation model is used to predict the recommended financial products that are suitable for the user.

[0050] Optionally, the target product recommendation model is obtained by searching from the node layer based on the user identifier in the product recommendation request, including: each block node broadcasts the user identifier in the product recommendation request in the blockchain layer so that each block representative sends a search request to the corresponding connected distributed host in the node layer; the distributed host that matches the user identifier is taken as the target distributed host, and the product recommendation model in the target distributed host is taken as the target product recommendation model.

[0051] Optionally, a target product recommendation model is used to predict recommended financial products suitable for the user, including: retrieving behavioral feature attributes of the user to be recommended from a local database through a target distributed host; and using the target product recommendation model to predict and obtain recommended financial products suitable for the user based on the behavioral feature attributes of the user to be recommended, wherein the number of recommended financial products is at least one.

[0052] Specifically, in this embodiment, when a product recommendation request is received through the blockchain layer, since the request includes a user identifier, such as the user name "Zhang San," the user name is broadcast across the blockchain layer to locate the target distributed host in the node layer where the user information exists. The product recommendation model in the target distributed host is then used as the target product recommendation model to respond to this request. For example, when computing node A receives a product recommendation request, it extracts the user identifier "user name Zhang San" and generates a search request, which is sent to block representative A. Block representative A broadcasts the search request across the blockchain layer, so block representative B also receives the search request. Each block representative sends the search request to the directly receiving distributed host. Each distributed host queries its local database based on the user name to determine if the user information is stored. If it is, a match is confirmed. For example, if it is determined that distributed host A stores the user Zhang San's information, then distributed host A is used as the target distributed host, and the product recommendation model trained and stored in distributed host A is used as the target product recommendation model. Of course, this embodiment is only an example and does not limit the specific method for determining the target product recommendation model.

[0053] In this embodiment, after determining the target distributed host and the target product recommendation model, the target distributed host retrieves the behavioral characteristic attributes of the user to be recommended from the local database based on the user's name. There may be missing features, which can be padded with zeros and normalized in this embodiment. The normalized feature attributes are then input into the target product recommendation model. The target product recommendation model is trained to obtain the output results containing recommended financial products suitable for the user and the corresponding recommendation probabilities. Product recommendations are then made based on the data results.

[0054] In this embodiment, a decentralized blockchain federated learning architecture is used to train the financial product recommendation model, thereby avoiding malicious attacks when data is transmitted to a central server for training, ensuring data security and privacy. Furthermore, the distributed host employs a multi-feature input model during local training, enabling the trained model to accurately predict financial products that better match user preferences.

[0055] Example 2

[0056] Figure 3 This is a flowchart of a financial product recommendation method provided in Embodiment 2 of the present invention. Based on the above embodiments, after predicting recommended financial products suitable for the user using a target product recommendation model, this embodiment further includes: receiving the selected recommendation model and visually displaying the recommended financial products on the front end according to the recommendation model. Figure 3 As shown, the method includes:

[0057] In step S201, each distributed host generates a local dataset based on the user information collected locally, performs initial training on the local initial model based on the local dataset to obtain local initial parameters, and sends the local initial parameters to the corresponding connected block node.

[0058] Optionally, each distributed host generates a local dataset based on the user information collected locally, including: each distributed host collects user registration information and user purchase information from its local database; preprocesses the user registration information to obtain user behavior feature attributes, and obtains user tags based on the user purchase information; and labels each user's user tags onto the user behavior feature attributes to generate the dataset.

[0059] In step S202, each block node calculates and obtains the upload parameters based on the initial parameters directly received and the initial parameters sent by the remaining block nodes, and sends the upload parameters to the corresponding connected distributed host.

[0060] Optionally, before each block node calculates and obtains the upload parameters based on the initial parameters directly received and the initial parameters sent by the remaining block nodes, the process further includes: each block node directly receiving the initial parameters sent by the corresponding connected distributed host through a block representative, sending the directly received initial parameters to the corresponding connected computing node, and broadcasting the initial parameters at the blockchain layer; each block node receiving the initial parameters sent by the block representatives of the remaining block nodes via broadcast through a block representative, and sending the broadcast-received initial parameters to the corresponding connected computing node through a block representative.

[0061] In step S203, each distributed host performs calibration training on the initial model after local initial training based on the uploaded parameters to obtain a local product recommendation model.

[0062] Step S204: When a product recommendation request is received through the blockchain layer, the target product recommendation model is obtained by searching the node layer according to the user identifier in the product recommendation request, and the target product recommendation model is used to predict the recommended financial products that are suitable for the user.

[0063] Optionally, the target product recommendation model is obtained by searching from the node layer based on the user identifier in the product recommendation request, including: each block node broadcasts the user identifier in the product recommendation request in the blockchain layer so that each block representative sends a search request to the corresponding connected distributed host in the node layer; the distributed host that matches the user identifier is taken as the target distributed host, and the product recommendation model in the target distributed host is taken as the target product recommendation model.

[0064] Optionally, a target product recommendation model is used to predict recommended financial products suitable for the user, including: retrieving behavioral feature attributes of the user to be recommended from a local database through a target distributed host; and using the target product recommendation model to predict and obtain recommended financial products suitable for the user based on the behavioral feature attributes of the user to be recommended, wherein the number of recommended financial products is at least one.

[0065] Step S205: Receive the selected recommendation model and visualize the recommended financial products on the front end based on the recommendation model.

[0066] Specifically, the recommendation model in this embodiment includes comprehensive recommendation or targeted recommendation. When a user has sufficient funds and needs to purchase multiple financial products, they can select comprehensive recommendation on the front-end interface. When the user receives the selection instruction for comprehensive recommendation, all recommended financial products output by the financial product recommendation model will be sorted in descending order of recommendation probability and displayed on the front end. Conversely, when a user has insufficient funds and only needs to purchase one financial product, they can select targeted recommendation on the front-end interface. When the user receives the selection instruction for targeted recommendation, the financial product with the highest recommendation probability in the financial product recommendation model will be displayed on the front end.

[0067] In this implementation, the predicted recommended financial products can be personalized based on different user purchasing needs, making the displayed products more aligned with user requirements and improving user experience. Furthermore, if none of the recommended financial products are suitable for the user, an option to exclude them can be selected on the interface. This allows the distributed host to retrain and modify the model based on user feedback, thereby further improving the model's prediction accuracy.

[0068] In this embodiment, a decentralized blockchain federated learning architecture is used to train the financial product recommendation model, thereby avoiding malicious attacks when data is transmitted to a central server for training, ensuring data security and privacy. Furthermore, the distributed host employs a multi-feature input model during local training, enabling the trained model to accurately predict financial products that better match user preferences. Based on different user purchasing needs, the predicted recommended financial products are personalized, making the displayed products more aligned with user needs and improving the user experience.

[0069] Example 3

[0070] Figure 4 This is a schematic diagram of a financial product recommendation device provided in Embodiment 3 of the present invention. Figure 4 As shown, the device includes: a dataset acquisition module 310, an upload parameter acquisition module 320, a product recommendation model acquisition module 330, and a financial product recommendation module 340.

[0071] The dataset acquisition module 310 is used by each distributed host to generate a local dataset based on the user information collected locally, to perform initial training on the local initial model based on the local dataset to obtain local initial parameters, and to send the local initial parameters to the corresponding connected block node. The initial model includes a multi-feature input model.

[0072] The uploading parameter acquisition module 320 is used by each block node to calculate and obtain the uploading parameters based on the initial parameters directly received and the initial parameters sent by the remaining block nodes, and then send the uploading parameters to the corresponding connected distributed host.

[0073] The product recommendation model acquisition module 330 is used by each distributed host to perform correction training on the initial model after local initial training based on the uploaded parameters to obtain the local product recommendation model.

[0074] The financial product recommendation module 340 is used to search for the target product recommendation model from the node layer based on the user identifier in the product recommendation request when a product recommendation request is received through the blockchain layer, and then use the target product recommendation model to predict the recommended financial products that are suitable for the user.

[0075] Optionally, a dataset collection module is provided, which is used by each distributed host to collect data from the local database to obtain user registration information and information on users' purchase of financial products;

[0076] Preprocess user registration information to obtain user behavior characteristics and attributes, and obtain user tags based on user purchase information of financial products.

[0077] Generate a dataset by labeling each user's user tags onto user behavior feature attributes.

[0078] Optionally, the block nodes include compute nodes and block representatives connected to the compute nodes; the device also includes an initial parameter receiving module, which is used for each block node to directly receive the initial parameters sent by the corresponding connected distributed host through the block representative, and to send the directly received initial parameters to the corresponding connected compute nodes, and to broadcast the initial parameters in the blockchain layer.

[0079] Each block node receives the initial parameters broadcast by the block representatives of the remaining block nodes through its block representative, and then forwards the received initial parameters to the corresponding connected compute nodes through its block representative.

[0080] Optionally, a parameter acquisition module is provided to allow the computing nodes in each block node to obtain the weights of each block node.

[0081] The computing nodes in each block node perform weighted aggregation of the directly received initial parameters and the broadcast received initial parameters to obtain the parameters to be sent up.

[0082] The compute nodes in each block node send the parameters to the corresponding connected block representative, and then the block representative sends them to the distributed host connected to the corresponding block node.

[0083] Optionally, the device also includes a block update module, which is used for the block representative in each block node to update its own block according to the received uploaded parameters;

[0084] The block representatives in each block node broadcast the received parameters in the blockchain layer so that the remaining block nodes can update their own blocks.

[0085] Optionally, the financial product recommendation module includes a target product recommendation model search submodule, which is used by each block node to broadcast the user identifier in the product recommendation request in the blockchain layer, so that each block representative can send a search request to the corresponding distributed host in the node layer.

[0086] The distributed host that matches the user identifier is taken as the target distributed host, and the product recommendation model in the target distributed host is taken as the target product recommendation model.

[0087] Optionally, the financial product recommendation module includes a financial product recommendation sub-module, which is used to retrieve the behavioral characteristic attributes of the users to be recommended from the local database through the target distributed host;

[0088] A target product recommendation model is used to predict and obtain recommended financial products that are suitable for the user based on the behavioral characteristics of the user to be recommended. The number of recommended financial products is at least one.

[0089] Optionally, the device also includes a personalized display module for receiving the selected recommendation model, wherein the recommendation model includes comprehensive recommendations or targeted recommendations;

[0090] Based on the recommendation model, the recommended financial products are displayed visually on the front end.

[0091] Optionally, the multi-feature input model includes: a multi-feature input module, a feature extraction module, and a fusion module.

[0092] The financial product recommendation device provided in this embodiment of the invention can execute the financial product recommendation method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0093] Example 4

[0094] Figure 5 A schematic diagram of an electronic device 10 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0095] like Figure 5As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0096] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0097] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as the incremental synchronization method for database table structures.

[0098] In some embodiments, the incremental synchronization method applied to the database table structure can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the incremental synchronization method for the database table structure described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to perform the incremental synchronization method for the database table structure by any other suitable means (e.g., by means of firmware).

[0099] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0100] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0101] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0102] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0103] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0104] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through a communication network. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0105] Example 5

[0106] This invention also provides a computer program product, including a computer program that, when executed by a processor, implements the financial product recommendation method as provided in any embodiment of this application.

[0107] In implementing the computer program product, computer program code for performing the operations of this invention can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0108] It should be noted that in the embodiments of this application, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, they do not mean that the applicant has used or necessarily used the solution.

[0109] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.

Claims

1. A method for recommending financial products, characterized in that, An architecture for blockchain federated learning, comprising a blockchain layer and a node layer, wherein the node layer deploys at least two distributed hosts, and the blockchain layer deploys block nodes connected to each of the distributed hosts, the method comprising: Each of the distributed hosts generates a local dataset based on user information collected locally, performs initial training on a local initial model based on the local dataset to obtain local initial parameters, and sends the local initial parameters to the corresponding connected block node. The initial model includes a multi-feature input model. Each block node calculates and obtains the upload parameters based on the initial parameters directly received and the initial parameters sent by the remaining block nodes, and sends the upload parameters to the corresponding connected distributed host; Each of the distributed hosts performs calibration training on the initial model after local initial training based on the uploaded parameters to obtain a local product recommendation model; When a product recommendation request is received through the blockchain layer, a target product recommendation model is obtained from the node layer based on the user identifier in the product recommendation request, and the target product recommendation model is used to predict recommended financial products that are suitable for the user.

2. The method according to claim 1, characterized in that, Each of the distributed hosts generates a local dataset based on user information collected locally, including: Each of the distributed hosts collects user registration information and user purchase information from its local database; The user registration information is preprocessed to obtain user behavior characteristic attributes, and the user tags are obtained based on the user's purchase information of financial products. The dataset is generated by labeling each user's user tags onto the user behavior feature attributes.

3. The method according to claim 1, characterized in that, The block node includes a compute node and a block representative connected to the compute node; Before each block node calculates and obtains the uploaded parameters based on the directly received initial parameters and the initial parameters sent by the remaining block nodes, the process further includes: Each block node directly receives the initial parameters sent by the corresponding connected distributed host through the block representative, sends the directly received initial parameters to the corresponding connected computing node, and broadcasts the initial parameters in the blockchain layer; Each block node receives the initial parameters broadcast by the block representatives of the remaining block nodes through the block representative, and then sends the broadcast-received initial parameters to the corresponding connected computing node through the block representative.

4. The method according to claim 3, characterized in that, Each block node calculates and obtains the upload parameters based on the directly received initial parameters and the initial parameters sent by the remaining block nodes, and sends the upload parameters to the corresponding connected distributed host, including: The computing nodes in each block node obtain the weights of each block node; The computing nodes in each block node perform weighted aggregation of the directly received initial parameters and the broadcast received initial parameters based on the weight to obtain the parameters to be sent up. The computing nodes in each block node send the uploaded parameters to the corresponding connected block representative, and then the block representative sends them to the distributed host connected to the block node.

5. The method according to claim 4, characterized in that, The method further includes: Each block node's block representative updates its own block based on the received uploaded parameters; The block representative in each of the block nodes broadcasts the received uploaded parameters in the blockchain layer so that the remaining block nodes can update their own blocks.

6. The method according to claim 3, characterized in that, The step of searching for and obtaining the target product recommendation model from the node layer based on the user identifier in the product recommendation request includes: Each of the block nodes broadcasts the user identifier in the product recommendation request to the blockchain layer, so that each of the block representatives sends a lookup request to the corresponding distributed host in the node layer; The distributed host that matches the user identifier is taken as the target distributed host, and the product recommendation model in the target distributed host is taken as the target product recommendation model.

7. The method according to claim 6, characterized in that, The step of using the target product recommendation model to predict recommended financial products suitable for users includes: The behavioral characteristic attributes of the users to be recommended are retrieved from the local database using the target distributed host. The target product recommendation model is used to predict and obtain recommended financial products that are suitable for the user based on the behavioral characteristics of the user to be recommended, wherein the number of recommended financial products is at least one.

8. The method according to claim 1, characterized in that, After predicting the recommended financial products suitable for the user using the target product recommendation model, the process further includes: Receive the selected recommendation model, wherein the recommendation model includes comprehensive recommendation or targeted recommendation; The recommended financial products are then visualized on the front end based on the recommendation model.

9. The method according to any one of claims 1-8, characterized in that, The multi-feature input model includes: a multi-feature input module, a feature extraction module, and a fusion module.

10. A financial product recommendation device, characterized in that, An architecture for blockchain federated learning, comprising a blockchain layer and a node layer, wherein the node layer deploys at least two distributed hosts, and the blockchain layer deploys block nodes connected to each of the distributed hosts, the device comprising: The dataset acquisition module is used by each distributed host to generate a local dataset based on the user information collected locally, to perform initial training on the local initial model based on the local dataset to obtain local initial parameters, and to send the local initial parameters to the corresponding connected block node. The initial model includes a multi-feature input model. The uploading parameter acquisition module is used for each block node to calculate and obtain uploading parameters based on the directly received initial parameters and the initial parameters sent by the remaining block nodes, and to send the uploading parameters to the corresponding connected distributed host. The product recommendation model acquisition module is used by each of the distributed hosts to perform correction training on the initial model after local initial training according to the uploaded parameters to obtain a local product recommendation model. The financial product recommendation module is used to, when a product recommendation request is received through the blockchain layer, search the node layer to obtain a target product recommendation model based on the user identifier in the product recommendation request, and use the target product recommendation model to predict recommended financial products that are suitable for the user.

11. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-9.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that are used to cause a processor to execute the method of any one of claims 1-9.

13. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method of any one of claims 1-9.

Images