Financial product recommendation method and device based on multi-source heterogeneous data and storage medium

Abstract

The invention discloses a financial product recommendation method and device based on multi-source heterogeneous data and a storage medium. The method relates to the field of artificial intelligence, and comprises the following steps: acquiring multi-source heterogeneous data of a target user from a plurality of data sources under the condition of acquiring authorization of the target user of a target financial institution; according to the target product recommendation knowledge graph, obtaining association relationship characteristics among the multi-source heterogeneous data; acquiring multi-modal data according to the multi-source heterogeneous data, and acquiring multi-modal features corresponding to the multi-modal data by using a federated learning model; determining input data according to the association relationship characteristics and the multi-modal characteristics, inputting the input data into a preset time sequence model, and outputting financial product recommendation information corresponding to the target user; and displaying the financial product recommendation information to the target user by using an augmented reality technology. Through the method and the device, the problem of relatively poor recommendation timeliness of financial products in related technologies is solved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence, and more specifically, to a method, apparatus, and storage medium for recommending financial products based on multi-source heterogeneous data. Background Technology

[0002] Financial product recommendation is a crucial component of banking operations. It aims to recommend the most suitable financial products and services to customers by analyzing their personal data, financial situation, risk preferences, and transaction history. To improve the accuracy and personalization of recommendations, existing methods typically employ data analytics and machine learning models, such as user profiling, personalized ranking algorithms, and deep learning neural networks. These methods learn customer preferences from training datasets to provide more tailored product recommendations.

[0003] However, in existing technologies, whether based on traditional statistical models or machine learning models, although training on training datasets can improve recommendation accuracy, the reliance on periodic training and updates, coupled with the inherent lag in these datasets (e.g., training and iterating only every few months), means that the model's weight parameters are not updated in a timely manner. Consequently, the prediction results are difficult to adjust promptly according to changes in customer preferences, and existing financial product recommendation methods still suffer from significant shortcomings in terms of timeliness. For example, when customer preferences have changed, recommendations may not reflect this immediately due to the long model update cycle. Continuing to recommend based on old data will result in products that are no longer suitable, reducing the relevance and accuracy of the recommendations.

[0004] There is currently no effective solution to the problem of poor timeliness in the recommendation methods for financial products in related technologies. Summary of the Invention

[0005] The main purpose of this application is to provide a method, apparatus and storage medium for recommending financial products based on multi-source heterogeneous data, so as to solve the problem of poor timeliness of financial product recommendations in related technologies.

[0006] To achieve the above objectives, according to one aspect of this application, a method for recommending financial products based on multi-source heterogeneous data is provided. The method includes: obtaining multi-source heterogeneous data of the target user from multiple data sources, with authorization from the target user of the target financial institution; obtaining relational features between the multi-source heterogeneous data based on a target product recommendation knowledge graph, wherein the target product recommendation knowledge graph is obtained by updating a product recommendation knowledge graph based on multiple data sources; obtaining multimodal data based on the multi-source heterogeneous data, and using a federated learning model to obtain multimodal features corresponding to the multimodal data; determining input data based on the relational features and multimodal features, and inputting the input data into a preset time series model to output financial product recommendation information corresponding to the target user; and displaying the financial product recommendation information to the target user using augmented reality technology.

[0007] Furthermore, the multiple data sources include at least user identity information, user financial information, and news event information. Updating the product recommendation knowledge graph based on multiple data sources includes: continuously monitoring data changes across multiple data sources; when any data source changes, using an information extraction algorithm to extract new data from the changed data source to determine entity extraction results and relation extraction results; and updating the product recommendation knowledge graph based on the entity extraction results and relation extraction results.

[0008] Furthermore, updating the product recommendation knowledge graph based on the entity extraction results and relation extraction results includes: mapping the entity extraction results and relation extraction results to quantum superposition state data; using a quantum search algorithm to traverse all paths in the product recommendation knowledge graph in parallel to determine the update target; and updating the product recommendation knowledge graph based on the update target and the quantum superposition state data.

[0009] Furthermore, multimodal data includes data from multiple different modalities. Obtaining multimodal features corresponding to multimodal data using a federated learning model includes: co-training multiple participants on training sets of the modalities corresponding to each participant using a pre-defined federated learning framework to obtain multiple local models; aggregating the multiple local models to determine the federated learning model; and inputting the multimodal data into the federated learning model to obtain multimodal features.

[0010] Furthermore, the federated learning model is determined by aggregating multiple local models, including: determining the modal weights of each local model based on its corresponding mode; encoding the model parameters in each local model into quantum state parameters; and aggregating the quantum state parameters based on the modal weights using quantum teleportation methods to obtain the federated learning model.

[0011] Furthermore, the preset time series model is a long short-term memory neural network model. Inputting the input data into the preset time series model and outputting financial product recommendation information corresponding to the target user includes: converting the input data into quantum state input data and converting the gating unit of the long short-term memory neural network model into quantum gate operations; using quantum gate operations for parallel computation to predict the quantum state probability amplitude corresponding to the quantum state input data; and determining the financial product recommendation information corresponding to the target user based on the quantum state probability amplitude.

[0012] Furthermore, the financial product recommendation information includes introductory information for multiple target financial products and a recommendation index for each target financial product. Presenting this information to target users using augmented reality technology involves: retrieving each target financial product from a target product recommendation knowledge graph to determine the target entities corresponding to the target financial products and the target relationships between these entities; determining the information transmission chain of the target financial products based on the target entities and target relationships; and using augmented reality technology to present the information transmission chain to target users in animated form, setting different display identifiers for the information transmission chains of different target financial products based on the recommendation index.

[0013] To achieve the above objectives, according to another aspect of this application, a financial product recommendation device based on multi-source heterogeneous data is provided. The device includes: a data acquisition unit, used to acquire multi-source heterogeneous data of a target user from multiple data sources, provided that authorization has been obtained from the target user of the target financial institution; a relation extraction unit, used to acquire the association features between the multi-source heterogeneous data based on a target product recommendation knowledge graph, wherein the target product recommendation knowledge graph is obtained by updating a product recommendation knowledge graph based on multiple data sources; a modality extraction unit, used to acquire multimodal data based on the multi-source heterogeneous data and to acquire the multimodal features corresponding to the multimodal data using a federated learning model; a recommendation prediction unit, used to determine input data based on the association features and multimodal features, input the input data into a preset time series model, and output financial product recommendation information corresponding to the target user; and an information display unit, used to display the financial product recommendation information to the target user using augmented reality technology.

[0014] Furthermore, the multiple data sources include at least user identity information, user financial information, and news event information. The relationship extraction unit includes: a change detection module, used to continuously detect data changes from multiple data sources; a change extraction module, used to extract new data from the data source with data changes when any data source changes, using an information extraction algorithm to determine the entity extraction result and the relationship extraction result; and a graph update module, used to update the product recommendation knowledge graph based on the entity extraction result and the relationship extraction result.

[0015] Furthermore, the graph update module includes: a first update submodule, used to map entity extraction results and relation extraction results to quantum superposition state data; a second update submodule, used to traverse all paths in the product recommendation knowledge graph in parallel according to the quantum search algorithm to determine the update target; and a third update submodule, used to update the product recommendation knowledge graph according to the update target and the quantum superposition state data.

[0016] Furthermore, the multimodal data includes data from multiple different modalities. The modality extraction unit includes: a collaborative training module, used to collaboratively train the training set of the modality corresponding to each participant using a preset federated learning framework to obtain multiple local models; a model aggregation module, used to aggregate the multiple local models to determine the federated learning model; and a modality feature module, used to input the multimodal data into the federated learning model to obtain multimodal features.

[0017] Furthermore, the model aggregation module is based on multiple local models and includes: a weight determination submodule, which determines the modal weights of each local model based on the modality of each local model; a parameter encoding submodule, which encodes the model parameters in each local model into quantum state parameters; and a quantum aggregation submodule, which aggregates the quantum state parameters based on the modal weights using quantum teleportation to obtain the federated learning model.

[0018] Furthermore, the preset time series model is a long short-term memory neural network model, and the recommendation prediction unit includes: an input transformation module, used to convert input data into quantum state input data and convert the gating unit of the long short-term memory neural network model into quantum gate operations; a parallel computing module, used to perform parallel computing using quantum gate operations to predict the quantum state probability amplitude corresponding to the quantum state input data; and a probability amplitude measurement module, used to determine the financial product recommendation information corresponding to the target user based on the quantum state probability amplitude.

[0019] Furthermore, the financial product recommendation information includes introductory information for multiple target financial products and a recommendation index for each target financial product. The information display unit includes: a target retrieval module, used to retrieve each target financial product in the target product recommendation knowledge graph to determine the target entity corresponding to the target financial product and the target association between the target entities; a transmission chain determination module, used to determine the information transmission chain of the target financial product based on the target entity and the target association; and an augmented reality module, used to use augmented reality technology to display the information transmission chain to the target user in the form of animation, and to set different display labels for the information transmission chain of different target financial products based on the recommendation index.

[0020] According to another aspect of this application, a computer-readable storage medium is provided, which includes a stored program, wherein, when the program is running, it controls the device where the computer-readable storage medium is located to execute any financial product recommendation method based on multi-source heterogeneous data.

[0021] According to another aspect of this application, an electronic device is provided, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include methods for performing any financial product recommendation method based on multi-source heterogeneous data.

[0022] According to another aspect of this application, a computer program product is provided, including computer instructions that, when executed by a processor, implement the steps of the financial product recommendation method based on multi-source heterogeneous data as described above.

[0023] In this embodiment, with the authorization of the target user from the target financial institution, multi-source heterogeneous data of the target user is obtained from multiple data sources; the correlation features between the multi-source heterogeneous data are obtained based on the target product recommendation knowledge graph, wherein the target product recommendation knowledge graph is obtained by updating the product recommendation knowledge graph based on multiple data sources; multimodal data is obtained based on the multi-source heterogeneous data, and multimodal features corresponding to the multimodal data are obtained using a federated learning model; input data is determined based on the correlation features and multimodal features, and the input data is input into a preset time series model to output financial product recommendation information corresponding to the target user; augmented reality technology is used to display the financial product recommendation information to the target user, thus solving the technical problem of poor timeliness of financial product recommendations in related technologies.

[0024] By leveraging a target product recommendation knowledge graph, timely updates based on multiple data sources are achieved, thereby extracting the correlation features of multiple data sources. Simultaneously, a federated learning model is used to extract multimodal features from heterogeneous multi-source data, providing information dimensions not available in single-modal data. A pre-defined time series model is used to predict financial product recommendation information based on correlation features and multimodal features, thus enabling timely response to data changes and efficient prediction, improving the timeliness of financial product recommendations. Attached Figure Description

[0025] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0026] Figure 1A hardware structure block diagram of a computer terminal for implementing a financial product recommendation method based on multi-source heterogeneous data is shown.

[0027] Figure 2 This is a flowchart of a financial product recommendation method based on multi-source heterogeneous data provided in the embodiments of this application;

[0028] Figure 3 This is a schematic diagram of a financial product recommendation device based on multi-source heterogeneous data provided in the embodiments of this application;

[0029] Figure 4 This is a structural block diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0030] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0031] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application 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 this application 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 apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0032] First, some nouns or terms that appear in the description of the embodiments of this application shall be interpreted as follows:

[0033] It should be noted that the information collected in this application (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) are information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of this data all comply with relevant laws, regulations, and standards, necessary confidentiality measures have been taken, and they do not violate public order and good morals. Corresponding access points are provided for users to choose to authorize or refuse. For example, interfaces are set up between this system and relevant users or organizations, providing users with corresponding access points to choose to agree to or refuse automated decision-making results; if the user chooses to refuse, the process proceeds to the expert decision-making stage.

[0034] Example 1

[0035] According to an embodiment of this application, a method embodiment for recommending financial products based on multi-source heterogeneous data is also provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0036] The method embodiment provided in Embodiment 1 of this application can be executed on a mobile terminal, computer terminal, or similar computing device. Figure 1 A hardware block diagram of a computer terminal (or mobile device) for implementing a financial product recommendation method based on multi-source heterogeneous data is shown. Figure 1 As shown, the computer terminal 10 (or mobile device) may include one or more processors 102 (shown as 102a, 102b, ..., 102n in the figure) 102 (processor 102 may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0037] It should be noted that the aforementioned one or more processors 102 and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10 (or mobile device). As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0038] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the financial product recommendation method based on multi-source heterogeneous data in this embodiment of the application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the aforementioned financial product recommendation method based on multi-source heterogeneous data. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0039] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0040] The display can be, for example, a touchscreen liquid crystal display (LCD) that allows the user to interact with the user interface of the computer terminal 10 (or mobile device).

[0041] Under the aforementioned operating environment, this application provides the following: Figure 2 The method for recommending financial products based on multi-source heterogeneous data is shown. Figure 2 This is a flowchart of a financial product recommendation method based on multi-source heterogeneous data according to Embodiment 1 of this application.

[0042] Step S201: With the authorization of the target user from the target financial institution, obtain multi-source heterogeneous data of the target user from multiple data sources.

[0043] Optionally, multiple data sources can be the target financial institution's application terminals, the target financial institution's databases, social media, industry news reports, etc. Multi-source heterogeneous data refers to a collection of data from multiple different sources and with different data structures or formats (e.g., structured data, semi-structured data, text data, image data, tabular data).

[0044] For example, transaction records of target users' financial products (including transaction time, product, and amount), textual evaluations of financial products by target users, and audio recordings of interactions between target users and customer service can be collected through the target financial institution's application terminals. Data on multiple financial products and personal information of target users (including age, gender, images of financial product purchase contracts, asset verification images, and investment intention questionnaire data) can also be collected through the target financial institution's database. Forms and texts related to financial products can be obtained through industry news reports, and textual evaluations of target users' consumption preferences can be obtained through social media.

[0045] Step S202: Obtain the relationship features between multi-source heterogeneous data based on the target product recommendation knowledge graph. The target product recommendation knowledge graph is obtained by updating the product recommendation knowledge graph based on multiple data sources.

[0046] Optionally, the target product recommendation knowledge graph is one of the core components of the financial product recommendation method based on multi-source heterogeneous data in this application embodiment. It is used to integrate multi-source heterogeneous data from multiple data sources. Since the core of a knowledge graph is a graph data structure composed of entities (nodes) and relations (edges), it is very suitable for representing and storing complex relationship networks. Through knowledge fusion, knowledge disambiguation, etc., the knowledge graph can integrate information from different data sources. In addition, the entities and relations in the knowledge graph can be independently expanded and updated. Therefore, when data changes are detected in the data source, new nodes or edges can be added quickly, which can respond to data changes in a timely manner compared with traditional databases. Obtaining the association relationship features between multi-source heterogeneous data based on the target product recommendation knowledge graph can be done by extracting the triple data involving the target user in the target product recommendation knowledge graph and using all triple data as association relationship features.

[0047] Step S203: Obtain multimodal data based on multi-source heterogeneous data, and use a federated learning model to obtain the multimodal features corresponding to the multimodal data.

[0048] Optionally, multi-source heterogeneous data can be categorized and organized according to its modality to obtain multimodal data stored by modality. The federated learning model can include multiple local models, each used to make predictions based on data of one modality. For example, the federated learning model may include a first local model for processing text data, a second local model for processing tabular data, a third local model for processing image data, a fourth local model for processing audio data, and a fifth local model for processing image data. Different local models processing the same modality can be assigned the same or different weights. The local models in the federated learning model can be multi-label classification algorithm models; that is, the prediction result of each local model can be the probability value of multiple financial products. Multimodal features can be the output of the federated learning model; for example, the prediction results of single modalities from different local models can be concatenated to obtain multimodal features.

[0049] Step S204: Determine the input data based on the correlation characteristics and multimodal characteristics, input the input data into the preset time series model, and output the financial product recommendation information corresponding to the target user.

[0050] Optionally, the relationship features are derived from the target product recommendation knowledge graph, reflecting the data features between different data sources, such as the relationship paths between users and products, and between products. The multimodal features include features extracted from different modal data such as text, images, and audio. The relationship features and multimodal features can be concatenated by feature concatenation to obtain the input data.

[0051] For example, when the association features are data in the form of triples, the timestamp of each triple when it was created in the target product recommendation knowledge graph is obtained. All triples are then processed into text vectors to obtain multiple relation vectors. These relation vectors are arranged in chronological order according to their timestamps, and multimodal features are added to the end of the arranged relation vectors to obtain time-series input data. For example, the input data could be [VEC1, VEC2, VEC3, ..., VECN, UNION], where VECN represents the Nth relation vector and UNION represents the vectorized multimodal feature. In other words, time-series data is constructed according to the order in which the triples were generated, and the multimodal features are used as the most recent time-series data to construct the input data. The preset time-series model can be a model trained based on a Long Short-Term Memory (LSTM) neural network.

[0052] Step S205: Use augmented reality technology to display financial product recommendation information to target users.

[0053] For example, the interface generated by augmented reality technology can display financial product recommendations in real time, including product name, type, expected return, investment period, and other information about the financial products. The financial product recommendations appear in the form of 3D labels, charts, or virtual cards. Through augmented reality technology, it provides target users with an intuitive and convenient interactive method, presenting financial product recommendations in a visual way and improving the efficiency of target users in understanding financial product recommendations.

[0054] In summary, the target product recommendation knowledge graph enables timely updates based on multiple data sources, thereby extracting the correlation features of multiple data sources. Furthermore, a federated learning model is used to extract multimodal features from heterogeneous multi-source data, providing information dimensions not available with single-modal data. A pre-defined time series model is then used to predict financial product recommendation information based on correlation features and multimodal features. This achieves timely response to data changes and efficient prediction, improving the timeliness of financial product recommendations.

[0055] To improve the timeliness of the target product recommendation knowledge graph, optionally, multiple data sources include at least user identity information, user financial information, and news event information. Updating the product recommendation knowledge graph based on multiple data sources includes: continuously monitoring data changes across multiple data sources; when any data source changes, using an information extraction algorithm to extract new data from the changed data source, determining the entity extraction results and relation extraction results; and updating the product recommendation knowledge graph based on the entity extraction results and relation extraction results.

[0056] Optionally, the product recommendation knowledge graph can be a pre-built knowledge graph. It can detect data changes in the data source by setting up a real-time data stream listener, periodically scraping data, or subscribing to a real-time update service provided by the data source. When a change is detected in data from a certain data source, an information extraction algorithm is invoked to extract entities and relationships from the updated data, identifying newly emerging entities and relationships.

[0057] For example, if a target financial institution updates the news event information "New financial product X launched, with a yield of A and a holding period of B", then the entity extraction results are financial product X, yield A, and holding period B, and the relation extraction results are X-yield attribute-A and X-term attribute-B. Based on the entity extraction results and relation extraction results, the product recommendation knowledge graph is updated, and corresponding entity nodes and relation edges are added, thereby obtaining the target product recommendation knowledge graph.

[0058] In summary, by detecting changes in the data source and extracting new data using information extraction algorithms, the product recommendation knowledge graph is automatically and efficiently updated, thus improving its timeliness.

[0059] To improve the timeliness of the target product recommendation knowledge graph, optionally, updating the product recommendation knowledge graph based on entity extraction results and relation extraction results includes: mapping entity extraction results and relation extraction results to quantum superposition state data; using a quantum search algorithm to traverse all paths in the product recommendation knowledge graph in parallel to determine the update target; and updating the product recommendation knowledge graph based on the update target and the quantum superposition state data.

[0060] Optionally, entity extraction and relation extraction results can be converted into quantum states. Each entity extraction result can be considered as a qubit, while relation extraction results are represented by connections between qubits (e.g., quantum gate operations or quantum entanglement). Quantum superposition allows an entity extraction result to simultaneously represent multiple attributes. Quantum superposition can be used to represent multiple attributes of an entity extraction result or relation extraction results of multiple entity extraction results as a single quantum superposition state. Searching for all possible paths in the target product recommendation knowledge graph using traditional search methods is a very time-consuming process. However, quantum search algorithms can traverse all paths in parallel, quickly determining all paths related to entity or relation extraction results in the target product recommendation knowledge graph with far fewer search iterations than traditional search methods, thus rapidly determining the update target.

[0061] For example, if the entity extraction result is a user's "holding" relationships, this user may have holding relationships with multiple financial products. By mapping this to quantum superposition data, all these relationships can simultaneously exist in a single quantum vector. Then, a quantum search algorithm can quickly determine the user's node in the product recommendation knowledge graph, and by adding corresponding edges to that user node based on the quantum superposition data, updates can be achieved quickly.

[0062] In summary, by mapping entities and relationships to quantum superposition states and then using a quantum search algorithm for parallel traversal and updating, real-time updates of the product recommendation knowledge graph are achieved, thereby significantly improving the timeliness of the target product recommendation knowledge graph.

[0063] To improve the accuracy of financial product recommendation information, optionally, multimodal data includes data from multiple different modalities. Obtaining multimodal features corresponding to the multimodal data using a federated learning model includes: co-training multiple local models on the training sets of the modalities corresponding to the participants using a preset federated learning framework; aggregating the multiple local models to determine the federated learning model; and inputting the multimodal data into the federated learning model to obtain multimodal features.

[0064] For example, multimodal data typically includes information of various modalities such as text (e.g., consultation texts), numerical data (e.g., financial statements, transaction records), images (e.g., ID cards, asset certificates), and audio (communication records between users and customer service). Each data modality can reveal customer behavior, preferences, and risk characteristics from different perspectives. Each participant uses its own labeled training dataset and a pre-defined federated learning framework to train a local model, learning the features in the corresponding modality of data, such as customer risk preferences and financial health status. Each participant encrypts the update parameters of its local model (e.g., homomorphic encryption) and uploads the encrypted model parameters to the federated server, ensuring that the model parameters do not reveal the participants' data details during the aggregation process. The federated server aggregates the model parameters from different participants to generate a global model, i.e., the federated learning model. Multimodal data is input into the federated learning model, and the prediction results output by the federated learning model are used as multimodal features.

[0065] In summary, a federated learning model was constructed using local models of multiple modalities, which enables more comprehensive and detailed extraction of customer features. Multimodal features capture details that are difficult to discover in single-modal data, thereby improving the accuracy of financial product recommendation information.

[0066] To improve the timeliness of financial product recommendation information, optionally, the federated learning model is determined by aggregating multiple local models, including: determining the modal weights of each local model based on the modality corresponding to each local model; encoding the model parameters in each local model into quantum state parameters; and aggregating the quantum state parameters according to the modal weights using quantum teleportation methods to obtain the federated learning model.

[0067] Optionally, a modality weight rule table can be pre-constructed, and modality weights corresponding to different local models can be determined based on the modality weight rule table. The modality weights reflect the relative importance of data of different modalities in recommendation. For example, if image data (such as promotional materials for new financial products) has an increased impact on user decisions in the recent market, modality weights for image modalities can be added to the modality weight rule table. Encoding model parameters as quantum state parameters can be achieved by converting model parameters into quantum state tensors and using quantum Fourier transform for dimensionality compression. Alternatively, a key can be generated through quantum key distribution, and the quantum state parameters can be encrypted using this key. The quantum state parameters can then be transmitted to a central node using a quantum teleportation protocol. The central node can recover the model parameters through decoding operations and then aggregate these model parameters according to the modality weights to obtain a federated learning model. The above methods can improve the aggregation efficiency by more than 40%.

[0068] In summary, by using modal weights and quantum teleportation, the aggregation efficiency of the federated learning model is improved, thereby enhancing the timeliness of financial product recommendation information.

[0069] To improve the prediction efficiency of financial product recommendation information, optionally, the preset time series model is a long short-term memory neural network model. Inputting input data into the preset time series model and outputting financial product recommendation information corresponding to the target user includes: converting the input data into quantum state input data and converting the gating unit of the long short-term memory neural network model into quantum gate operations; using quantum gate operations for parallel computation to predict the quantum state probability amplitude corresponding to the quantum state input data; and determining the financial product recommendation information corresponding to the target user based on the quantum state probability amplitude.

[0070] For example, quantum mapping can be used to map input data onto the ground state of a quantum system, obtaining quantum state input data, thus laying the data foundation for parallel computing. The gating units and node update functions of a Long Short-Term Memory (LSTM) neural network model can be converted into quantum gate operations. The parallelism of these gate operations enables simultaneous computation across multiple time steps. Furthermore, quantum entanglement can be used to handle time-series dependencies, reducing the computational complexity of prediction from O(n) to O(log n). The prediction results from parallel computation using quantum gate operations are output as quantum state probability amplitudes. By measuring these amplitudes, financial product recommendation information (e.g., target financial products and their recommendation probability values) can be obtained. This method can reduce the time required for a single prediction of 1000-dimensional features from 120ms using traditional probability calculations to 8ms.

[0071] In summary, quantum parallel computing was used to obtain the quantum state probability amplitude of financial products that match the input data and the probability of recommending financial products, thereby improving the prediction efficiency of financial product recommendation information.

[0072] To improve the timeliness of financial product recommendation information, optionally, the financial product recommendation information includes introductory information of multiple target financial products and a recommendation index for each target financial product. Presenting the financial product recommendation information to target users using augmented reality technology includes: retrieving each target financial product in a target product recommendation knowledge graph to determine the target entities corresponding to the target financial products and the target relationships between these entities; determining the information transmission chain of the target financial products based on the target entities and target relationships; and using augmented reality technology to present the information transmission chain to target users in animation form, setting different display labels for the information transmission chains of different target financial products based on the recommendation index.

[0073] Optionally, augmented reality technology can be combined with a target product recommendation knowledge graph to display financial product recommendation information in an intuitive and interactive way. The dynamically updated target product recommendation knowledge graph performs a deep search on each target financial product in the recommendation information, identifying entities associated with that target financial product. The relationships between these entities (e.g., introductory information entities, user entities, news event entities, yield entities, and the relationship between the target financial product and these entities) form an information transmission chain. This information transmission chain reflects all relevant information about the target financial product in the knowledge graph. The augmented reality interface can intuitively display key information such as the introductory information and yield of the target financial product through visual elements such as color, shape, and animation effects. In the information transmission chain, each entity and relationship is presented as a node and edge. The recommendation index, i.e., the recommendation probability of the target financial product in the financial product recommendation information center, can be displayed by adjusting the size, brightness, or edge thickness of nodes, the flashing frequency of the information transmission chain, and the background color of the information transmission chain, allowing users to quickly see the comparison of recommendation probabilities of different target financial products. It can also collect user operation data in the augmented reality interface (e.g., the operation of the timeline sliding control) and use the operation data to trigger a query of the target product recommendation knowledge graph, and display the query results in the augmented reality interface.

[0074] In summary, augmented reality technology not only makes the presentation of financial product recommendations more vivid and intuitive, but also enables users to quickly understand the introduction and recommendation index of the target financial product through animated information transmission chains. This facilitates target users to quickly understand the details of the financial product recommendations, shortens the viewing time, and thus improves the timeliness of the recommendations.

[0075] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0076] Example 2

[0077] This application also provides a financial product recommendation device based on multi-source heterogeneous data. It should be noted that this device can be used to execute the financial product recommendation method based on multi-source heterogeneous data provided in this application. The following describes the financial product recommendation device based on multi-source heterogeneous data provided in this application.

[0078] According to embodiments of this application, an apparatus for implementing the above-described financial product recommendation method based on multi-source heterogeneous data is also provided, such as... Figure 3 As shown, the device includes:

[0079] The data acquisition unit 301 is used to acquire multi-source heterogeneous data of the target user from multiple data sources, provided that the target user of the target financial institution has been authorized.

[0080] The relation extraction unit 302 is used to obtain the association features between multi-source heterogeneous data based on the target product recommendation knowledge graph, wherein the target product recommendation knowledge graph is obtained by updating the product recommendation knowledge graph based on multiple data sources.

[0081] The modality extraction unit 303 is used to obtain multimodal data based on multi-source heterogeneous data and to obtain the multimodal features corresponding to the multimodal data using a federated learning model.

[0082] The recommendation prediction unit 304 is used to determine the input data based on the correlation characteristics and multimodal characteristics, and input the input data into the preset time series model to output the financial product recommendation information corresponding to the target user.

[0083] Information display unit 305 is used to display financial product recommendation information to target users using augmented reality technology.

[0084] The financial product recommendation device based on multi-source heterogeneous data provided in this application embodiment acquires multi-source heterogeneous data of the target user from multiple data sources through a data acquisition unit 301, with the authorization of the target user from the target financial institution. A relation extraction unit 302 extracts the association features between the multi-source heterogeneous data based on a target product recommendation knowledge graph, which is obtained by updating the product recommendation knowledge graph based on multiple data sources. A modality extraction unit 303 acquires multimodal data based on the multi-source heterogeneous data and uses a federated learning model to obtain the corresponding multimodal features. A recommendation prediction unit 304 determines the input data based on the association features and multimodal features, inputs the input data into a preset time series model, and outputs financial product recommendation information corresponding to the target user. An information display unit 305 uses augmented reality technology to display the financial product recommendation information to the target user, thus solving the problem of poor timeliness of financial product recommendations in related technologies. This improves the timeliness of financial product recommendation results.

[0085] Optionally, in the financial product recommendation device based on multi-source heterogeneous data provided in this application embodiment, the multiple data sources include at least user identity information, user financial information, and news event information. The relationship extraction unit includes: a change detection module, used to continuously detect data changes in multiple data sources; a change extraction module, used to extract new data from the data source with data changes when any data source changes, using an information extraction algorithm to determine the entity extraction result and the relationship extraction result; and a graph update module, used to update the product recommendation knowledge graph based on the entity extraction result and the relationship extraction result.

[0086] Optionally, in the financial product recommendation device based on multi-source heterogeneous data provided in this application embodiment, the graph update module includes: a first update submodule, used to map entity extraction results and relation extraction results to quantum superposition state data; a second update submodule, used to traverse all paths in the product recommendation knowledge graph in parallel according to a quantum search algorithm to determine the update target; and a third update submodule, used to update the product recommendation knowledge graph according to the update target and the quantum superposition state data.

[0087] Optionally, in the financial product recommendation device based on multi-source heterogeneous data provided in this application embodiment, the multimodal data includes data of multiple different modalities. The modality extraction unit includes: a collaborative training module, used to collaboratively train the training set of the modality corresponding to the participant using a preset federated learning framework to obtain multiple local models; a model aggregation module, used to aggregate the multiple local models to determine the federated learning model; and a modality feature module, used to input the multimodal data into the federated learning model to obtain multimodal features.

[0088] Optionally, in the financial product recommendation device based on multi-source heterogeneous data provided in this application embodiment, multiple local models are aggregated. The model aggregation module includes: a weight determination submodule, used to determine the modal weights corresponding to each local model based on the modality corresponding to each local model; a parameter encoding submodule, used to encode the model parameters in each local model into parameters of a quantum state; and a quantum aggregation submodule, used to aggregate the parameters of the quantum state in a quantum teleportation manner based on the modal weights to obtain a federated learning model.

[0089] Optionally, in the financial product recommendation device based on multi-source heterogeneous data provided in this application embodiment, the preset time series model is a long short-term memory neural network model, and the recommendation prediction unit includes: an input conversion module, used to convert input data into quantum state input data and convert the gating unit of the long short-term memory neural network model into quantum gate operations; a parallel computing module, used to perform parallel computing using quantum gate operations to predict the quantum state probability amplitude corresponding to the quantum state input data; and a probability amplitude measurement module, used to determine the financial product recommendation information corresponding to the target user based on the quantum state probability amplitude.

[0090] Optionally, in the financial product recommendation device based on multi-source heterogeneous data provided in this application embodiment, the financial product recommendation information includes introduction information of multiple target financial products and recommendation index of each target financial product. The information display unit includes: a target retrieval module, used to retrieve each target financial product in the target product recommendation knowledge graph, determine the target entity corresponding to the target financial product and the target association relationship between the target entities; a transmission chain determination module, used to determine the information transmission chain of the target financial product based on the target entity and the target association relationship; and an augmented reality module, used to use augmented reality technology to display the information transmission chain to the target user in the form of animation, and set different display labels for the information transmission chains of different target financial products based on the recommendation index.

[0091] It should be noted that the data acquisition unit 301, relation extraction unit 302, modality extraction unit 303, recommendation prediction unit 304, and information display unit 305 mentioned above correspond to steps S201 to S205 in Embodiment 1. The instances and application scenarios implemented by the units and corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1. It should be noted that the above modules or units can be hardware or software components stored in memory (e.g., memory 104) and processed by one or more processors (e.g., processors 102a, 102b, ..., 102n). The above modules can also be part of a device and run in the computer terminal 10 provided in Embodiment 1.

[0092] Example 3

[0093] Embodiments of this application may provide an electronic device. Figure 4 This is a structural block diagram of an electronic device according to an embodiment of this application. Figure 4 As shown, the electronic device may include: one or more ( Figure 4 (Only one is shown) processor 1002, memory 1004, memory controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0094] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the methods and apparatus in the embodiments of this application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby implementing the above-described methods. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0095] The processor can access information and applications stored in memory via a transmission device to execute the following steps: With authorization from the target user of the target financial institution, acquire multi-source heterogeneous data of the target user from multiple data sources; obtain the correlation features between the multi-source heterogeneous data based on the target product recommendation knowledge graph, wherein the target product recommendation knowledge graph is obtained by updating the product recommendation knowledge graph based on multiple data sources; acquire multimodal data based on the multi-source heterogeneous data, and use a federated learning model to obtain the multimodal features corresponding to the multimodal data; determine the input data based on the correlation features and multimodal features, and input the input data into a preset time series model to output financial product recommendation information corresponding to the target user; and use augmented reality technology to display the financial product recommendation information to the target user.

[0096] The processor can also call the information and application programs stored in the memory through the transmission device to perform the following steps: continuously detect data changes from multiple data sources; when any data source changes, use an information extraction algorithm to extract the new data from the data source that has changed, and determine the entity extraction result and the relation extraction result; update the product recommendation knowledge graph based on the entity extraction result and the relation extraction result.

[0097] The processor can also call the information and application programs stored in the memory through the transmission device to perform the following steps: mapping the entity extraction results and relation extraction results to quantum superposition state data; traversing all paths in the product recommendation knowledge graph in parallel according to the quantum search algorithm to determine the update target; and updating the product recommendation knowledge graph according to the update target and the quantum superposition state data.

[0098] The processor can also call the information and application stored in the memory through the transmission device to perform the following steps: multiple participants use a preset federated learning framework to collaboratively train the training set of the modality corresponding to each participant to obtain multiple local models; aggregate the multiple local models to determine the federated learning model; input the multimodal data into the federated learning model to obtain multimodal features.

[0099] The processor can also call the information and application programs stored in the memory through the transmission device to perform the following steps: determine the mode weights corresponding to the local models based on the modes corresponding to each local model; encode the model parameters in each local model into quantum state parameters; and aggregate the quantum state parameters according to the mode weights to obtain a federated learning model.

[0100] The processor can also access information and applications stored in the memory via a transmission device to perform the following steps: converting input data into quantum state input data and converting the gating units of the long short-term memory neural network model into quantum gate operations; performing parallel computation using quantum gate operations to predict the quantum state probability amplitude corresponding to the quantum state input data; and determining the financial product recommendation information corresponding to the target user based on the quantum state probability amplitude.

[0101] The processor can also call the information and application programs stored in the memory through the transmission device to perform the following steps: retrieval of each target financial product in the target product recommendation knowledge graph to determine the target entity corresponding to the target financial product and the target association between the target entities; determination of the information transmission chain of the target financial product based on the target entity and the target association; display of the information transmission chain to the target user in the form of animation using augmented reality technology, and setting different display labels for the information transmission chains of different target financial products based on the recommendation index.

[0102] This application provides a solution for recommending financial products based on multi-source heterogeneous data. With the authorization of the target user from the target financial institution, multi-source heterogeneous data of the target user is obtained from multiple data sources. The relationship features between the multi-source heterogeneous data are obtained based on a target product recommendation knowledge graph, which is obtained by updating the product recommendation knowledge graph based on multiple data sources. Multimodal data is obtained from the multi-source heterogeneous data, and multimodal features corresponding to the multimodal data are obtained using a federated learning model. Input data is determined based on the relationship features and multimodal features, and the input data is input into a preset time series model to output financial product recommendation information corresponding to the target user. Augmented reality technology is used to display the financial product recommendation information to the target user, thus solving the technical problem of poor timeliness in financial product recommendations in related technologies.

[0103] Those skilled in the art will understand that Figure 4 The structure shown is for illustrative purposes only. Electronic devices can also be smartphones, tablets, handheld computers, mobile internet devices (MIDs), PADs, and other terminal devices. Figure 4 This does not limit the structure of the aforementioned electronic device. For example, electronic devices may also include components that are more... Figure 4 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 4 The different configurations shown.

[0104] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0105] Example 4

[0106] Embodiments of this application also provide a storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the financial product recommendation method based on multi-source heterogeneous data provided in Embodiment 1.

[0107] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0108] Optionally, in this embodiment, the computer-readable storage medium is configured to store program code for performing the following steps: obtaining multi-source heterogeneous data of the target user from multiple data sources after obtaining authorization from the target user of the target financial institution; obtaining the correlation features between the multi-source heterogeneous data based on the target product recommendation knowledge graph, wherein the target product recommendation knowledge graph is obtained by updating the product recommendation knowledge graph based on multiple data sources; obtaining multimodal data based on the multi-source heterogeneous data, and using a federated learning model to obtain the multimodal features corresponding to the multimodal data; determining input data based on the correlation features and multimodal features, and inputting the input data into a preset time series model to output financial product recommendation information corresponding to the target user; and displaying the financial product recommendation information to the target user using augmented reality technology.

[0109] Optionally, in this embodiment, the computer-readable storage medium is further configured to store program code for performing the following steps: continuously detecting data changes from multiple data sources; when any data source changes, using an information extraction algorithm to extract new data from the data source where the data change occurred, and determining entity extraction results and relation extraction results; updating the product recommendation knowledge graph based on the entity extraction results and relation extraction results.

[0110] Optionally, in this embodiment, the computer-readable storage medium is further configured to store program code for performing the following steps: mapping entity extraction results and relation extraction results to quantum superposition state data; traversing all paths in the product recommendation knowledge graph in parallel according to a quantum search algorithm to determine the update target; and updating the product recommendation knowledge graph according to the update target and the quantum superposition state data.

[0111] Optionally, in this embodiment, the computer-readable storage medium is further configured to store program code for performing the following steps: multiple participants co-train the training set of the modality corresponding to each participant using a preset federated learning framework to obtain multiple local models; aggregate the multiple local models to determine the federated learning model; and input multimodal data into the federated learning model to obtain multimodal features.

[0112] Optionally, in this embodiment, the computer-readable storage medium is further configured to store program code for performing the following steps: determining the mode weights corresponding to each local model based on the mode corresponding to each local model; encoding the model parameters in each local model into parameters of the quantum state; and aggregating the parameters of the quantum state according to the mode weights using quantum teleportation methods to obtain a federated learning model.

[0113] Optionally, in this embodiment, the computer-readable storage medium is further configured to store program code for performing the following steps: converting input data into quantum state input data and converting the gating unit of the long short-term memory neural network model into quantum gate operations; performing parallel computation using quantum gate operations to predict the quantum state probability amplitude corresponding to the quantum state input data; and determining the financial product recommendation information corresponding to the target user based on the quantum state probability amplitude.

[0114] Optionally, in this embodiment, the computer-readable storage medium is further configured to store program code for performing the following steps: retrieving each target financial product in the target product recommendation knowledge graph to determine the target entity corresponding to the target financial product and the target association between the target entities; determining the information transmission chain of the target financial product based on the target entity and the target association; displaying the information transmission chain to the target user in the form of animation using augmented reality technology, and setting different display labels for the information transmission chains of different target financial products based on the recommendation index.

[0115] This application also provides a computer program product that, when executed on a data processing device, is suitable for performing steps of a financial product recommendation method based on multi-source heterogeneous data.

[0116] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0117] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0118] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0119] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0120] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0121] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0122] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for recommending financial products based on multi-source heterogeneous data, characterized in that, include: With authorization from the target financial institution's target user, multi-source heterogeneous data of the target user is obtained from multiple data sources; The association features between the multi-source heterogeneous data are obtained based on the target product recommendation knowledge graph, wherein the target product recommendation knowledge graph is obtained by updating the product recommendation knowledge graph based on the multiple data sources; Multimodal data is obtained based on the multi-source heterogeneous data, and multimodal features corresponding to the multimodal data are obtained using a federated learning model; Based on the correlation features and the multimodal features, the input data is determined and input into a preset time series model to output financial product recommendation information corresponding to the target user; Augmented reality technology is used to display the financial product recommendation information to the target user.

2. The method according to claim 1, characterized in that, The multiple data sources include at least user identity information, user financial information, and news event information. Updating the product recommendation knowledge graph based on the multiple data sources includes: Continuously monitor data changes from the multiple data sources; When any of the data sources changes, an information extraction algorithm is used to extract the new data from the data source that has changed, and to determine the entity extraction result and the relation extraction result. The product recommendation knowledge graph is updated based on the entity extraction results and the relationship extraction results.

3. The method according to claim 2, characterized in that, The product recommendation knowledge graph is updated based on the entity extraction results and the relation extraction results, including: The entity extraction results and the relation extraction results are mapped to quantum superposition state data; The update target is determined by traversing all paths in the product recommendation knowledge graph in parallel using a quantum search algorithm. The product recommendation knowledge graph is updated based on the update target and the quantum superposition state data.

4. The method according to claim 1, characterized in that, The multimodal data includes data from multiple different modalities. The multimodal features corresponding to the multimodal data obtained using a federated learning model include: Multiple local models are obtained by using a pre-defined federated learning framework to collaboratively train the training set of the modality corresponding to each participant. The federated learning model is determined by aggregating multiple local models. The multimodal data is input into the federated learning model to obtain the multimodal features.

5. The method according to claim 4, characterized in that, The federated learning model is determined by aggregating multiple local models, including: Based on the mode corresponding to each local model, determine the mode weights corresponding to the local model; The model parameters in each of the local models are encoded as parameters of the quantum state; The parameters of the quantum state are aggregated according to the modal weights to obtain the federated learning model.

6. The method according to claim 1, characterized in that, The preset time series model is a long short-term memory neural network model. Inputting the input data into the preset time series model and outputting financial product recommendation information corresponding to the target user includes: The input data is converted into quantum state input data, and the gating unit of the long short-term memory neural network model is converted into a quantum gate operation; Parallel computation is performed using the quantum gate operation to predict the quantum state probability amplitude corresponding to the input data of the quantum state; The financial product recommendation information corresponding to the target user is determined based on the quantum state probability amplitude.

7. The method according to claim 1, characterized in that, The financial product recommendation information includes introductory information for multiple target financial products and a recommendation index for each target financial product. Displaying the financial product recommendation information to the target user using augmented reality technology includes: Each target financial product is retrieved from the target product recommendation knowledge graph to determine the target entity corresponding to the target financial product and the target association relationship between the target entities; Based on the target entity and the target relationship, determine the information transmission chain of the target financial product; The information transmission chain is presented to the target user in the form of animation using the augmented reality technology, and different display labels are set for the information transmission chains of different target financial products based on the recommendation index.

8. A financial product recommendation device based on multi-source heterogeneous data, characterized in that, include: The data acquisition unit is used to acquire multi-source heterogeneous data of the target user from multiple data sources, provided that the target user of the target financial institution has been authorized. The relation extraction unit is used to obtain the association relationship features between the multi-source heterogeneous data based on the target product recommendation knowledge graph, wherein the target product recommendation knowledge graph is obtained by updating the product recommendation knowledge graph based on the multiple data sources; The modality extraction unit is used to obtain multimodal data based on the multi-source heterogeneous data, and to obtain the multimodal features corresponding to the multimodal data using a federated learning model. The recommendation prediction unit is used to determine input data based on the correlation features and the multimodal features, input the input data into a preset time series model, and output financial product recommendation information corresponding to the target user; The information display unit is used to display the financial product recommendation information to the target user using augmented reality technology.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored executable program, wherein, when the executable program is executed, it controls the device on which the computer-readable storage medium is located to perform the financial product recommendation method based on multi-source heterogeneous data as described in any one of claims 1 to 7.

10. An electronic device, characterized in that, include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include methods for performing the financial product recommendation method based on multi-source heterogeneous data as described in any one of claims 1 to 7.

11. A computer program product comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the steps of the financial product recommendation method based on multi-source heterogeneous data as described in any one of claims 1 to 7.

Images