Financial product determination method and device, electronic equipment and computer program product

Abstract

The invention discloses a financial product determination method and device, electronic equipment and a computer program product. Relates to the field of financial science and technology, and the method comprises the steps: obtaining user information of a target user under the condition of authorization of the target user, obtaining M candidate financial products and product data of each candidate financial product from a data lake, and obtaining M groups of product data; the M groups of product data and the user information are input into a product prediction model, M product output vectors are obtained through processing, and the product prediction model is used for constructing matching vectors of M candidate financial products and target users according to the M groups of product data and the user information, generating M product output vectors by using the matching vectors of the M candidate financial products and the target user; and screening the M candidate financial products according to the M product output vectors to obtain a target financial product. Through the method and the device, the technical problem that the screening accuracy is low when the financial product matched with the target user is screened in the prior art is solved.

Description

Technical Field

[0001] This application relates to the field of financial technology, and more specifically, to a method, apparatus, electronic device, and computer program product for determining a financial product. Background Technology

[0002] With the development of financial technology, the supply side of financial products has ushered in an unprecedented boom. In order to meet the increasingly diversified needs of users, financial products and services are constantly being innovated. However, there is a serious problem of product homogenization. Coupled with the trend of personalized user needs, it is difficult for users to choose the most suitable financial products and services when faced with a wide variety of financial products.

[0003] To address these issues, related technologies utilize rule-based engine algorithms to achieve a basic match between products and users to some extent, or employ collaborative filtering algorithms to capture subtle, hard-to-define intrinsic relationships within user groups, revealing implicit user needs for financial institutions, or utilize machine learning models to process nonlinear data. These methods are more flexible than traditional statistical methods and can adapt to more complex user demand scenarios.

[0004] While the aforementioned technologies have achieved significant results in improving screening performance, they still suffer from the following key drawbacks: First, relying on methods such as collaborative filtering easily overlooks the multidimensional nature of user needs, leading to a lack of targeted screening results and a tendency to apply a one-size-fits-all approach, failing to fully reflect the fine-grained matching between users' personalized needs and product attributes. Second, these technologies often adopt a static perspective on historical interactions, while user activity within financial institutions carries temporal characteristics, resulting in incomplete capture of these temporal dependencies and limiting the speed and accuracy of response to changes in user needs. Finally, when processing large-scale heterogeneous data, these technologies struggle to establish efficient multi-dimensional feature fusion mechanisms, thus impacting user experience.

[0005] There is currently no effective solution to the technical problem of low accuracy in screening financial products that match target users in related technologies. Summary of the Invention

[0006] The main objective of this application is to provide a method, apparatus, electronic device, and computer program product for determining financial products, so as to solve the technical problem of low screening accuracy when screening financial products that match target users in related technologies.

[0007] To achieve the above objectives, according to one aspect of this application, a method for determining financial products is provided. The method includes: obtaining user information of the target user with the user's authorization; obtaining M candidate financial products and product data for each candidate financial product from a data lake, resulting in M ​​sets of product data, where M is a positive integer; inputting the M sets of product data and user information into a product prediction model, processing to obtain M product output vectors, where each product output vector represents the degree of matching between each candidate financial product and the target user; the product prediction model constructs matching vectors between the M candidate financial products and the target user based on the M sets of product data and user information, and generates M product output vectors using these matching vectors; and selecting the target financial product from the M candidate financial products based on the M product output vectors.

[0008] Optionally, before inputting the M sets of product data and user information into the product prediction model, the method further includes: extracting static information of the target user from the user information, performing vector transformation on the static information to obtain a first vector; extracting behavioral information of the target user from the user information, converting the behavioral information into a behavioral sequence, and performing vector transformation on the behavioral sequence to obtain a second vector; and concatenating the first vector and the second vector to obtain a user feature vector.

[0009] Optionally, before inputting the M sets of product data and user information into the product prediction model, the method further includes: for a set of product data of a candidate financial product, extracting discrete data of the candidate financial product from the set of product data, performing vector transformation on the discrete data to obtain a third vector; extracting continuous data of the candidate financial product from the set of product data, performing vector mapping on the continuous data to obtain a fourth vector, and concatenating the third vector and the fourth vector to obtain the product feature vector of the candidate financial product.

[0010] Optionally, the product prediction model includes an input module, N attention modules, and an output module, where N is a positive integer. M sets of product data and user information are input into the product prediction model, and the process yields M product output vectors. This includes: for a candidate financial product, the input module obtains M product feature vectors corresponding to a set of product data for the candidate financial product, and obtains user feature vectors corresponding to the user information; the N attention modules extract feature vectors from the product feature vectors and user feature vectors respectively, and each attention module calculates N matching vectors based on the extracted feature vectors; the output module processes the N matching vectors to obtain the product output vector of the candidate financial product.

[0011] Optionally, when there are N attention modules, including a first attention module, a second attention module, and a third attention module, the N attention modules respectively extract feature vectors from the product feature vector and the user feature vector, and each attention module calculates N matching vectors based on the extracted feature vectors, including: the first attention module extracts a product risk feature vector from the product feature vector and a user risk feature vector from the user feature vector, uses the product risk feature vector as a key-value vector and the user risk feature vector as a query vector, and calculates a risk fit vector based on the key-value vector and the query vector; the second attention module extracts a user behavior feature vector from the user feature vector, uses the user behavior feature vector as a query vector and the product feature vector as a key-value vector, and calculates a behavior association vector based on the key-value vector and the query vector; obtains a state vector, and the third attention module calculates an aggregation vector based on the state vector and the product feature vector; the risk fit vector, the behavior association vector, and the aggregation vector constitute N matching vectors.

[0012] Optionally, the output module processes the N matching vectors to obtain the product output vector of the candidate financial product, including: extracting the risk fit vector, behavior correlation vector, and aggregation vector from the N matching vectors; obtaining a preset function, and concatenating the risk fit vector, behavior correlation vector, and aggregation vector based on the preset function to obtain the fused feature vector; obtaining a preset projection matrix, and calculating the dot product of the fused feature vector and the preset projection matrix to obtain the product output vector.

[0013] Optionally, selecting the target financial product from M candidate financial products based on M product output vectors includes: normalizing each product output vector to obtain M product predicted scores; sorting the M product predicted scores in descending order to obtain a score sequence, and obtaining the product preset score of the preset position from the score sequence; and determining the candidate financial product associated with the product preset score of the preset position as the target financial product.

[0014] To achieve the above objectives, according to another aspect of this application, a financial product determination apparatus is provided. The apparatus includes: an acquisition unit, configured to acquire user information of a target user with authorization, and acquire M candidate financial products and product data for each candidate financial product from a data lake, resulting in M ​​sets of product data, where M is a positive integer; an input unit, configured to input the M sets of product data and user information into a product prediction model, processing it to obtain M product output vectors, wherein each product output vector represents the degree of matching between each candidate financial product and the target user, the product prediction model constructs matching vectors between the M candidate financial products and the target user based on the M sets of product data and user information, and generates M product output vectors using the matching vectors between the M candidate financial products and the target user; and a filtering unit, configured to filter the target financial product from the M candidate financial products based on the M product output vectors.

[0015] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is running, the device where the computer-readable storage medium is located executes any of the above-mentioned financial products.

[0016] According to another aspect of the present invention, an electronic device is also provided, including one or more processors and a memory, the memory storing an executable program, and the processor for running the program, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method for determining any of the above-described financial products.

[0017] According to another aspect of the present invention, a computer program product is also provided, the computer program product including a computer program, wherein when the computer program is executed by a processor, it implements the method for determining any of the above-mentioned financial products.

[0018] In this embodiment, a method for determining financial products is adopted. With the authorization of the target user, the user information of the target user is obtained, and M candidate financial products and product data for each candidate financial product are retrieved from a data lake, resulting in M ​​sets of product data, where M is a positive integer. The M sets of product data and user information are input into a product prediction model, which processes them to obtain M product output vectors. Each product output vector represents the degree of matching between each candidate financial product and the target user. The product prediction model constructs matching vectors between the M candidate financial products and the target user based on the M sets of product data and user information, and generates M product output vectors using these matching vectors. The target financial product is then selected from the M candidate financial products based on the M product output vectors. This solves the technical problem of low selection accuracy when selecting financial products matching the target user in related technologies. By inputting the product data and user information of multiple candidate financial products into the product prediction model, processing them to obtain corresponding product output vectors, and using these product output vectors to select the target financial product from the candidate financial products, the technical effect of improving the selection accuracy of financial products matching the target user is achieved. Attached Figure Description

[0019] 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:

[0020] Figure 1 It is a hardware structure block diagram of a computer terminal (or mobile device) used to implement a method for determining financial products;

[0021] Figure 2 This is a flowchart of a method for determining financial products according to embodiments of this application;

[0022] Figure 3 This is a schematic diagram of an optional financial product determination method provided according to an embodiment of this application;

[0023] Figure 4 This is a schematic diagram of a financial product determination device provided according to an embodiment of this application;

[0024] Figure 5 This is a structural block diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0025] 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.

[0026] 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.

[0027] It should be noted that all information (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.) involved in this application are information and data authorized by the user or fully authorized by all parties. For example, this system has interfaces with relevant users or organizations to provide users with corresponding operation data for them to choose to agree to or refuse automated decision-making results. Before obtaining relevant information, a request for obtaining the information needs to be sent to the aforementioned user or organization through the interface, and the relevant information is obtained after receiving consent from the aforementioned user or organization; if the user chooses to refuse, the expert decision-making process is initiated. Users can view the purpose of data use in real time through authorization decoding and have the right to withdraw authorization or delete data at any time. After the authorization is withdrawn, the system will terminate the relevant data processing within 24 hours.

[0028] It should be noted that the information collected in this application 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 the relevant data all comply with the relevant laws, regulations and standards of the relevant regions, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation access points for users to choose to authorize use or refuse use.

[0029] Example 1

[0030] According to an embodiment of this application, a method embodiment for determining a financial product 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.

[0031] 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 This is a hardware structure block diagram of a computer terminal (or mobile device) used to implement a method for determining financial products, such as... Figure 1 As shown, computer terminal 10 (or mobile device) may include one or more ( Figure 1 The processor 102 (which may include, but is not limited to, a microprocessor MCU (Microcontroller Unit) or a programmable gate array (FPGA)) is shown as 102a, 102b, ..., 102n. It also includes a memory 104 for storing data and a transmission device 106 for communication functions. In addition, it may include: a display, an input / output interface, a Universal Serial Bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a keyboard, a cursor control device, 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.

[0032] 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).

[0033] 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 determination method in this embodiment. 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 determination method. 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.

[0034] 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) and a network interface, which can be connected 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.

[0035] The display can be, for example, a touchscreen liquid crystal display (LCD), which allows the user to interact with the user interface of the computer terminal 10 (or mobile device).

[0036] Under the aforementioned operating environment, this application provides the following: Figure 2 The method for determining the financial products shown. Figure 2 This is a flowchart of a method for determining financial products according to embodiments of this application, such as... Figure 2 As shown, the method includes the following steps:

[0037] Step S201: With the authorization of the target user, obtain the user information of the target user, and obtain M candidate financial products and product data of each candidate financial product from the data lake to obtain M sets of product data, where M is a positive integer.

[0038] It's important to note that, to provide more precise financial products to target users, financial institutions need to obtain authorization from these users during the initial data collection phase to use and analyze their personal information and transaction data. After user authorization, they can obtain user information, such as static and dynamic data related to the target user, including basic personal information, financial status, risk preferences, and user behavior logs. Simultaneously, they can also obtain multiple candidate financial products and product data for each candidate product from the financial institution's data lake. A data lake is an environment for storing massive amounts of diverse data, capable of storing structured and unstructured data in its native format, providing data flexibility and accessibility. This data lake can include multi-source data such as user information, product data, and user transaction and behavior logs. Candidate financial products can include multiple categories such as wealth management products and funds to meet the needs of users at different stages and in different financial situations.

[0039] Furthermore, for each candidate financial product, product data can include its static attributes (such as product type) and dynamic performance (such as volatility index). By collecting and organizing user information and product data, a comprehensive user profile and product feature library can be established, laying a solid foundation for feature extraction and matching in subsequent multi-head attention mechanisms.

[0040] Step S202: Input M sets of product data and user information into the product prediction model, process them to obtain M product output vectors, where each product output vector is used to characterize the matching degree between each candidate financial product and the target user. The product prediction model is used to construct M matching vectors between candidate financial products and target users based on M sets of product data and user information, and use the M matching vectors between candidate financial products and target users to generate M product output vectors.

[0041] Specifically, after obtaining product data and user information, a product prediction model based on a multi-head attention mechanism can be used to process the aforementioned static and dynamic characteristics of users and static attributes of products. Multiple attention heads focus on the matching of user preferences with product risk levels, the correlation between recent user behavior sequences and product types, and the fit between product attributes and the environment. Through parallel computing and feature fusion, the model comprehensively considers the diversified needs of users and product characteristics, evaluates product attributes from multiple perspectives, and thus uncovers the deep matching relationship between users and candidate financial products, i.e., outputs the corresponding product output vector. The product output vector is a quantitative representation of the degree of matching between each candidate financial product and the target user. Each vector, through parallel processing of the multi-head attention mechanism, not only reflects the adaptability of products and users in multiple dimensions such as risk preferences, return expectations, and liquidity needs, but also considers temporal behavior, enhancing the personalization and timeliness of the screening.

[0042] Step S203: Select the target financial product from the M candidate financial products based on the M product output vectors.

[0043] Specifically, after obtaining the product output vector of each candidate financial product, these product output vectors can be used to screen the financial products. That is, the product output vectors can be normalized and transformed into easily comparable predicted interaction scores (such as purchase probabilities). The candidate products are then ranked according to these scores, and the Top N products (that is, the top N products in the queue) are finally selected as the target financial products for the target users.

[0044] It should be noted that, in order to ensure that the screening results meet the requirements of financial institutions and avoid screening high-risk products for users with insufficient risk tolerance or products that exceed the financial capacity of users with insufficient funds, the target financial products initially screened can be verified a second time to eliminate products that do not match the user's risk level or whose minimum purchase amount exceeds the user's financial capacity. This makes the screening results both personalized and in line with the needs of financial institutions.

[0045] The method for determining financial products provided in this application involves obtaining user information of the target user with their authorization, and acquiring M candidate financial products and product data for each candidate financial product from a data lake, resulting in M ​​sets of product data, where M is a positive integer. The M sets of product data and user information are input into a product prediction model, which is then processed to obtain M product output vectors. Each product output vector represents the degree of matching between each candidate financial product and the target user. The product prediction model constructs matching vectors between the M candidate financial products and the target user based on the M sets of product data and user information, and generates M product output vectors using these matching vectors. The target financial product is then selected from the M candidate financial products based on the M product output vectors. This method solves the technical problem of low selection accuracy when selecting financial products matching the target user in related technologies. By inputting product data and user information of multiple candidate financial products into the product prediction model, processing them to obtain corresponding product output vectors, and then using these product output vectors to select the target financial product from the candidate financial products, the technical effect of improving the selection accuracy of financial products matching the target user is achieved.

[0046] Optionally, in the method for determining financial products provided in the embodiments of this application, before inputting M sets of product data and user information into the product prediction model, the method further includes: extracting static information of the target user from the user information, performing vector transformation on the static information to obtain a first vector; extracting behavioral information of the target user from the user information, converting the behavioral information into a behavioral sequence, and performing vector transformation on the behavioral sequence to obtain a second vector; and concatenating the first vector and the second vector to obtain a user feature vector.

[0047] To help product prediction models better analyze relevant data, product data and user information can be processed. Specifically, the static information portion can first be separated from the acquired user information. This static information can include the user's basic attributes and a snapshot of their financial situation, reflecting the user's personal characteristics and financial status. Then, the extracted static information is vectorized, mapping each static attribute to a numerical value or vector. Through pre-trained embedding layers or encoding rules, non-numerical attributes are converted into numerical representations, thereby constructing a 64-dimensional first vector.

[0048] Furthermore, recent behavioral information of the target user is extracted from the user information. For example, behavioral data can include the user's transaction and browsing events within the past 30 days, each event bearing a timestamp to indicate the order and time interval of the behavior. These behavioral events are then arranged chronologically to construct a user behavior sequence. Using timestamp encoding and embedded representation of behavioral events, the behavior sequence is converted into a 128-dimensional second vector. This second vector not only captures the user's behavioral patterns but also retains the time-series information of the behavior, helping the model identify dynamic changes in user needs and preferences. Finally, the transformed first vector (the vector of static user information) and the second vector (the vector of the user behavior sequence) are concatenated to form a 192-dimensional user feature vector. The concatenation operation directly links the dimensions of the two vectors, allowing static and dynamic behavioral features to be represented in the same vector, providing comprehensive user information support for subsequent product selection.

[0049] This embodiment constructs a user feature vector that comprehensively reflects users' static attributes and dynamic behaviors. This not only enables the model to accurately capture users' behavioral characteristics at different points in time, but also helps the model understand users' multi-dimensional needs such as risk preferences, return expectations, and liquidity requirements. As a result, it generates a more accurate list of financial products to screen, thereby improving the product screening level of financial institutions and customer satisfaction.

[0050] Optionally, in the method for determining financial products provided in the embodiments of this application, before inputting M sets of product data and user information into the product prediction model, the method further includes: for a set of product data of a candidate financial product, extracting discrete data of the candidate financial product from the set of product data, performing vector transformation on the discrete data to obtain a third vector; extracting continuous data of the candidate financial product from the set of product data, performing vector mapping processing on the continuous data to obtain a fourth vector, and concatenating the third vector and the fourth vector to obtain the product feature vector of the candidate financial product.

[0051] While processing user information, product data can also be processed. Specifically, firstly, discrete data of discrete attributes is identified and extracted from a set of product data for a candidate financial product. This discrete data can include product type, risk level, product term, etc., reflecting the classification characteristics of financial products. Then, the extracted discrete data is vectorized. Through predefined embedding layers or encoding rules, each discrete attribute is mapped to a low-dimensional dense vector, for example, encoded as a 64-dimensional third vector, where each dimension corresponds to a product type. By numerically representing the characteristics of product types, the model can understand the category information of product attributes.

[0052] Furthermore, continuous attributes, such as expected rate of return, fixed interest rate, and redemption fee rate, are extracted from the product data of the candidate financial product. These data are numerical and can accurately reflect the risk-return characteristics of the product. Then, data standardization or normalization methods are used to map the continuous data to a unified numerical range, such as the [0, 1] interval, to eliminate the influence of dimensions and improve the efficiency and accuracy of model processing. Then, through a fully connected layer in the neural network or a custom mapping function, the standardized continuous attributes are converted into numerical feature codes, forming a representation with the same dimension as the discrete data vector, that is, a 64-dimensional fourth vector. Finally, the third vector (the vector of discrete data) and the fourth vector (the vector of continuous data) are concatenated to integrate the multiple dimensions of product attributes into a 128-dimensional product feature vector.

[0053] This embodiment constructs a product feature vector for each candidate financial product, providing comprehensive and refined input for the subsequent multi-head attention mechanism. This enables the model to consider multi-dimensional information simultaneously, improving matching accuracy and personalization.

[0054] Optionally, in the method for determining financial products provided in this application embodiment, the product prediction model includes an input module, N attention modules, and an output module, where N is a positive integer. Inputting M sets of product data and user information into the product prediction model and processing them to obtain M product output vectors includes: for a candidate financial product, the input module obtains M product feature vectors corresponding to a set of product data for the candidate financial product, and obtains user feature vectors corresponding to the user information; the N attention modules extract feature vectors from the product feature vectors and user feature vectors respectively, and each attention module calculates N matching vectors based on the extracted feature vectors; the output module processes the N matching vectors to obtain the product output vector of the candidate financial product.

[0055] After obtaining the vectors corresponding to user information and product data, these vectors can be input into the product prediction model. The product prediction model can receive data through the input module, perform feature extraction and matching through the attention module, and generate the final score through the output module, producing a product output vector for each candidate financial product. Specifically, the input module first obtains the pre-processed and transformed product feature vector corresponding to each candidate financial product, and simultaneously obtains the user feature vector from the target user, and inputs it into the attention module for processing.

[0056] Furthermore, the core component of the product prediction model—the attention module—can be used to process the aforementioned feature vectors. Each attention module focuses on a specific dimension in the model input, such as the matching between user risk preferences and product risk levels, or the relationship between user behavior sequences and product types. By focusing on specific subsets or aspects of the product and user feature vectors, and calculating attention weights, the input feature vectors are weighted and summed to generate multiple matching vectors. Each matching vector reflects the degree of matching between the user and the product in that dimension.

[0057] After the attention module finishes processing, the output module receives all matching vectors and performs comprehensive processing on them. The vectors are merged into a composite vector through vector concatenation, and then mapped through a fully connected layer or output projection matrix to generate a product output vector used to measure the overall fit between candidate financial products and user needs.

[0058] This embodiment utilizes the multi-head attention mechanism of the product prediction model to analyze and match the multi-dimensional characteristics of users and financial products, gain a deep understanding of the details of user needs and the characteristics of products, and then outputs product output vectors. Based on these vectors, personalized target financial products are generated, which significantly improves the service quality and competitiveness of financial institutions.

[0059] Optionally, in the method for determining financial products provided in this application embodiment, when the N attention modules include a first attention module, a second attention module, and a third attention module, the N attention modules respectively extract feature vectors from the product feature vector and the user feature vector, and each attention module calculates N matching vectors based on the extracted feature vectors, including: the first attention module extracts a product risk feature vector from the product feature vector and a user risk feature vector from the user feature vector, uses the product risk feature vector as a key-value vector and the user risk feature vector as a query vector, and calculates a risk fit vector based on the key-value vector and the query vector; the second attention module extracts a user behavior feature vector from the user feature vector, uses the user behavior feature vector as a query vector and the product feature vector as a key-value vector, and calculates a behavior association vector based on the key-value vector and the query vector; obtains a state vector, and the third attention module calculates an aggregation vector based on the state vector and the product feature vector; the risk fit vector, the behavior association vector, and the aggregation vector constitute N matching vectors.

[0060] It should be noted that by designing three attention modules for the product prediction model, each focusing on multiple key matching dimensions, and then using parallel computing and feature fusion, the deep matching relationship between users and products is mined. Specifically, during processing by the attention modules, the first attention module first extracts a risk-related sub-vector from the product feature vector, namely the product risk feature vector p_risk, which can include dimensions such as product risk level and historical volatility. Then, it extracts the user's risk preference and its quantitative description of risk tolerance from the user feature vector, forming the user risk feature vector u_risk. Then, the first attention module uses p_risk as a key-value vector (i.e., key vector and value vector) and u_risk as a query vector. By calculating the attention weight between the query vector and the key vector, the user's preference for products with different risk levels can be obtained, i.e., the risk fit vector. The attention weight is obtained by dividing the dot product of the two by the square root of the feature dimension, and then normalized by the softmax function, finally obtaining the risk fit vector head1. The risk fit vector accurately reflects the degree of matching between a user's risk preference and the product's risk attributes, and can be calculated using the following formula: head1=softmax(Q_1·K_ ) / sqrt(d_k)·V_1=Softmax((u_risk·p_ris ) / sqrt(d))·p_risk, where the user risk feature vector u_risk is the query vector Q_1, the product risk feature vector p_risk is the key vector K_1, the product risk feature vector p_risk is the value vector V_1, and the feature dimension d_k=32.

[0061] Furthermore, the second attention module primarily focuses on the correlation between recent user behavior and product type. It extracts sub-vectors closely related to the user's recent transactions, browsing, and query behaviors from the user feature vector, namely the user behavior feature vector u_behavior. Then, using the user behavior feature vector as the query vector and the product feature vector as the key-value vector, a behavior correlation vector head2 is obtained through multi-head attention. This vector quantifies the correlation between user behavior and product attributes, and can be calculated using the following formula:

[0062] head2=Softmax(Q_2·K_ / sqrt(d))·V_2;

[0063] Among them, the user behavior feature vector is Q_2, and the product feature vectors are K_2 and V_2.

[0064] Furthermore, the third attention module focuses on the impact of environmental dynamics on product performance. When this module processes data, it first retrieves the state vector `market` reflecting the current state from the data lake. Then, it performs a dot product operation with each row of the product feature vector `P` (i.e., each product attribute) of the candidate financial products to evaluate the performance potential of each product under the current conditions. The dot product result is then converted into weights using the softmax function. These weights are then multiplied by the product feature vectors and summed to obtain the aggregate vector `head3`, which can be calculated using the following formula: `head3 = Softmax(market·market)`. Finally, the risk fit vector head1, the behavior association vector head2, and the aggregation vector head3 are combined to form a matching vector, which can represent the evaluation results of the product user matching degree from three key perspectives.

[0065] This embodiment utilizes a multi-head attention mechanism to analyze the intrinsic relationship between user characteristics and product attributes from three different dimensions, effectively overcoming the shortcomings in handling multi-dimensional user needs and temporal dependencies, thereby generating more accurate target financial products and improving user experience.

[0066] Optionally, in the method for determining financial products provided in this application embodiment, the output module processes N matching vectors to obtain the product output vector of the candidate financial product, which includes: extracting risk matching vector, behavior correlation vector and aggregation vector from the N matching vectors; obtaining a preset function, concatenating the risk matching vector, behavior correlation vector and aggregation vector based on the preset function to obtain a fused feature vector; obtaining a preset projection matrix, calculating the dot product of the fused feature vector and the preset projection matrix to obtain the product output vector.

[0067] After obtaining the matching vectors, these vectors can be fused. Specifically, the risk fit vector head1 generated by the first attention module, the behavioral correlation vector head2 generated by the second attention module, and the aggregation vector head3 generated by the third attention module can be extracted from the matching vectors. The risk fit vector represents the degree of fit between the user's risk preference and the risk level of the candidate financial product, quantifying the matching degree between the product and the user in terms of risk tolerance, and helping the model understand the user's risk tendency. The behavioral correlation vector can reflect the correlation between the user's recent behavioral sequence and the product type. By capturing the user's behavioral patterns and preferences, the model can assess the user's demand tendency for a specific product type.

[0068] Then, the risk fit vector, behavioral correlation vector, and aggregation vector are concatenated using a preset function (Concat function) to form a multi-dimensional vector that comprehensively reflects the matching relationship between users and products, i.e., the fused feature vector. Finally, the dot product between the fused feature vector and the preset projection matrix W_o is calculated, mapping the fused feature vector to a fixed-size output space, i.e., transforming the fused feature vector to obtain the product output vector. The product output vector H can be calculated using the following formula: H = Concat(head1, head2, head3) · W_o. It should be noted that each dimension of the output vector can represent the matching score between the product and the user on different dimensions. The final score comprehensively considers user needs and product characteristics, providing an accurate suitability score for each candidate financial product.

[0069] This embodiment generates a product output vector that matches all candidate financial products with specific user needs by accurately extracting and fusing user matching features. This not only improves the accuracy of product selection but also makes the screening results meet the user's personalized needs, thereby enhancing the user experience and overall efficiency of financial institutions' product screening services.

[0070] Optionally, in the method for determining financial products provided in this application embodiment, selecting the target financial product from M candidate financial products based on M product output vectors includes: normalizing each product output vector to obtain M product prediction scores; sorting the M product prediction scores in descending order to obtain a score sequence, and obtaining the preset product score of a preset position from the score sequence; and determining the candidate financial product associated with the preset product score of the preset position as the target financial product.

[0071] Specifically, after obtaining the product output vector for each candidate product, to ensure fairness in product score comparison, all product output vectors are normalized. This rescaling of each component of the vector according to a specific range (e.g., 0 to 1) eliminates dimensional differences, facilitating direct comparison. After normalization, each product output vector can be converted into a corresponding predicted product score, which intuitively reflects the target user's matching degree for each candidate financial product. To efficiently identify the most matching product, all predicted product scores can be sorted in descending order to generate a score sequence. Based on business needs and model performance, by pre-setting a ranking threshold, such as Top 5 or Top 10, the top few highest scores from the descending score sequence are selected as the candidate financial products that best match the user's characteristics.

[0072] This embodiment avoids unfair comparisons caused by differences in the rating scales of different products by adjusting the product output vector to a uniform scale, thereby quickly identifying the most attractive financial products, optimizing the screening process, reducing computational resource consumption, improving user experience satisfaction, and greatly enhancing screening efficiency and personalization.

[0073] This application also provides an optional method for determining financial products. Figure 3 This is a schematic diagram of an optional financial product determination method provided according to an embodiment of this application, such as... Figure 3 As shown, the method includes:

[0074] After user authorization, user information of the target user who needs to select financial products can be obtained from the data lake of financial institutions, such as static and dynamic data related to the target user. At the same time, multiple candidate financial products and product data of each candidate financial product can also be obtained. Among them, product data can include: product static attributes and dynamic data, and user information can include: static data and behavior logs.

[0075] Furthermore, the data is converted into a vectorized representation that can be input into the model. This involves cleaning and standardizing the collected data, extracting static information of the target user from the user information, and then performing vector transformation on this static information to obtain the first vector. The target user's behavioral information is then converted into a behavioral sequence, and this sequence is further vectorized to obtain the second vector. The first and second vectors are then concatenated to obtain the user feature vector. Simultaneously, discrete data of each candidate financial product is extracted from its product data, and this discrete data is vectorized to obtain the third vector. Continuous data of the candidate financial products is then vectorized to obtain the fourth vector. The third and fourth vectors are then concatenated to obtain the product feature vector of the candidate financial product.

[0076] Furthermore, the aforementioned vectors are processed using a product prediction model. This model includes an input module, three attention modules, and an output module. The input module obtains product feature vectors corresponding to a set of product data for candidate financial products and user feature vectors corresponding to user information. The attention modules extract feature vectors from both the product and user feature vectors, and each attention module calculates a matching vector based on the extracted feature vectors. The output module processes the matching vectors to obtain the product output vector for each candidate financial product. It should be noted that the first attention module focuses on the matching degree between user risk level and product risk level, and between user return expectation and product return rate; the second attention module captures the correlation between recent user behavior sequences and products; and the third attention module focuses on context awareness.

[0077] Finally, the product output vector of each candidate financial product is normalized to output the product prediction score. All products are sorted in descending order of prediction score, and the candidate financial products associated with the preset scores of the products in the preset order in the score sequence are determined as the target financial products.

[0078] This embodiment inputs product data and user information of multiple candidate financial products into a product prediction model, processes them to obtain corresponding product output vectors, and uses the product output vectors to filter out target financial products from the candidate financial products, thereby achieving the technical effect of improving the accuracy of filtering financial products that match the target user.

[0079] 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.

[0080] Example 2

[0081] This application also provides a device for determining financial products. It should be noted that this device can be used to execute the method for determining financial products provided in this application. The following describes the device for determining financial products provided in this application.

[0082] According to an embodiment of this application, an apparatus for implementing the above-described method for determining financial products is also provided. Figure 4 This is a schematic diagram of a financial product determination device provided according to an embodiment of this application, such as... Figure 4 As shown, the device includes: an acquisition unit 40, an input unit 41, and a filtering unit 42.

[0083] The acquisition unit 40 is used to acquire the user information of the target user with the authorization of the target user, and to acquire M candidate financial products and product data of each candidate financial product from the data lake to obtain M sets of product data, where M is a positive integer;

[0084] Input unit 41 is used to input M sets of product data and user information into the product prediction model, process it to obtain M product output vectors, wherein each product output vector is used to characterize the matching degree between each candidate financial product and the target user, and the product prediction model is used to construct M matching vectors between candidate financial products and target users based on M sets of product data and user information, and use the M matching vectors between candidate financial products and target users to generate M product output vectors.

[0085] The filtering unit 42 is used to filter the target financial product from the M candidate financial products based on the M product output vectors.

[0086] The financial product determination device provided in this application embodiment obtains user information of the target user with the authorization of the target user by acquisition unit 40, and obtains M candidate financial products and product data of each candidate financial product from the data lake to obtain M sets of product data, where M is a positive integer; input unit 41 inputs the M sets of product data and user information into a product prediction model, processes it to obtain M product output vectors, where each product output vector is used to characterize the matching degree between each candidate financial product and the target user, the product prediction model is used to construct the matching vectors between the M candidate financial products and the target user based on the M sets of product data and user information, and uses the matching vectors between the M candidate financial products and the target user to generate M product output vectors; screening unit 42 selects the target financial product from the M candidate financial products based on the M product output vectors, which solves the technical problem of low screening accuracy when screening financial products matching the target user in related technologies. By inputting the product data and user information of multiple candidate financial products into the product prediction model, processing it to obtain the corresponding product output vectors, and using the product output vectors to screen the target financial product from the candidate financial products, the technical effect of improving the screening accuracy of financial products matching the target user is achieved.

[0087] Optionally, in the financial product determination device provided in the embodiments of this application, the device further includes: a first extraction unit, used to extract static information of the target user from the user information before inputting M sets of product data and user information into the product prediction model, and to perform vector transformation on the static information to obtain a first vector; a second extraction unit, used to extract behavioral information of the target user from the user information, convert the behavioral information into a behavioral sequence, and perform vector transformation on the behavioral sequence to obtain a second vector; and a concatenation unit, used to concatenate the first vector and the second vector to obtain a user feature vector.

[0088] Optionally, in the financial product determination device provided in the embodiments of this application, the device further includes: a third extraction unit, used to extract discrete data of a candidate financial product from a set of product data for a candidate financial product before inputting M sets of product data and user information into the product prediction model, and to perform vector transformation on the discrete data to obtain a third vector; and a fourth extraction unit, used to extract continuous data of the candidate financial product from a set of product data, to perform vector mapping processing on the continuous data to obtain a fourth vector, and to concatenate the third vector and the fourth vector to obtain the product feature vector of the candidate financial product.

[0089] Optionally, in the financial product determination device provided in this application embodiment, the input unit 41 includes: a first acquisition module, used to acquire M product feature vectors corresponding to a set of product data of a candidate financial product and acquire user feature vectors corresponding to user information for a candidate financial product; a first extraction module, used to extract feature vectors from the product feature vectors and user feature vectors by N attention modules respectively, and to calculate N matching vectors by each attention module based on the extracted feature vectors; and a processing module, used to process the N matching vectors to obtain the product output vector of the candidate financial product.

[0090] Optionally, in the financial product determination device provided in this application embodiment, the input unit 41 includes: a second extraction module, used by a first attention module to extract a product risk feature vector from a product feature vector, extract a user risk feature vector from a user feature vector, use the product risk feature vector as a key-value vector, use the user risk feature vector as a query vector, and calculate a risk matching vector based on the key-value vector and the query vector; a third extraction module, used by the second attention module to extract a user behavior feature vector from a user feature vector, use the user behavior feature vector as a query vector, use the product feature vector as a key-value vector, and calculate a behavior association vector based on the key-value vector and the query vector; a second acquisition module, used to acquire a state vector, and the third attention module calculates an aggregation vector based on the state vector and the product feature vector; and a composition module, used to construct N matching vectors from the risk matching vector, the behavior association vector, and the aggregation vector.

[0091] Optionally, in the financial product determination device provided in this application embodiment, the input unit 41 includes: a fourth extraction module, used to extract risk matching vector, behavioral correlation vector and aggregation vector from N matching vectors; a third acquisition module, used to acquire a preset function, and concatenate the risk matching vector, behavioral correlation vector and aggregation vector based on the preset function to obtain a fused feature vector; and a fourth acquisition module, used to acquire a preset projection matrix, calculate the dot product of the fused feature vector and the preset projection matrix to obtain the product output vector.

[0092] Optionally, in the financial product determination device provided in this application embodiment, the screening unit 42 includes: a processing module, used to normalize the output vector of each product to obtain M product prediction scores; a sorting module, used to sort the M product prediction scores in descending order to obtain a score sequence, and obtain the preset product score of the preset position from the score sequence; and a determination module, used to determine the candidate financial products associated with the preset product score of the preset position as the target financial products.

[0093] It should be noted that the acquisition unit 40, input unit 41, and filtering unit 42 mentioned above correspond to steps S201 to S203 in Embodiment 1. The instances and application scenarios implemented by the above 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 units can also be part of a device and run in the computer terminal 10 provided in Embodiment 1.

[0094] Example 3

[0095] Embodiments of this application may provide a computer terminal, which may be any computer terminal device in a group of computer terminals. Optionally, in this embodiment, the aforementioned computer terminal may also be replaced with a mobile terminal or an electronic device, etc.

[0096] Optionally, in this embodiment, the computer terminal may be located in at least one of a plurality of network devices in a computer network.

[0097] In this embodiment, the computer terminal described above can execute the program code for the following steps in the method for determining financial products: With the authorization of the target user, obtain the target user's user information, and obtain M candidate financial products and product data for each candidate financial product from the data lake, resulting in M ​​sets of product data, where M is a positive integer; input the M sets of product data and user information into a product prediction model, process it to obtain M product output vectors, where each product output vector is used to characterize the degree of matching between each candidate financial product and the target user; the product prediction model is used to construct matching vectors between the M candidate financial products and the target user based on the M sets of product data and user information, and generate M product output vectors using the matching vectors between the M candidate financial products and the target user; select the target financial product from the M candidate financial products based on the M product output vectors.

[0098] Optionally, the computer terminal described above can execute the program code for the following steps in the method for determining financial products: extracting static information of the target user from user information, performing vector transformation on the static information to obtain a first vector; extracting behavioral information of the target user from user information, converting the behavioral information into a behavioral sequence, and performing vector transformation on the behavioral sequence to obtain a second vector; concatenating the first vector and the second vector to obtain a user feature vector.

[0099] Optionally, the computer terminal described above can execute the program code for the following steps in the method for determining financial products: for a set of product data of a candidate financial product, extract discrete data of the candidate financial product from the set of product data, perform vector transformation on the discrete data to obtain a third vector; extract continuous data of the candidate financial product from the set of product data, perform vector mapping processing on the continuous data to obtain a fourth vector, and concatenate the third vector and the fourth vector to obtain the product feature vector of the candidate financial product.

[0100] Optionally, the aforementioned computer terminal can execute the program code for the following steps in the method for determining financial products: For a candidate financial product, the input module obtains M product feature vectors corresponding to a set of product data of the candidate financial product, and obtains user feature vectors corresponding to user information; N attention modules extract feature vectors from the product feature vectors and user feature vectors respectively, and each attention module calculates N matching vectors based on the extracted feature vectors; the output module processes the N matching vectors to obtain the product output vector of the candidate financial product.

[0101] Optionally, the aforementioned computer terminal can execute the program code for the following steps in the method for determining financial products: The first attention module extracts a product risk feature vector from the product feature vector and a user risk feature vector from the user feature vector; the product risk feature vector is used as a key-value vector, and the user risk feature vector is used as a query vector; a risk fit vector is calculated based on the key-value vector and the query vector. The second attention module extracts a user behavior feature vector from the user feature vector; the user behavior feature vector is used as a query vector, and the product feature vector is used as a key-value vector; a behavior association vector is calculated based on the key-value vector and the query vector. A state vector is obtained, and the third attention module calculates an aggregation vector based on the state vector and the product feature vector. N matching vectors are formed by the risk fit vector, the behavior association vector, and the aggregation vector.

[0102] Optionally, the computer terminal described above can execute the program code for the following steps in the method for determining financial products: extracting risk-matching vectors, behavioral correlation vectors, and aggregation vectors from N matching vectors; obtaining a preset function, concatenating the risk-matching vectors, behavioral correlation vectors, and aggregation vectors based on the preset function to obtain a fused feature vector; obtaining a preset projection matrix, calculating the dot product of the fused feature vector and the preset projection matrix to obtain the product output vector.

[0103] Optionally, the computer terminal described above can execute the program code for the following steps in the method for determining financial products: normalize the output vector of each product to obtain M product prediction scores; sort the M product prediction scores in descending order to obtain a score sequence, and obtain the preset product score of the preset position from the score sequence; and determine the candidate financial products associated with the preset product score of the preset position as the target financial products.

[0104] Optionally, Figure 5 This is a structural block diagram of an electronic device according to an embodiment of this application. Figure 5 As shown, the electronic device may include: one or more ( Figure 5 (Only one is shown) processor 502, memory 504, memory controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0105] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the financial product determination method and apparatus in this application embodiment. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby realizing the aforementioned financial product determination method. 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.

[0106] The processor can access the information and application programs stored in the memory via the transmission device to execute the steps described above in the method for determining the aforementioned financial product.

[0107] Those skilled in the art will understand that Figure 5 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 5 This does not limit the structure of the aforementioned electronic device. For example, electronic devices may also include components that are more... Figure 5 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 5 The different configurations shown.

[0108] 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.

[0109] Example 4

[0110] 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 method for determining financial products provided in Embodiment 1.

[0111] 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.

[0112] Optionally, in this embodiment, the storage medium is configured to store program code for performing the following steps: with the authorization of the target user, obtain the target user's user information, and obtain M candidate financial products and product data for each candidate financial product from the data lake to obtain M sets of product data, where M is a positive integer; input the M sets of product data and user information into the product prediction model, process to obtain M product output vectors, where each product output vector is used to characterize the matching degree between each candidate financial product and the target user, the product prediction model is used to construct matching vectors between the M candidate financial products and the target user based on the M sets of product data and user information, and generate M product output vectors using the matching vectors between the M candidate financial products and the target user; select the target financial product from the M candidate financial products based on the M product output vectors.

[0113] This application also provides a computer program product that, when executed on a data processing device, is suitable for performing the steps of a method for determining a financial product.

[0114] 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.

[0115] 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.

[0116] 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 coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of units or modules may be electrical or other forms.

[0117] 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.

[0118] 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.

[0119] 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 of 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.

[0120] 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 determining a financial product, characterized in that, include: With the authorization of the target user, the user information of the target user is obtained, and M candidate financial products and product data of each candidate financial product are obtained from the data lake to obtain M sets of product data, where M is a positive integer; The M sets of product data and the user information are input into the product prediction model and processed to obtain M product output vectors. Each product output vector is used to characterize the matching degree between each candidate financial product and the target user. The product prediction model is used to construct the matching vectors between the M candidate financial products and the target user based on the M sets of product data and the user information, and to generate the M product output vectors using the matching vectors between the M candidate financial products and the target user. The target financial product is selected from the M candidate financial products based on the M product output vectors.

2. The method according to claim 1, characterized in that, Before inputting the M sets of product data and the user information into the product prediction model, the method further includes: Extract static information of the target user from the user information, and perform vector transformation on the static information to obtain a first vector; Extract the target user's behavioral information from the user information, convert the behavioral information into a behavioral sequence, and perform vector transformation on the behavioral sequence to obtain a second vector; The first vector and the second vector are concatenated to obtain the user feature vector.

3. The method according to claim 1, characterized in that, Before inputting the M sets of product data and the user information into the product prediction model, the method further includes: For a set of product data for a candidate financial product, extract discrete data of the candidate financial product from the set of product data, and perform vector transformation on the discrete data to obtain a third vector; Continuous data of the candidate financial products are extracted from the set of product data. The continuous data is then processed by vector mapping to obtain a fourth vector. The third vector and the fourth vector are then concatenated to obtain the product feature vector of the candidate financial product.

4. The method according to claim 1, characterized in that, The product prediction model includes an input module, N attention modules, and an output module, where N is a positive integer. The M sets of product data and the user information are input into the product prediction model, and the resulting M product output vectors are processed to obtain: For a candidate financial product, the input module obtains M product feature vectors corresponding to a set of product data of the candidate financial product, and obtains the user feature vector corresponding to the user information; The N attention modules extract feature vectors from the product feature vector and the user feature vector respectively, and each attention module calculates N matching vectors based on the extracted feature vectors; The output module processes the N matching vectors to obtain the product output vector of the candidate financial product.

5. The method according to claim 4, characterized in that, When the N attention modules include a first attention module, a second attention module, and a third attention module, the N attention modules respectively extract feature vectors from the product feature vector and the user feature vector, and each attention module calculates N matching vectors based on the extracted feature vectors, including: The first attention module extracts a product risk feature vector from the product feature vector and a user risk feature vector from the user feature vector. The product risk feature vector is used as a key-value vector and the user risk feature vector is used as a query vector. A risk fit vector is calculated based on the key-value vector and the query vector. The second attention module extracts user behavior feature vectors from the user feature vectors, uses the user behavior feature vectors as query vectors, uses the product feature vectors as key-value vectors, and calculates behavior association vectors based on the key-value vectors and the query vectors. The state vector is obtained, and the third attention module calculates the aggregate vector based on the state vector and the product feature vector. The N matching vectors are composed of the risk-matching vector, the behavior-related vector, and the aggregation vector.

6. The method according to claim 4, characterized in that, The product output vector of the candidate financial product obtained by the output module based on N matching vectors includes: Extract risk-matching vectors, behavior-related vectors, and aggregation vectors from the N matching vectors; Obtain a preset function, and concatenate the risk matching vector, the behavior association vector, and the aggregation vector based on the preset function to obtain a fused feature vector; Obtain a preset projection matrix, calculate the dot product of the fused feature vector and the preset projection matrix, and obtain the product output vector.

7. The method according to claim 1, characterized in that, The target financial products are selected from the M candidate financial products based on the M product output vectors, including: The output vector of each product is normalized to obtain M product prediction scores; The predicted scores of the M products are sorted in descending order to obtain a score sequence, and the preset scores of the products in a preset position are obtained from the score sequence. The candidate financial products associated with the preset scores of the products in the preset order are determined as the target financial products.

8. A device for determining a financial product, characterized in that, include: The acquisition unit is used to acquire the user information of the target user with the authorization of the target user, and to acquire M candidate financial products and product data of each candidate financial product from the data lake to obtain M sets of product data, where M is a positive integer; The input unit is used to input the M sets of product data and the user information into the product prediction model, process them to obtain M product output vectors, wherein each product output vector is used to characterize the matching degree between each candidate financial product and the target user, and the product prediction model is used to construct the matching vectors between the M candidate financial products and the target user based on the M sets of product data and the user information, and generate the M product output vectors using the matching vectors between the M candidate financial products and the target user; The filtering unit is used to filter the target financial product from the M candidate financial products based on the M product output vectors.

9. An electronic device, characterized in that, include: Memory, which stores executable programs; A processor for running the program, wherein the program, when running, performs the method for determining a financial product according to any one of claims 1 to 7.

10. 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 method for determining the financial product according to any one of claims 1 to 7.

Images