Financial product recommendation method and device based on comparative learning, equipment and medium
By comparing the heterogeneous data processing of learning methods, extracting and converting the feature vectors of financial products, the problem of low feature accuracy in traditional methods is solved, and a higher accuracy of financial products recommendation is achieved.
Patent Information
- Application Number
- CN202411068073.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-05
- Publication Date
- 2024-12-17
AI Technical Summary
Traditional contrast learning methods are difficult to effectively process heterogeneous data in financial products, resulting in low feature accuracy and low recommendation accuracy.
Using the financial product recommendation method based on comparison learning, different types of data are processed through heterogeneous data feature extraction layer, feature conversion layer and vector similarity calculation layer, meaningful feature vectors are extracted, and the attention mechanism and FM feature intersection are crossed. Methods improve feature accuracy.
It improves the accuracy of financial product recommendations, can effectively process different types of data, and improves the accuracy of recommendation results.
Smart Images
Figure CN119149806A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the fields of data processing and financial technology, and in particular to a financial product recommendation method, device, computer equipment and storage medium based on contrastive learning. Background Art
[0002] With the development of computer technology, intelligent systems such as product recommendation and product search are gradually applied to the financial field. For example, the financial product recommendation system can recommend financial products to users based on the purchase needs, product needs and other information input by users. In the financial field, it is usually necessary to compare the similarities and differences between different financial products to obtain more accurate recommendation results. Since the data sources of financial products are diverse and heterogeneous, it is usually necessary to analyze heterogeneous data such as text, numbers, and images at the same time. However, traditional contrastive learning methods can usually only process data of the same type, resulting in low accuracy of features extracted from product data, and thus low accuracy of recommended products. Therefore, how to improve the accuracy of data features and thus improve the accuracy of financial product recommendations has become an urgent problem to be solved. Summary of the invention
[0003] The present application provides a financial product recommendation method, apparatus, computer equipment and storage medium based on contrastive learning to improve the accuracy of data features and thereby improve the accuracy of financial product recommendations.
[0004] In a first aspect, the present application provides a financial product recommendation method based on contrastive learning, the method comprising:
[0005] Acquire data to be processed, wherein the data to be processed includes at least one data category;
[0006] Based on the preset product recommendation model and the data category, the feature vectors corresponding to each financial product in the preset product database and the data to be processed are processed to determine the target financial product, and recommend the target financial product to the user.
[0007] In a second aspect, the present application further provides a financial product recommendation device based on contrastive learning, the device comprising:
[0008] A data acquisition module, used to acquire data to be processed, wherein the data to be processed includes at least one data category;
[0009] The product determination module is used to process the feature vectors corresponding to each financial product in the preset product database and the data to be processed based on the preset product recommendation model and the data category, determine the target financial product, and recommend the target financial product to the user.
[0010] In a third aspect, the present application also provides a computer device, comprising a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program and implement the financial product recommendation method based on contrastive learning as described above when executing the computer program.
[0011] In a fourth aspect, the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements the financial product recommendation method based on contrastive learning as described above.
[0012] The present application discloses a financial product recommendation method, device, computer equipment and storage medium based on contrastive learning, obtaining data to be processed, wherein the data to be processed includes at least one data category; based on a preset product recommendation model and the data category, processing the feature vectors corresponding to each financial product in a preset product database and the data to be processed, determining a target financial product, and recommending the target financial product to the user. The present application can process the data to be processed containing different data categories and the data corresponding to the financial product according to the product recommendation model, thereby realizing targeted processing of different types of data, thereby improving the accuracy of the recommendation results. BRIEF DESCRIPTION OF THE DRAWINGS
[0013] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0014] Figure 1 is a schematic flow chart of a first embodiment of a financial product recommendation method based on contrastive learning provided by an embodiment of the present application;
[0015] Figure 2 is a schematic flow chart of a second embodiment of a financial product recommendation method based on contrastive learning provided in an embodiment of the present application;
[0016] Figure 3 A schematic block diagram of a financial product recommendation device based on contrastive learning provided in an embodiment of the present application;
[0017] Figure 4 A schematic block diagram of the structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0018] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0019] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.
[0020] It should be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.
[0021] It should be further understood that the term “and / or” used in the specification and appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.
[0022] The embodiments of the present application provide a financial product recommendation method, device, computer equipment and storage medium based on contrastive learning. The financial product recommendation method based on contrastive learning can be applied to a server, and the to-be-processed data containing different data categories and the data corresponding to the financial products are processed according to the product recommendation model, so as to achieve targeted processing of different types of data, thereby improving the accuracy of the recommendation results. The server can be an independent server or a server cluster.
[0023] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.
[0024] See also Figure 1 , Figure 1 This is a schematic flow chart of a financial product recommendation method based on contrastive learning provided by an embodiment of the present application. The financial product recommendation method based on contrastive learning can be applied to a server to process the to-be-processed data containing different data categories and the data corresponding to the financial products according to the product recommendation model, thereby achieving targeted processing of different types of data and improving the accuracy of the recommendation results.
[0025] like Figure 1As shown, the financial product recommendation method based on contrastive learning specifically includes steps S101 to S102.
[0026] S101, obtaining data to be processed, wherein the data to be processed includes at least one data category;
[0027] In one embodiment, the data to be processed may be obtained from user input information. For example, the user inputs relevant requirements of the financial product that the user wants to purchase, and the relevant requirements input by the user are sorted to obtain the data to be processed;
[0028] In one embodiment, the data to be processed can also be obtained by analyzing user portrait data, purchased product data, investment tendency and other information.
[0029] In one embodiment, the data to be processed includes data of at least one data category, such as image data, text data, etc. For example, financial product data generally includes data of categories such as text, numbers, and images. Medical diagnosis data also includes data of multiple categories such as text, numbers, and images.
[0030] S102: Based on a preset product recommendation model and the data category, the feature vectors corresponding to each financial product in a preset product database and the data to be processed are processed to determine a target financial product, and the target financial product is recommended to the user.
[0031] In one embodiment, the product recommendation model includes a heterogeneous data feature extraction layer, a feature conversion layer, and a vector similarity calculation layer.
[0032] In one embodiment, the heterogeneous data feature extraction layer can process the data to be processed containing different types of data and extract meaningful feature vectors. Different encoding methods are used for different types of data. For example, for image data, a convolutional neural network model is used for feature extraction; for text data, a word vector model or a topic model is used for feature extraction.
[0033] In one embodiment, the feature conversion layer is used to convert the feature vectors of heterogeneous data so as to facilitate comparison in the same space. A feature crossover method based on FM (Factor i zat i on Machine) is adopted to process multiple features simultaneously, and an Attention mechanism is introduced to enable the model to focus on important information.
[0034] In one embodiment, the vector similarity calculation layer is used to calculate the similarity between the two converted vectors and obtain their similarity scores. A Simese neural network is used to process the two input data respectively through the same learning parameters, and then output their similarity scores to indicate the degree of similarity between them.
[0035] In one embodiment, recommending the target financial product to the user may be by displaying the identifier of the financial product to the user, or by displaying the comparison result and similarity between the data to be processed and the relevant data of the target financial product to the user for reference.
[0036] Furthermore, before the feature vectors corresponding to each financial product in a preset product database and the data to be processed are processed based on the preset product recommendation model and the data category to determine the target financial product and recommend the target financial product to the user, it also includes: obtaining a training data set and a verification data set; based on a preset pre-trained model, processing the training data set to obtain predicted recommended products; based on the verification data set and the predicted recommended products, obtaining a cross-entropy loss function value of the pre-trained model; and when the cross-entropy loss function value is less than a preset loss threshold, determining the pre-trained model as a product recommendation model.
[0037] In one embodiment, the pre-training model includes a heterogeneous data feature extraction layer, a feature conversion layer, and a vector similarity calculation layer. The heterogeneous data feature extraction layer performs feature extraction on the training data set to obtain a feature vector corresponding to the training data set. The feature conversion layer performs feature conversion on the feature vector corresponding to the training data set to obtain a converted vector. The vector similarity calculation layer calculates a similarity score between the converted vector and the vector corresponding to the financial product. The predicted recommended product is determined based on the similarity score.
[0038] In one embodiment, the validation data set stores the best recommended products corresponding to each training data. The predicted recommended products are compared with the recommended products in the validation data set to obtain the cross entropy loss function value of the pre-training model.
[0039] In one embodiment, when the cross entropy loss function value is less than or equal to a preset loss threshold, the pre-trained model is determined as a product recommendation model; when the cross entropy loss function value is greater than the loss threshold, the parameters of the pre-trained model are adjusted to obtain a new pre-trained model, which is retrained using the training data set until the cross entropy loss function value is less than or equal to the cross entropy loss function value threshold. The loss threshold can be freely set by the user according to actual needs.
[0040] In a specific embodiment, a cross entropy loss function and an Adam (Adaptive Moment Estimation, optimization algorithm) optimizer can be used for model training, and a stochastic gradient descent algorithm can be used to update model parameters. The cross entropy loss function value between the predicted value and the true value corresponding to the current batch of training data is calculated, the gradient of the loss relative to the model parameters is calculated, and the Adam optimizer is used to update the parameters of the model according to the calculated gradient.
[0041] The above embodiment provides a financial product recommendation method, device, computer equipment and storage medium based on contrastive learning, obtaining data to be processed, wherein the data to be processed includes at least one data category; based on a preset product recommendation model and the data category, processing the feature vectors corresponding to each financial product in a preset product database and the data to be processed, determining the target financial product, and recommending the target financial product to the user. The present application can process the data to be processed containing different data categories and the data corresponding to the financial product according to the product recommendation model, thereby realizing targeted processing of different types of data, and thereby improving the accuracy of the recommendation results.
[0042] See also Figure 2 , Figure 2 This is a schematic flow chart of a financial product recommendation method based on contrastive learning provided by an embodiment of the present application. The financial product recommendation method based on contrastive learning can be applied to a server, and is used to perform targeted feature extraction on a data set according to the data category of each data set of the data to be processed through the heterogeneous data feature extraction layer of the product recommendation model, obtain an initial feature vector, and improve the accuracy of the feature vector; then, the feature conversion layer of the product recommendation model performs feature conversion and weighting processing on the obtained initial feature vector, and the attention of different initial feature vectors can be determined according to the weight coefficient. This makes the target feature vector obtained according to the data to be processed more accurate, thereby improving the accuracy of financial product recommendations.
[0043] like Figure 2 As shown, the financial product recommendation method based on contrastive learning specifically includes steps S201 to S204.
[0044] S201, based on the heterogeneous data feature extraction layer of the product recommendation model, processing the data to be processed to obtain at least one initial feature vector;
[0045] Furthermore, the heterogeneous data feature extraction layer based on the product recommendation model processes the data to be processed to obtain at least one initial feature vector, including: classifying the data to be processed based on the heterogeneous data feature extraction layer to obtain a data set corresponding to each of the data categories; acquiring a feature extraction model corresponding to each of the data categories; and extracting features from the data set corresponding to each of the data categories based on the feature extraction model corresponding to each of the data categories to obtain at least one of the initial feature vectors.
[0046] In one embodiment, the data to be processed includes data of at least one data category, such as image data, text data, voice data, etc.
[0047] In one embodiment, the heterogeneous data feature extraction layer can extract meaningful feature vectors from heterogeneous data and use different feature extraction models for different types of data.
[0048] In one embodiment, the categories of data to be processed are identified, and the data to be processed are grouped according to the data categories to obtain a data set corresponding to each data category.
[0049] In one embodiment, each data category corresponds to a feature extraction model, and feature extraction is performed on the data set corresponding to the data category according to the feature extraction model corresponding to each data category to obtain an initial feature vector corresponding to each data set. For example, for image data in the data to be processed, a convolutional neural network model can be used for feature extraction, and for text data, a word vector model or a topic model can be used for feature extraction.
[0050] S202: Based on the feature conversion layer of the product recommendation model, process at least one of the initial feature vectors to obtain a target feature vector;
[0051] Furthermore, the feature conversion layer based on the product recommendation model processes at least one of the initial feature vectors to obtain a target feature vector, including: based on the feature conversion layer, performing feature conversion on all of the initial feature vectors to obtain at least one feature to be weighted; based on the attention mechanism, performing weighted processing on at least one of the features to be weighted to obtain the target feature vector.
[0052] In one embodiment, the feature conversion layer is used to convert the initial feature vectors of the data set corresponding to each data category to facilitate comparison in the same space. A feature crossover method based on FM (Factorization Machine) can be used to process multiple features at the same time, and an Attention mechanism is introduced to enable the model to focus on important information.
[0053] In a specific embodiment, it is assumed that the extracted i-th initial feature vector is x i , through the feature conversion layer, they are converted into vectors v through FM i , which is the weighted vector. Then the Attention mechanism is added so that the model can pay attention to the weight coefficient α of each feature i , and obtain the weighted target feature vector v~1.
[0054] In one embodiment, the feature conversion formula based on FM is:
[0055] v i =W[x i φ(x i )]
[0056] Among them, φ(x i ) represents the feature x i The transformation function performed may include feature encoding, normalization or other preprocessing steps.
[0057]
[0058] Where n is the number of initial eigenvectors.
[0059] S203, based on the vector similarity calculation layer of the product recommendation model, performing similarity calculation on the target feature vector and the feature vectors corresponding to each financial product in the product database to obtain a similarity score;
[0060] Furthermore, the vector similarity calculation layer based on the product recommendation model performs similarity calculation on the target feature vector and the feature vectors corresponding to each financial product in the product database to obtain a similarity score, including: based on the similarity calculation layer and a first preset formula, regularizing and linearly transforming the target feature vector and the vectors corresponding to each financial product to obtain a first vector corresponding to the target feature vector and a second vector corresponding to the financial product; performing similarity calculation based on a second preset formula, the first vector and the second vector to obtain the similarity score.
[0061] In one embodiment, the vector similarity calculation layer is used to calculate a similarity score between the target feature vector and the vector corresponding to the financial product.
[0062] In a specific embodiment, first, the target feature vector and the vector corresponding to the financial product are converted according to a first preset formula to obtain a first vector h1 corresponding to the target feature vector and a second vector h2 corresponding to the financial product feature vector. Then, according to a second preset formula, the similarity of the first vector and the second vector is calculated to obtain a similarity score to indicate the degree of similarity between the target feature and the vector of the financial product.
[0063] Furthermore, the first preset formula is:
[0064]
[0065]
[0066] Among them, v~1 represents the weighted target feature vector; Represents the vector corresponding to the financial product; W h represents the weight matrix; b h is a bias vector; h1 and h2 represent the first vector and the second vector respectively; ⊙ represents an element-wise multiplication operation; u is a random Dropout vector used to set some elements to 0;
[0067] The second preset formula is:
[0068]
[0069] Among them, o represents the output value, that is, the similarity score; e represents the base of the natural logarithm.
[0070] In one embodiment, Dropout is a regularization technique used to prevent neural networks from overfitting. Dropout is achieved by randomly discarding (i.e., setting activation to zero) the output of a portion of neurons during training. This can prevent the network from being overly dependent on specific samples in the training data, thereby improving the generalization ability of the model.
[0071] In one embodiment, the weight W h and bias b h are the parameters of the fully connected layers, which are adjusted during training by a learning algorithm (such as gradient descent) to minimize the loss function of the model.
[0072] S204: Determine the target financial product in the product database based on the similarity score.
[0073] In one embodiment, the similarity score is compared with a preset similarity threshold to obtain a comparison result. In the comparison result, the target feature vector and the feature vector of the financial product whose similarity score exceeds the similarity threshold and has the highest similarity score are determined, and the financial product is used as the target financial product.
[0074] In the above embodiment, the heterogeneous data feature extraction layer of the product recommendation model performs targeted feature extraction on the data set according to the data category of each data set of the data to be processed, obtains the initial feature vector, and improves the accuracy of the feature vector; then, the feature conversion layer of the product recommendation model performs feature conversion and weighting processing on the obtained initial feature vector, and the attention of different initial feature vectors can be determined according to the weight coefficient. This makes the target feature vector obtained according to the data to be processed more accurate, thereby improving the recommendation accuracy of financial products.
[0075] See also Figure 3 , Figure 3 The embodiment of the present application provides a schematic block diagram of a financial product recommendation device based on contrastive learning, which is used to execute the aforementioned financial product recommendation method based on contrastive learning. The financial product recommendation device based on contrastive learning can be configured on a server.
[0076] like Figure 3 As shown, the financial product recommendation device 300 based on contrastive learning includes:
[0077] The data acquisition module 301 is used to acquire data to be processed, wherein the data to be processed includes at least one data category;
[0078] The product determination module 302 is used to process the feature vectors corresponding to each financial product in the preset product database and the data to be processed based on the preset product recommendation model and the data category, determine the target financial product, and recommend the target financial product to the user.
[0079] Furthermore, the product determination module 302 includes:
[0080] An initial feature vector obtaining unit, configured to process the data to be processed based on the heterogeneous data feature extraction layer of the product recommendation model to obtain at least one initial feature vector;
[0081] a target feature vector obtaining unit, configured to process at least one of the initial feature vectors based on the feature conversion layer of the product recommendation model to obtain a target feature vector;
[0082] A similarity score obtaining unit, configured to perform similarity calculation on the target feature vector and the feature vectors corresponding to each financial product in the product database based on the vector similarity calculation layer of the product recommendation model to obtain a similarity score;
[0083] A target financial product determining unit is configured to determine the target financial product in the product database based on the similarity score.
[0084] Furthermore, the initial feature vector obtaining unit includes:
[0085] A data set acquisition subunit, used for classifying the data to be processed based on the heterogeneous data feature extraction layer to obtain data sets corresponding to each data category;
[0086] A feature extraction model acquisition subunit, used to acquire a feature extraction model corresponding to each of the data categories;
[0087] The initial feature vector obtaining subunit is used to extract features from the data set corresponding to each of the data categories based on the feature extraction model corresponding to each of the data categories to obtain at least one of the initial feature vectors.
[0088] Furthermore, the target feature vector obtaining unit includes:
[0089] A feature to be weighted obtaining subunit, used for performing feature conversion on all the initial feature vectors based on the feature conversion layer to obtain at least one feature to be weighted;
[0090] The target feature vector obtaining subunit is used to perform weighted processing on at least one of the features to be weighted based on the attention mechanism to obtain the target feature vector.
[0091] Furthermore, the similarity score obtaining unit includes:
[0092] A vector transformation subunit, configured to perform regularization and linear transformation on the target feature vector and the vectors corresponding to each financial product based on the similarity calculation layer and the first preset formula, to obtain a first vector corresponding to the target feature vector and a second vector corresponding to the financial product;
[0093] The similarity calculation subunit is used to perform similarity calculation based on a second preset formula, the first vector and the second vector to obtain the similarity score.
[0094] Furthermore, the first preset formula is:
[0095]
[0096]
[0097] in, represents the weighted target feature vector; Represents the vector corresponding to the financial product; W h represents the weight matrix; b his a bias vector; h1 and h2 represent the first vector and the second vector respectively; ⊙ represents an element-wise multiplication operation; u is a random Dropout vector used to set some elements to 0;
[0098] The second preset formula is:
[0099]
[0100] Among them, o represents the output value, that is, the similarity score; e represents the base of the natural logarithm.
[0101] Furthermore, the financial product recommendation device 300 based on contrastive learning further includes a model training module, and the model training module includes:
[0102] A data set acquisition unit, used to acquire a training data set and a verification data set;
[0103] A product prediction unit, used to process the training data set based on a preset pre-training model to obtain predicted recommended products;
[0104] A loss function value obtaining unit, used to obtain a cross entropy loss function value of the pre-trained model based on the verification data set and the predicted recommended product;
[0105] A model determination unit is used to determine the pre-trained model as a product recommendation model when the cross entropy loss function value is greater than or equal to a preset loss threshold.
[0106] It should be noted that those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described device and each module can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.
[0107] The above-mentioned device can be implemented in the form of a computer program. Figure 4 Runs on the computer device shown.
[0108] See also Figure 4 , Figure 4 1 is a schematic block diagram of a computer device provided in an embodiment of the present application. The computer device may be a server.
[0109] See also Figure 4 The computer device includes a processor, a memory and a network interface connected through a system bus, wherein the memory may include a non-volatile storage medium and an internal memory.
[0110] The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions, and when the program instructions are executed, the processor can execute any financial product recommendation method based on contrastive learning.
[0111] The processor is used to provide computing and control capabilities and support the operation of the entire computer equipment.
[0112] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor, the processor can execute any financial product recommendation method based on contrastive learning.
[0113] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0114] It should be understood that the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.
[0115] In one embodiment, the processor is used to run a computer program stored in the memory to implement the following steps:
[0116] Acquire data to be processed, wherein the data to be processed includes at least one data category;
[0117] Based on the preset product recommendation model and the data category, the feature vectors corresponding to each financial product in the preset product database and the data to be processed are processed to determine the target financial product, and recommend the target financial product to the user.
[0118] In one embodiment, when the processor processes the feature vectors corresponding to each financial product in a preset product database and the data to be processed based on the preset product recommendation model and the data category to determine the target financial product, it is used to implement:
[0119] Based on the heterogeneous data feature extraction layer of the product recommendation model, the data to be processed is processed to obtain at least one initial feature vector;
[0120] Based on the feature conversion layer of the product recommendation model, processing at least one of the initial feature vectors to obtain a target feature vector;
[0121] Based on the vector similarity calculation layer of the product recommendation model, similarity calculation is performed between the target feature vector and the feature vectors corresponding to each financial product in the product database to obtain a similarity score;
[0122] Based on the similarity score, the target financial product is determined in the product database.
[0123] In one embodiment, when the processor implements the heterogeneous data feature extraction layer based on the product recommendation model and processes the data to be processed to obtain at least one initial feature vector, it is used to implement:
[0124] Based on the heterogeneous data feature extraction layer, the data to be processed are classified to obtain data sets corresponding to each data category;
[0125] Acquire a feature extraction model corresponding to each of the data categories;
[0126] Based on the feature extraction model corresponding to each of the data categories, feature extraction is performed on the data set corresponding to each of the data categories to obtain at least one of the initial feature vectors.
[0127] In one embodiment, when the processor implements the feature conversion layer based on the product recommendation model and processes at least one of the initial feature vectors to obtain a target feature vector, it is configured to implement:
[0128] Based on the feature conversion layer, performing feature conversion on all the initial feature vectors to obtain at least one feature to be weighted;
[0129] Based on the attention mechanism, weighted processing is performed on at least one of the features to be weighted to obtain the target feature vector.
[0130] In one embodiment, when the processor implements the vector similarity calculation layer based on the product recommendation model, performs similarity calculation on the target feature vector and the feature vectors corresponding to each financial product in the product database to obtain a similarity score, it is used to implement:
[0131] Based on the similarity calculation layer and the first preset formula, regularize and linearly transform the target feature vector and the vectors corresponding to each financial product to obtain a first vector corresponding to the target feature vector and a second vector corresponding to the financial product;
[0132] Based on a second preset formula, the first vector, and the second vector, a similarity calculation is performed to obtain the similarity score.
[0133] In one embodiment, the first preset formula is:
[0134]
[0135]
[0136] in, represents the weighted target feature vector; Represents the vector corresponding to the financial product; W h represents the weight matrix; b h is a bias vector; h1 and h2 represent the first vector and the second vector respectively; ⊙ represents an element-wise multiplication operation; u is a random Dropout vector used to set some elements to 0;
[0137] The second preset formula is:
[0138]
[0139] Among them, o represents the output value, that is, the similarity score; e represents the base of the natural logarithm.
[0140] In one embodiment, before the processor processes the feature vectors corresponding to each financial product in a preset product database and the data to be processed based on the preset product recommendation model and the data category, determines the target financial product, and recommends the target financial product to the user, it is further configured to implement:
[0141] Get training data set and validation data set;
[0142] Based on a preset pre-training model, the training data set is processed to obtain predicted recommended products;
[0143] Based on the validation data set and the predicted recommended product, obtaining a cross entropy loss function value of the pre-trained model;
[0144] When the cross entropy loss function value is greater than or equal to a preset loss threshold, the pre-trained model is determined as a product recommendation model.
[0145] A computer-readable storage medium is also provided in an embodiment of the present application, wherein the computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and the processor executes the program instructions to implement any one of the financial product recommendation methods based on contrastive learning provided in the embodiments of the present application.
[0146] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiment, such as a hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a smart memory card (Smart Medi a Card, SMC), a secure digital (Secure Digital, SD) card, a flash memory card (Flash Card), etc., equipped on the computer device.
[0147] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.
Claims
1. A financial product recommendation method based on contrastive learning, It is characterized in that include: Acquire data to be processed, wherein the data to be processed includes at least one data category; Based on the preset product recommendation model and the data category, the feature vectors corresponding to each financial product in the preset product database and the data to be processed are processed to determine the target financial product, and recommend the target financial product to the user.
2. The financial product recommendation method based on contrastive learning according to claim 1, It is characterized in that The method of processing the feature vectors corresponding to each financial product in a preset product database and the data to be processed based on the preset product recommendation model and the data category to determine the target financial product includes: Based on the heterogeneous data feature extraction layer of the product recommendation model, the data to be processed is processed to obtain at least one initial feature vector; Based on the feature conversion layer of the product recommendation model, processing at least one of the initial feature vectors to obtain a target feature vector; Based on the vector similarity calculation layer of the product recommendation model, similarity calculation is performed between the target feature vector and the feature vectors corresponding to each financial product in the product database to obtain a similarity score; Based on the similarity score, the target financial product is determined in the product database.
3. The financial product recommendation method based on contrastive learning according to claim 2, It is characterized in that The heterogeneous data feature extraction layer based on the product recommendation model processes the data to be processed to obtain at least one initial feature vector, including: Based on the heterogeneous data feature extraction layer, the data to be processed are classified to obtain data sets corresponding to each data category; Acquire a feature extraction model corresponding to each of the data categories; Based on the feature extraction model corresponding to each of the data categories, feature extraction is performed on the data set corresponding to each of the data categories to obtain at least one of the initial feature vectors.
4. The financial product recommendation method based on contrastive learning according to claim 2, It is characterized in that The feature conversion layer based on the product recommendation model processes at least one of the initial feature vectors to obtain a target feature vector, including: Based on the feature conversion layer, performing feature conversion on all the initial feature vectors to obtain at least one feature to be weighted; Based on the attention mechanism, weighted processing is performed on at least one of the features to be weighted to obtain the target feature vector.
5. The financial product recommendation method based on contrastive learning according to claim 2, It is characterized in that The vector similarity calculation layer based on the product recommendation model performs similarity calculation on the target feature vector and the feature vectors corresponding to each financial product in the product database to obtain a similarity score, including: Based on the similarity calculation layer and the first preset formula, regularize and linearly transform the target feature vector and the vectors corresponding to each financial product to obtain a first vector corresponding to the target feature vector and a second vector corresponding to the financial product; Based on a second preset formula, the first vector, and the second vector, a similarity calculation is performed to obtain the similarity score.
6. The financial product recommendation method based on contrastive learning according to claim 5, It is characterized in that The first preset formula is: in, represents the weighted target feature vector; Represents the vector corresponding to the financial product; W h represents the weight matrix; b h is the bias vector; h 1 、h 2 denote the first vector and the second vector respectively; ⊙ denotes an element-wise multiplication operation; u is a random Dropout vector used to set some elements to 0; The second preset formula is: Among them, o represents the output value, that is, the similarity score; e represents the base of the natural logarithm.
7. The financial product recommendation method based on contrastive learning according to any one of claims 1 to 6, It is characterized in that Before processing the feature vectors corresponding to each financial product in the preset product database and the data to be processed based on the preset product recommendation model and the data category, determining the target financial product, and recommending the target financial product to the user, the method further includes: Get training data set and validation data set; Based on a preset pre-training model, the training data set is processed to obtain predicted recommended products; Based on the validation data set and the predicted recommended product, obtaining a cross entropy loss function value of the pre-trained model; When the cross entropy loss function value is greater than or equal to a preset loss threshold, the pre-trained model is determined as a product recommendation model.
8. A financial product recommendation device based on contrastive learning, It is characterized in that include: A data acquisition module, used to acquire data to be processed, wherein the data to be processed includes at least one data category; The product determination module is used to process the feature vectors corresponding to each financial product in the preset product database and the data to be processed based on the preset product recommendation model and the data category, determine the target financial product, and recommend the target financial product to the user.
9. A computer device, It is characterized in that The computer device includes a memory and a processor; The memory is used to store computer programs; The processor is used to execute the computer program and implement the financial product recommendation method based on contrastive learning as described in any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements the financial product recommendation method based on contrastive learning as described in any one of claims 1 to 7.