Financial product recommendation method and device, storage medium and program product
By acquiring characteristic and historical data of users and products and constructing a full-sample-space multi-task network, we solved the problem of sparse training samples in deep learning networks and improved the accuracy and efficiency of financial product recommendations.
Patent Information
- Application Number
- CN202511107684.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-08
- Publication Date
- 2025-09-19
AI Technical Summary
In existing technologies, the training samples of deep learning networks in financial product recommendations are relatively sparse, resulting in poor training results and difficulty in achieving refined user-product matching.
By obtaining user feature data, product feature data, historical user reach intention data and historical product conversion data, we generate training sets and test sets, use the full sample space multi-task network for iterative training, build a financial product recommendation model, and output the matching degree between the target user and each financial product.
It improves the accuracy and efficiency of financial product recommendations. By expanding network training samples, it improves the reliability of ESMM network training results and achieves precise matching of users and products.
Smart Images
Figure CN120672431A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of computer data processing technology, is applicable to the field of financial technology, and in particular to a financial product recommendation method, device, storage medium and program product. Background Art
[0002] Currently, in the financial product recommendation scenario, potential product audiences can be identified by mining user characteristics, or users and products can be matched based on expert rules or data analysis methods. These methods are difficult to provide more refined two-way matching results between individuals and products, and require too much manual analysis process.
[0003] In the existing technology, recommendation strategies can be formulated quickly and directly based on deep learning networks. However, existing deep learning networks are usually trained with data samples of users successfully using products. Such sample data is relatively small and sparse, resulting in poor network training results. Summary of the Invention
[0004] The present invention provides a financial product recommendation method, device, storage medium and program product, which can realize accurate financial product recommendation for users and improve the efficiency of financial product recommendation.
[0005] According to one aspect of the present invention, a method for recommending financial products is provided, the method comprising:
[0006] Obtain user feature data, product feature data, historical user reach intention data, and historical product conversion data, and generate a training set and a test set based on the user feature data, product feature data, historical user reach intention data, and historical product conversion data;
[0007] Iteratively training the full sample space multi-task network using the training set and the test set to obtain a financial product recommendation model;
[0008] The target user feature data and product feature data to be recommended are input into the financial product recommendation model, and the matching degree between the target user and each financial product is output through the financial product recommendation model.
[0009] According to one aspect of the present invention, a financial product recommendation device is provided, the device comprising:
[0010] A data acquisition module is used to acquire user feature data, product feature data, historical user reach intention data, and historical product conversion data, and generate a training set and a test set based on the user feature data, product feature data, historical user reach intention data, and historical product conversion data;
[0011] A network training module, configured to iteratively train the full-sample-space multi-task network using the training set and the test set to obtain a financial product recommendation model;
[0012] The matching degree determination module is used to input the target user feature data and product feature data to be recommended into the financial product recommendation model, and output the matching degree between the target user and each financial product through the financial product recommendation model.
[0013] According to another aspect of the present invention, an electronic device is provided, comprising:
[0014] at least one processor; and
[0015] a memory communicatively connected to the at least one processor; wherein,
[0016] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can perform the financial product recommendation method according to any embodiment of the present invention.
[0017] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the financial product recommendation method according to any embodiment of the present invention when executed.
[0018] According to another aspect of the present invention, a computer program product is provided. The computer program product includes a computer program. When the computer program is executed by a processor, the financial product recommendation method according to any embodiment of the present invention is implemented.
[0019] The technical solution provided by the embodiment of the present invention generates a training set and a test set by acquiring user feature data, product feature data, historical user reach intention data and historical product conversion data, and uses the training set and the test set to iteratively train the ESMM network to obtain a financial product recommendation model. The target user feature data and product feature data to be recommended are input into the financial product recommendation model, and the matching degree between the target user and each financial product is output through the financial product recommendation model. This technical means can solve the problem in the prior art that the deep learning network training samples are relatively sparse, resulting in poor training effect. By introducing historical user reach intention data, the network training samples can be effectively expanded, the reliability of the ESMM network training results can be improved, and the precise financial product recommendations for users can be achieved, thereby improving the recommendation efficiency of financial products.
[0020] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present invention, nor is it intended to limit the scope of the present invention. Other features of the present invention will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0022] Figure 1 is a flowchart of a financial product recommendation method provided according to an embodiment of the present invention;
[0023] Figure 2a is a flowchart of another financial product recommendation method provided according to an embodiment of the present invention;
[0024] Figure 2b 1 is a schematic diagram of the structure of an ESMM network provided according to an embodiment of the present invention;
[0025] Figure 3 2 is a schematic structural diagram of a financial product recommendation device provided according to an embodiment of the present invention;
[0026] Figure 4 2 is a schematic diagram of the structure of an electronic device for implementing the financial product recommendation method according to an embodiment of the present invention. DETAILED DESCRIPTION
[0027] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.
[0028] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0029] Figure 1 This is a flowchart of a financial product recommendation method provided by an embodiment of the present invention. This embodiment is applicable to recommending financial products to matching users. The method can be executed by a financial product recommendation device, which can be implemented in the form of hardware and / or software and can be configured in an electronic device. Figure 1 As shown, the method includes:
[0030] Step 110: Obtain user feature data, product feature data, historical user reach intention data, and historical product conversion data, and generate a training set and a test set based on the user feature data, product feature data, historical user reach intention data, and historical product conversion data.
[0031] In this embodiment, user feature data, such as user attribute information and behavior data, can be extracted from the financial product user database, and product feature data, such as product categories or tags, can be extracted from the financial product database. Specifically, the financial product can be a credit card or other product, which is not limited in this embodiment.
[0032] The historical user reach intention data refers to the data on historical users' willingness to use the reached products. The historical product conversion data refers to the data on successful product recommendations to historical users.
[0033] Optionally, detailed data of each reach channel and embedded point data can be obtained from the database to determine the reach intention of historical users, thereby obtaining historical user reach intention data. The historical user reach intention data can include the correspondence between historical users and their willingness to use different products.
[0034] In a specific embodiment, the embedded data may include outbound call contact duration data, click data on recommended products on the mobile client, and click data on links in the SMS contact portal, etc. If the outbound call connection duration exceeds a preset duration threshold (for example, 1 minute), it can be determined that the user has the intention to use the product; if the user clicks on the relevant product on the mobile client, it can be determined that the user has the intention to use the product; if the user accesses the product details page through the SMS portal, it can be determined that the user has the intention to use the product.
[0035] Step 120: Use the training set and the test set to iteratively train the Entire Space Multi-Task Model (ESMM) to obtain a financial product recommendation model.
[0036] In this embodiment, an ESMM network for bidirectional matching of users and products can be constructed. Specifically, multiple samples in the training set can be input into the ESMM network in sequence, and the training parameters can be continuously optimized and adjusted according to the network output results. Then, the samples in the test set can be input into the ESMM network, and the network performance can be verified according to the network output results.
[0037] If the network performance verification is successful, the network training is determined to be completed, and the trained ESMM is used as the financial product recommendation model; if the network performance verification fails, the training parameters of the network are continuously optimized and adjusted until the network performance verification is successful.
[0038] Step 130: Input the target user feature data and product feature data to be recommended into the financial product recommendation model, and output the matching degree between the target user and each financial product through the financial product recommendation model.
[0039] In this embodiment, after the financial product recommendation model outputs the matching degree between the target user and each financial product, it can assist relevant personnel or equipment to quickly and effectively recommend financial products to the target user based on the matching degree.
[0040] In this embodiment, by obtaining historical user reach intention data and historical product conversion data as the training set and test set of the ESMM network, the reach intention intermediate task can be introduced into the multi-task structure of the ESMM network, and the reach intention intermediate data can be fully utilized to enrich the network's training data samples and improve the effectiveness of the network training results.
[0041] The technical solution provided by the embodiment of the present invention generates a training set and a test set by acquiring user feature data, product feature data, historical user reach intention data and historical product conversion data, and uses the training set and the test set to iteratively train the ESMM network to obtain a financial product recommendation model. The target user feature data and product feature data to be recommended are input into the financial product recommendation model, and the matching degree between the target user and each financial product is output through the financial product recommendation model. This technical means can solve the problem in the prior art that the deep learning network training samples are relatively sparse, resulting in poor training effect. By introducing historical user reach intention data, the network training samples can be effectively expanded, the reliability of the ESMM network training results can be improved, and the precise financial product recommendations for users can be achieved, thereby improving the recommendation efficiency of financial products.
[0042] Figure 2a A flowchart of another financial product recommendation method provided by an embodiment of the present invention is applied to a server, such as Figure 2a As shown, the method includes:
[0043] Step 210: Obtain user feature data, product feature data, historical user reach intention data, and historical product conversion data, and concatenate the user feature data and product feature data to obtain a user feature vector and a product feature vector.
[0044] In one implementation of this embodiment, obtaining user characteristic data and product characteristic data includes: obtaining original user data and original product data from a user database and a product database, respectively; deleting abnormal data in the original user data and original product data, and converting continuous data into discrete data; and numbering discrete values in the processed user data and product data to obtain user characteristic data and product characteristic data.
[0045] In this embodiment, the raw user data may include user attribute label data (e.g., gender, age, etc.), user behavior data regarding financial products within a preset time period (e.g., user consumption behavior data for the past six months), etc. The raw product data may include basic product information data (e.g., product category label, channel label), product popularity data, etc. After obtaining the raw user data and raw product data, abnormal data may be deleted, and continuous data may be bucketed so that continuous data within a certain range has the same bucket value, thereby converting the continuous data into discrete data.
[0046] In this embodiment, specifically, in order to improve the data storage efficiency and the subsequent network model calculation efficiency, all the processed discrete forms of user data and product data can be sequentially numbered, and then the numbered user features and product features are spliced according to the features to obtain the user feature vector X u and product feature vector X i :
[0047]
[0048] Among them, d u is the total number of user features, d i is the total number of product features, X u and X i The component values are the eigenvalue numbers of the corresponding features.
[0049] The advantage of this setting is that by processing the acquired original user data and original product data to obtain feature data, it is convenient for the subsequent network model to process the input data, thereby improving the training efficiency of the network model.
[0050] Step 220: Generate a reach intention dataset based on the historical user reach intention data, user feature vectors, and product feature vectors; generate a reach conversion dataset based on the historical product conversion data, user feature vectors, and product feature vectors.
[0051] In this step, specifically, the reach intention dataset can be in the form of (X u ,X i ,Y l ), where Y l Represents a data label, with a value of 1 or 0, representing whether the user has the intention to use the product. The form of the reach conversion data set can be (X u ,X i ,Y v ), where Y v Represents a data label, with a value of 1 or 0, representing whether the product recommendation is successful or not.
[0052] Step 230: Concatenate the reach intention dataset and the reach conversion dataset to obtain a training set and a test set.
[0053] In this step, the above-mentioned reach intention dataset and reach conversion dataset can be spliced to obtain the training set and test set, which are in the form of (X u ,X i ,Y l ,Y v ).
[0054] The advantage of this setting is that by introducing intermediate data on reach intention into the training set and test set, the effective samples corresponding to the network model can be expanded, thereby improving the training efficiency and quality of the network model.
[0055] Step 240: Use the training set and the test set to iteratively train the ESMM network to obtain a financial product recommendation model.
[0056] In this embodiment, Figure 2b It can be a structural diagram of an ESMM network, such as Figure 2b As shown, the ESMM network includes a feature embedding layer, a reach intention network, and a reach intention conversion network. The feature embedding layer includes a user feature embedding layer and a product feature embedding layer. The reach intention network and the reach intention conversion network each deploy a dual-tower network, consisting of a user tower and a product tower. The reach intention network and the reach intention conversion network have similar structures but different network parameters. The user tower and product tower represent users and products, respectively, using low-dimensional dense vectors.
[0057] Optionally, the reach intention network and the reach intention conversion network are constructed based on deep structured semantic models (DSSM), and the two networks share the underlying user feature embedding layer and product feature embedding layer. The reach intention network is used to output the reach intention rate based on the input sample, and the reach intention conversion network is used to output the reach intention conversion rate based on the input sample. Finally, the reach intention rate is multiplied by the reach intention conversion rate to obtain the reach conversion rate corresponding to the input sample, and the network parameters are optimized and adjusted according to this reach conversion rate, ultimately forming an ESMM network that matches users and products in both directions.
[0058] The advantage of this setting is that this embodiment provides an ESMM network with a simple structure for bidirectional matching of user products, which can be quickly implemented in different application scenarios and different data sets. By utilizing the strong feature extraction and expression capabilities of the deep network, the robustness of the network training results can be improved.
[0059] In a specific embodiment, the user feature vector X obtained in the above steps can be u and product feature vector X i After passing through the user feature embedding layer and the product feature embedding layer respectively, the user feature embedding vector is obtained and product feature embedding vector This embedding vector is then continuously fed into two fully connected layers, and finally outputs the user representation vector and the product representation vector. The output of the first fully connected layer can be determined by the following expression:
[0060]
[0061] The output of the second fully connected layer can be determined by the following expression:
[0062]
[0063] Where: W and Z are the preset weight parameters, b and t are the preset biases, and f is the activation function. and is the user representation vector, and is the product characterization vector, and have the same dimensions, and have the same dimensions.
[0064] After obtaining the user representation vector and product representation vector, the user representation vector and product representation vector can be mapped to the corresponding vector space through the dual-tower network in the reach intention network and the reach intention conversion network, and the matching degree between the specific user and the specific product can be calculated in this vector space. Specifically, and Perform inner product calculation and and Perform inner product calculation, and then use sigmoid function to obtain the touch willingness rate p l and the conversion rate of willingness to reach s ,Right now:
[0065]
[0066] Finally, p l With p s Multiply to get the touch conversion rate p corresponding to the input sample v ,Right now:
[0067] p v =p l ×p s
[0068] In one implementation of this embodiment, the ESMM network is iteratively trained using the training set and the test set to obtain a financial product recommendation model, including:
[0069] Multiple samples in the training set and the test set are input into the ESMM network in sequence, and the reach intention intermediate task loss function and the reach conversion task loss function are constructed; the reach intention intermediate task loss function and the reach conversion task loss function are weightedly summed to obtain the target loss function; in the process of iterative processing of samples by the ESMM network, the target loss function is used to backpropagate and update the network parameters until the target loss function converges.
[0070] In a specific embodiment, the loss function of the intermediate task of reaching willingness can be constructed as loss l And the loss function of the conversion task loss v , the formula is as follows:
[0071]
[0072] Among them, N is the number of training samples in a single batch, y li is the label of the i-th reach intention sample, y vi is the label of the i-th touch conversion sample, p li is the willingness rate of the i-th sample to reach, p vi is the conversion rate of the i-th sample. l and loss v The weighted combination obtains the target loss function loss, namely:
[0073] loss=m l loss l +m v loss v
[0074] Among them, m l and m v is the preset weight coefficient.
[0075] The advantage of this setting is that by constructing the reach intention intermediate task loss function and the reach conversion task loss function, the reach intention network and the reach intention conversion network can be trained quickly and effectively to improve the training efficiency of the financial product recommendation model and the reliability of the training results.
[0076] Step 250: Input the target user feature data and product feature data to be recommended into the financial product recommendation model, and output the reach intention rate between the target user and each financial product through the reach intention network in the financial product recommendation model.
[0077] Step 260: Output the reach intention conversion rate between the target user and each financial product through the reach intention conversion network in the financial product recommendation model; multiply the reach intention rate by the reach intention conversion rate to obtain the matching degree between the target user and each financial product (that is, the reach conversion rate).
[0078] The advantage of this setting is that the matching degree between the target user and each financial product can be determined quickly and effectively, so that relevant personnel or financial software can recommend suitable financial products to the target user based on the matching degree, thereby improving the user's experience of using the financial software.
[0079] In one implementation of this embodiment, after obtaining the matching degree between the target user and each financial product, the matching degree between the target user and each financial product can be sorted from high to low; a product priority list corresponding to the target user is generated based on the sorting result, so that the product recommendation node recommends products to the target user based on the product priority list.
[0080] In this embodiment, the product recommendation node may be a recommendation module in financial software. The advantage of such a setting is that it can provide a refined two-way matching result between users and products, thereby improving the user experience of the financial software.
[0081] In another implementation of this embodiment, the characteristic data of multiple target users to be recommended and the characteristic data of multiple products can also be input into the financial product model to improve the financial product model so that it can output the matching degree between each user and each product, and then generate a corresponding user priority list for each product, so that relevant personnel can determine the potential users suitable for a specific product based on the user priority list.
[0082] The technical solution provided by the embodiment of the present invention obtains user feature data, product feature data, historical user reach intention data and historical product conversion data, and respectively concatenates the user feature data and product feature data to obtain user feature vectors and product feature vectors. A reach intention dataset is generated based on the historical user reach intention data, user feature vectors and product feature vectors; a reach conversion dataset is generated based on the historical product conversion data, user feature vectors and product feature vectors, and the reach intention dataset and the reach conversion dataset are concatenated to obtain a training set and a test set. The ESMM network is iteratively trained using the training set and the test set to obtain a financial product recommendation model. The target user feature data and product feature data to be recommended are input into the financial product recommendation model. The reach intention network in the financial product recommendation model outputs the reach intention rate between the target user and each financial product, and the reach intention conversion network in the financial product recommendation model outputs the reach intention conversion rate between the target user and each financial product. The reach intention rate is multiplied by the reach intention conversion rate to obtain the matching degree between the target user and each financial product. This technical solution can achieve precise financial product recommendations for users and improve the efficiency of financial product recommendations.
[0083] 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 relevant data comply with the relevant laws, regulations and standards of relevant countries and regions, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entrances for users to choose to authorize or refuse.
[0084] Figure 3 This is a schematic diagram of the structure of a financial product recommendation device provided by an embodiment of the present invention, wherein the device is applied to an electronic device, such as Figure 3 As shown, the device includes: a data acquisition module 310, a network training module 320 and a matching degree determination module 330.
[0085] Data acquisition module 310, for acquiring user feature data, product feature data, historical user reach intention data, and historical product conversion data, and generating training and test sets based on the user feature data, product feature data, historical user reach intention data, and historical product conversion data;
[0086] A network training module 320 is configured to iteratively train the ESMM network using the training set and the test set to obtain a financial product recommendation model;
[0087] The matching degree determination module 330 is used to input the target user feature data and product feature data to be recommended into the financial product recommendation model, and output the matching degree between the target user and each financial product through the financial product recommendation model.
[0088] The technical solution provided by the embodiment of the present invention generates a training set and a test set by acquiring user feature data, product feature data, historical user reach intention data and historical product conversion data, and uses the training set and the test set to iteratively train the ESMM network to obtain a financial product recommendation model. The target user feature data and product feature data to be recommended are input into the financial product recommendation model, and the matching degree between the target user and each financial product is output through the financial product recommendation model. This technical means can solve the problem in the prior art that the deep learning network training samples are relatively sparse, resulting in poor training effect. By introducing historical user reach intention data, the network training samples can be effectively expanded, the reliability of the ESMM network training results can be improved, and the precise financial product recommendations for users can be achieved, thereby improving the recommendation efficiency of financial products.
[0089] Based on the above embodiment, the ESMM network includes a feature embedding layer, a reach intention network and a reach intention conversion network; wherein, the feature embedding layer includes a user feature embedding layer and a product feature embedding layer; dual-tower networks are deployed in the reach intention network and the reach intention conversion network respectively; the dual-tower network consists of a user tower and a product tower.
[0090] The data acquisition module 310 includes:
[0091] The original data acquisition unit is used to acquire original user data and original product data from the user database and the product database respectively;
[0092] A data processing unit is used to delete abnormal data in the original user data and the original product data, and convert continuous data into discrete data; number the discrete values in the processed user data and product data to obtain user feature data and product feature data;
[0093] A data set generation unit is used to splice user feature data and product feature data respectively to obtain user feature vectors and product feature vectors; generate a reach intention data set based on the historical user reach intention data, user feature vectors and product feature vectors; generate a reach conversion data set based on the historical product conversion data, user feature vectors and product feature vectors; and splice the reach intention data set and the reach conversion data set to obtain a training set and a test set.
[0094] The network training module 320 includes:
[0095] An iterative training unit is used to input multiple samples in the training set and the test set into the ESMM network in sequence, and construct a reach intention intermediate task loss function and a reach conversion task loss function; perform weighted summation on the reach intention intermediate task loss function and the reach conversion task loss function to obtain a target loss function; in the process of iterative processing of samples by the ESMM network, use the target loss function to backpropagate and update the network parameters until the target loss function converges.
[0096] The matching degree determination module 330 includes:
[0097] The matching degree calculation unit is used to output the reaching intention rate between the target user and each financial product through the reaching intention network in the financial product recommendation model; output the reaching intention conversion rate between the target user and each financial product through the reaching intention conversion network in the financial product recommendation model; and multiply the reaching intention rate by the reaching intention conversion rate to obtain the matching degree between the target user and each financial product.
[0098] The device further comprises:
[0099] The list generation module is used to sort the matching degree between the target user and each financial product from high to low; generate a product priority list corresponding to the target user based on the sorting result, so that the product recommendation node recommends products to the target user based on the product priority list.
[0100] The above device can execute the methods provided by all the above embodiments of the present invention, and has the corresponding functional modules and beneficial effects of executing the above methods. For technical details not fully described in the embodiments of the present invention, please refer to the methods provided by all the above embodiments of the present invention.
[0101] Figure 4 A schematic diagram of the structure of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or claimed herein.
[0102] like Figure 4As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., which is communicatively connected to the at least one processor 11. The memory stores a computer program that can be executed by the at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 into the random access memory (RAM) 13. Various programs and data required for the operation of the electronic device 10 can also be stored in the RAM 13. The processor 11, ROM 12, and RAM 13 are connected to each other via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.
[0103] Multiple components in the electronic device 10 are connected to the I / O interface 15, including an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.
[0104] Processor 11 can be any general-purpose and / or specialized processing component with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any other suitable processor, controller, microcontroller, etc. Processor 11 executes the various methods and processes described above, such as the financial product recommendation method.
[0105] In some embodiments, the financial product recommendation method may be implemented as a computer program tangibly embodied in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the financial product recommendation method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to execute the financial product recommendation method in any other suitable manner (e.g., via firmware).
[0106] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.
[0107] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the computer program is executed by the processor, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The computer program may be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.
[0108] In the context of the present invention, computer-readable storage media can be tangible media that can contain or store a computer program for use with an instruction execution system, device or equipment or used in combination with an instruction execution system, device or equipment. Computer-readable storage media can include but are not limited to electronic, magnetic, optical, electromagnetic, infrared or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, computer-readable storage media can be machine-readable signal media. More specific examples of machine-readable storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.
[0109] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).
[0110] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.
[0111] A computing system may include a client and a server. The client and server are typically remote from each other and typically interact via a communication network. This client-server relationship arises through computer programs running on the respective computers, creating a client-server relationship. The server may be a cloud server, also known as a cloud computing server or cloud host. This server is a hosting product within a cloud computing service ecosystem that addresses the management difficulties and limited scalability of traditional physical hosting and VPS services.
[0112] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in the present invention can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved. This is not limited herein.
[0113] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.
Claims
1. A financial product recommendation method, characterized in that: The method comprises: Obtain user feature data, product feature data, historical user reach intention data, and historical product conversion data, and generate a training set and a test set based on the user feature data, product feature data, historical user reach intention data, and historical product conversion data; Iteratively training the full sample space multi-task network using the training set and the test set to obtain a financial product recommendation model; The target user feature data and product feature data to be recommended are input into the financial product recommendation model, and the matching degree between the target user and each financial product is output through the financial product recommendation model.
2. The method according to claim 1, characterized in that Obtain user feature data and product feature data, including: Obtain original user data and original product data from the user database and the product database respectively; Deleting abnormal data from the original user data and the original product data, and converting continuous data into discrete data; The processed discrete values in the user data and product data are numbered to obtain user feature data and product feature data.
3. The method according to claim 1, characterized in that Generate training and test sets based on the user feature data, product feature data, historical user reach intention data, and historical product conversion data, including: Concatenate user feature data and product feature data to obtain user feature vectors and product feature vectors; Generate a reach intention dataset based on the historical user reach intention data, user feature vectors, and product feature vectors; Generate a reach conversion data set based on the historical product conversion data, user feature vectors, and product feature vectors; The reach intention dataset and the reach conversion dataset are spliced to obtain a training set and a test set.
4. The method according to claim 1, wherein The full sample space multi-task network includes a feature embedding layer, a reach intention network, and a reach intention conversion network; Wherein, the feature embedding layer includes a user feature embedding layer and a product feature embedding layer; A dual-tower network is deployed in the reach intention network and the reach intention conversion network respectively; the dual-tower network consists of a user tower and a product tower.
5. The method according to claim 1, wherein The full sample space multi-task network is iteratively trained using the training set and the test set to obtain a financial product recommendation model, including: Input multiple samples in the training set and the test set into the full sample space multi-task network in sequence, and construct a reach intention intermediate task loss function and a reach conversion task loss function; Performing a weighted summation on the reach intention intermediate task loss function and the reach conversion task loss function to obtain a target loss function; In the process of iteratively processing samples by the full sample space multi-task network, the target loss function is used to back-propagate and update the network parameters until the target loss function converges.
6. The method according to claim 4, characterized in that Outputting the matching degree between the target user and each financial product through the financial product recommendation model includes: Outputting the reach intention rate between the target user and each financial product through the reach intention network in the financial product recommendation model; Outputting the reach intention conversion rate between the target user and each financial product through the reach intention conversion network in the financial product recommendation model; The degree of match between the target user and each financial product is obtained by multiplying the reach intention rate by the reach intention conversion rate.
7. The method according to claim 1, characterized in that After the financial product recommendation model outputs the matching degree between the target user and each financial product, the method further includes: Sort the matching degree between the target user and each financial product from high to low; A product priority list corresponding to the target user is generated according to the ranking result, so that the product recommendation node recommends products to the target user according to the product priority list.
8. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to perform the financial product recommendation method according to any one of claims 1 to 7.
9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the financial product recommendation method according to any one of claims 1 to 7 when executed.
10. A computer program product, characterized in that The computer program product comprises a computer program, which, when executed by a processor, implements the financial product recommendation method according to any one of claims 1 to 7.