Service recommendation method and device, computer equipment and storage medium

Abstract

The embodiment of the invention belongs to the field of artificial intelligence and financial self-service terminals, and relates to a service recommendation method, which comprises the following steps: acquiring user data to be processed, determining a target user group based on facial recognition data, determining weblog data of the target user group based on user behavior data, and recommending the weblog data to the target user group. And based on combination of the target user group and the weblog data, constructing a target user portrait, importing the target user portrait into a pre-trained service recommendation model, and performing push processing on the target user group through the pre-trained service recommendation model to obtain a target push service. The invention further provides a service recommendation device, computer equipment and a storage medium. According to the method, highly personalized service recommendation can be realized, the learning cost and selection difficulty of the user are reduced, the correlation of service recommendation and the user satisfaction are remarkably improved, meanwhile, the user experience is optimized, and the service efficiency is improved.

Description

Technical Field

[0001] The present application relates to the technical field of artificial intelligence and financial self-service terminals, and in particular to a service recommendation method, device, computer equipment and storage medium. Background Art

[0002] With the intelligent and digital development of banking services, more and more local banks have introduced smart ATMs, which can be used for self-service deposits and withdrawals, bill inquiries and other services. However, for elderly users who are accustomed to going to offline business outlets to handle business, there is a problem of low proficiency in intelligent operations. These users often encounter difficulties in choosing services when using smart ATMs.

[0003] The traditional solution to the difficulties encountered by elderly users in using intelligent banking services is for banks to provide on-site customer service staff to help elderly users understand and use smart ATMs or other self-service equipment, as well as to simplify the interfaces of smart ATMs and online services to make operations more intuitive and easy to understand.

[0004] However, this approach has high demands on human resources, increases the bank's operating costs, and fails to fully understand the specific needs and preferences of each user, resulting in inaccurate service recommendations. Summary of the invention

[0005] The purpose of the embodiments of the present application is to propose a service recommendation method, apparatus, computer device and storage medium to solve the technical problem that the specific needs and preferences of each user cannot be fully understood, resulting in inaccurate service recommendations.

[0006] In order to solve the above technical problems, the present application embodiment provides a service recommendation method, which adopts the following technical solution:

[0007] Acquire user data to be processed, wherein the user data to be processed includes facial recognition data and user behavior data;

[0008] determining a target user group based on the facial recognition data, and determining web log data of the target user group based on the user behavior data;

[0009] Building a target user profile based on combining the target user group with the network log data;

[0010] The target user portrait is imported into a pre-trained service recommendation model, and push processing is performed on the target user group through the pre-trained service recommendation model to obtain a target push service.

[0011] Furthermore, the step of determining the target user group based on the facial recognition data specifically includes:

[0012] Performing face detection on the face recognition data by using a Haar cascade detector to obtain a target face image;

[0013] Extracting features of the target facial image based on the model fitting method to obtain target facial features;

[0014] A user group corresponding to the target facial feature is obtained from a preset facial feature library to determine the target user group.

[0015] Furthermore, the step of inputting the target facial features into a preset facial feature library for comparison to determine the target user group specifically includes:

[0016] Performing principal component analysis on the target facial features to obtain a facial descriptor;

[0017] The user group corresponding to the facial descriptor is obtained from a preset facial feature library to determine the target user group. Furthermore, before the step of importing the target user portrait into a pre-trained service recommendation model and performing push processing on the target user group through the pre-trained service recommendation model to obtain the target push service, the method specifically includes:

[0018] Acquire a recommended training data set, wherein the recommended training data set is composed of a plurality of the user behavior data;

[0019] Performing feature encoding on the user behavior data and network log data to obtain push data features;

[0020] Estimating the push data features to obtain model parameters of a service recommendation model;

[0021] Building a self-attention network function, and embedding the self-attention network function into a hidden layer of a preset general model;

[0022] The model parameters of the service recommendation model are imported into the general model, and the general model is trained using the recommended training data set to obtain the pre-trained service recommendation model.

[0023] Furthermore, the pre-trained service recommendation model includes an input layer, a processing layer and an output layer. The step of performing push processing on the target user group through the pre-trained service recommendation model to obtain the target push service specifically includes:

[0024] Performing feature analysis on the target user portrait through the input layer to obtain a target feature vector;

[0025] Calculating the push degree of the target feature vector by the processing layer to obtain the service push score, and determining the target push service based on the service push score;

[0026] The target push service is output through the output layer.

[0027] Furthermore, the step of calculating the push degree of the target feature vector by the processing layer to obtain the service push score, and determining the target push service based on the service push score specifically includes:

[0028] Get service options;

[0029] The target feature vector and the service options are calculated by a collaborative filtering algorithm to obtain the service push score, where one service push score corresponds to one service option;

[0030] The service push scores are sorted, and the service option with the highest service push score is selected as the target push service.

[0031] Furthermore, after the step of performing push processing on the target user group through the pre-trained service recommendation model to obtain the target push service, the method further includes:

[0032] Obtain user feedback data;

[0033] Based on the target push service and the user feedback data, a back propagation algorithm is used to perform fitting to obtain a prediction error;

[0034] The prediction error is compared with a preset threshold. If the prediction error is greater than the preset threshold, the pre-trained service recommendation model is iteratively updated until the prediction error is less than or equal to the preset threshold.

[0035] In order to solve the above technical problems, the embodiment of the present application further provides a service recommendation device, which adopts the following technical solution:

[0036] An acquisition module, used to acquire user data to be processed, wherein the user data to be processed includes facial recognition data and user behavior data;

[0037] a determination module, configured to determine a target user group based on the facial recognition data, and to determine network log data of the target user group based on the user behavior data;

[0038] A construction module, used to construct a target user portrait based on the target user group and the network log data;

[0039] The processing module imports the target user portrait into a pre-trained service recommendation model, performs push processing on the target user group through the pre-trained service recommendation model, and obtains a target push service.

[0040] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0041] A computer device comprises a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the service recommendation method when executing the computer-readable instructions.

[0042] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0043] A computer-readable storage medium, characterized in that computer-readable instructions are stored on the computer-readable storage medium, and when the computer-readable instructions are executed by a processor, the steps of the service recommendation method are implemented.

[0044] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0045] The embodiment of the present application obtains user data to be processed, determines the target user group based on the facial recognition data, and determines the network log data of the target user group based on the user behavior data, constructs a target user portrait based on the target user group and the network log data, imports the target user portrait into a pre-trained service recommendation model, and pushes the target user group through the pre-trained service recommendation model to obtain a target push service. By combining facial recognition and user behavior data, a detailed user portrait is constructed and combined with historical network log data, the needs and preferences of the target user group can be accurately identified and analyzed, the accuracy of the service and user satisfaction are improved, and the user experience is optimized. For user groups who are not very proficient in use, it is more convenient to obtain banking services suitable for them through smart ATMs, thereby improving the efficiency and convenience of intelligent services. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. 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.

[0047] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0048] Figure 2 is a flow chart of an embodiment of a service recommendation method according to the present application;

[0049] Figure 3 is a structural diagram of an embodiment of a service recommendation device according to the present application;

[0050] Figure 4 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0051] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0052] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0053] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0054] like Figure 1 As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0055] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0056] Terminal devices 101, 102, 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, desktop computers, etc.

[0057] The server 105 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal devices 101 , 102 , and 103 .

[0058] It should be noted that the service recommendation method provided in the embodiment of the present application is generally executed by a server / terminal device, and accordingly, the service recommendation device is generally set in the server / terminal device.

[0059] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to the implementation requirements.

[0060] Continue to refer Figure 2 , shows a flow chart of an embodiment of a method for service recommendation according to the present application. The service recommendation method comprises the following steps:

[0061] Step S201, obtaining user data to be processed.

[0062] In this embodiment, the service recommendation method can be deployed in a smart ATM. The smart ATM can include several business functions, one service option corresponds to one business function, and the business functions can include service options such as deposits, withdrawals, account management, loans, and credit card services. The smart ATM can be constructed by a server or a server cluster. The server or server cluster can be any electronic device with image recognition, image processing, data transmission, and data storage functions. The service recommendation method runs on the electronic device (e.g. Figure 1The server / terminal device shown in the figure can obtain the user data to be processed through a wired connection or a wireless connection. It should be noted that the above wireless connection method may include but is not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future.

[0063] In this embodiment, the user data to be processed includes facial recognition data and user behavior data. Specifically, the facial recognition data may be a user's facial image captured by a camera on a smart ATM, and the user behavior data may be the user's operation data when using the smart ATM, such as selected services, interactive interfaces, input commands, etc.

[0064] Step S202, determining a target user group based on the facial recognition data, and determining network log data of the target user group based on the user behavior data.

[0065] In this embodiment, the target user group can be divided into an elderly user group and a general user group. Specifically, the elderly user group can be users over 55 years old, and the general user group can be users between 18 and 55 years old. The target user group can include user registration information, and the user registration information can be basic information such as the user's name, age, gender, and occupation.

[0066] In this embodiment, the above-mentioned network log data can be the access log of the above-mentioned target user group on the online banking, mobile banking or website collected by the above-mentioned smart ATM through the bank's internal system. The above-mentioned access log records the various operations of the user on the above-mentioned online banking, mobile banking or website (that is, the above-mentioned user behavior data). The above-mentioned access log may include the user's browsing history, click behavior and page dwell time.

[0067] Specifically, the above browsing history can be used to reveal the user's interest in and attention to different banking services and functions. For example, within a certain period of time, the higher the number of visits to the wealth management product page by the user in the online banking, the higher the user's interest in and attention to the wealth management service; the above clicking behavior can be used to understand the user's preferences for specific functions and services. For example, when the number of clicks on the transfer function by the user is greater than the preset click threshold, it indicates that the user has a high demand for the transfer function; when the number of clicks on the loan service by the user is less than the preset click threshold, it indicates that the user has a low demand for the loan service; the above page dwell time can be used to infer the user's interest in the current page content. For example, when the user's dwell time on the introduction page of the loan service is greater than the preset dwell time threshold, and the page's static time is less than the preset static time threshold, it indicates that the user is interested in the content of the loan service and spends more time reading the relevant information.

[0068] Specifically, regarding the specific implementation process of determining the target user group based on facial recognition data, this application will provide further detailed descriptions in subsequent specific embodiments, and will not be elaborated on here.

[0069] Step S203, constructing a target user portrait based on the target user group and the network log data.

[0070] In this embodiment, the target user group can be classified according to preset age ranges through machine learning algorithms such as K-means, DBSCAN or hierarchical clustering, and the target user group can be labeled to obtain elderly user labels and ordinary user labels. The elderly user label includes the elderly user group, and the ordinary user label includes the ordinary user group.

[0071] In a possible embodiment, in the case where the target user group is an ordinary user group, the browsing history, click behavior, page dwell time and other data of the users in the above-mentioned network log data are converted into numerical features (for example, the click frequency, number of visits and page dwell time in the network log data are standardized into scores or weights), and a feature vector of the ordinary user group is constructed based on the above-mentioned numerical features and the user registration information such as the name, age, gender, occupation and other user information of the users in the above-mentioned ordinary user group. The feature vector of the above-mentioned ordinary user group is imported into the above-mentioned ordinary user tag as a user behavior feature to obtain the above-mentioned target user portrait. The feature vector of the above-mentioned ordinary user group can be used to reflect the behavior pattern and service preference of the above-mentioned ordinary user group.

[0072] In a possible embodiment, in the case where the target user group is an elderly user group, the browsing history, click behavior, page dwell time and other data of the users in the above-mentioned network log data are converted into numerical features (for example, the click frequency, number of visits and page dwell time in the network log data are standardized into scores or weights), and at the same time, the transaction data of the above-mentioned elderly user group is obtained. The above-mentioned transaction data may include transaction frequency, consumption category, common payment methods, balance of payments and savings rate and other data. Based on the above-mentioned numerical features, transaction data and user registration information such as the name, age, gender, occupation and other user registration information of the users in the above-mentioned elderly user group, a feature vector of the elderly user group is constructed, and the feature vector of the above-mentioned elderly user group is imported into the above-mentioned elderly user tag as a user behavior feature to obtain the above-mentioned target user portrait. The feature vector of the above-mentioned elderly user group can be used to reflect the behavior pattern and service preference of the above-mentioned elderly user group.

[0073] Specifically, after obtaining the above facial recognition data, the smart teller machine can determine the user account based on the above facial recognition data, and collect transaction data such as transaction frequency, consumption category, commonly used payment methods, balance of payments and savings rate of the user account from the bank's internal system. More specifically, by counting the transaction frequency of the user, the user's consumption habits and behavioral patterns can be understood. For example, users who frequently conduct transactions prefer flexible financial management methods, while users who conduct fewer transactions tend to be more conservative financial management strategies; through the consumption category information in the user's transaction data, the user's consumption preferences and interests can be understood. For example, if the user frequently consumes in the fields of catering, travel, shopping, etc., it can be inferred that the user has a high interest and consumption demand in these fields; by analyzing the user's commonly used payment methods, the user's preference for different payment methods can be understood. For example, users who tend to use credit card payments pay more attention to points and discounts, while users who like to use mobile payments pay more attention to convenience and speed; by observing the user's balance of payments and savings rate, their financial goals can be inferred. For example, if the user has a high savings rate and a good balance of payments, it means that their financial goals are to accumulate savings, plan for the future, or achieve a specific goal, such as buying a house, educating children, etc.

[0074] Step S204, importing the target user portrait into the pre-trained service recommendation model, performing push processing on the target user group through the pre-trained service recommendation model, and obtaining the target push service.

[0075] In this embodiment, the pre-trained service recommendation model model can be a model based on the self-attention mechanism to establish a user recommendation model, and the target push service can include retirement planning and insurance product services, investment and savings plan services, discount and promotion notification services, and security and fraud protection services and other personalized financial services.

[0076] Specifically, regarding the above-mentioned importing of the target user portrait into a pre-trained service recommendation model, pushing processing to the target user group through the pre-trained service recommendation model, and obtaining the specific implementation process of the target push service, this application will provide further detailed descriptions of this in subsequent specific embodiments and will not be elaborated in detail here.

[0077] The present application obtains the user data to be processed, determines the target user group based on the facial recognition data, and determines the network log data of the target user group based on the user behavior data, constructs a target user portrait based on the combination of the target user group and the network log data, imports the target user portrait into a pre-trained service recommendation model, and pushes the target user group through the pre-trained service recommendation model to obtain a target push service. By combining facial recognition and user behavior data, a detailed user portrait is constructed and combined with historical network log data, the needs and preferences of the target user group can be accurately identified and analyzed, the accuracy of the service and user satisfaction are improved, and the user experience is optimized. For user groups who are not very proficient in use, it is more convenient to obtain banking services suitable for them through smart ATMs, thereby improving the efficiency and convenience of intelligent services.

[0078] In some optional implementations, step S201 includes the following steps:

[0079] Perform facial detection on facial recognition data through Haar cascade detector to obtain the target facial image;

[0080] Based on the model fitting method, feature extraction is performed on the target facial image to obtain the target facial features;

[0081] A user group corresponding to the target facial features is obtained from a preset facial feature library to determine the target user group.

[0082] In this embodiment, the Haar cascade detector can be a target detection algorithm based on Haar features and cascade structure, which is mainly used for detecting target objects in images. Specifically, the Haar cascade detector can perform fast object detection on the facial recognition data by using Haar features as feature sets to generate a target facial image. The Haar cascade detector is mainly used in real-time video processing, and can quickly and accurately identify the facial area in the image, that is, identify the existence and position of the user's face, and obtain the target facial image. The target facial image can be the user's face data.

[0083] In this embodiment, the above-mentioned model fitting method can be an Active Shape Model (ASM) and an Active Appearance Model (AAM) algorithm. The above-mentioned ASM can effectively locate the positions of facial features such as eyes, nose and mouth by using a point distribution model (PDM) to represent the shape of the object. The above-mentioned AAM is an extension of ASM. The above-mentioned AAM not only considers the shape of the target facial image, but also considers the texture. The above-mentioned AAM can extract features of the target facial image by establishing a statistical model of the relationship between facial shape and texture to obtain the above-mentioned target facial features.

[0084] In this embodiment, the above-mentioned preset facial feature library can be a database in the bank's internal system for storing and managing known user facial data. The above-mentioned user facial data can be captured, analyzed and saved by the above-mentioned smart teller machine when the user registers a bank account for the first time or performs a facial scan. The above-mentioned preset facial feature library can provide a reference benchmark during the facial recognition process of the smart teller machine, so that the smart teller machine can compare the above-mentioned facial recognition data with the user facial data in the library to verify the identity of the user.

[0085] In a possible embodiment, facial detection is first performed on the image collected by the above-mentioned smart ATM, that is, a Haar cascade detector is used to identify whether there is a facial area in the image. Once the face is detected, Active Appearance Model (AAM) is used as a model fitting method to extract location points with important semantic information in the facial image, such as the locations of eyes, nose, mouth, etc., to obtain target facial features, and the target facial features are input into a preset facial feature library for comparison to determine the target user group.

[0086] Specifically, regarding the specific implementation process of inputting the target facial features into a preset facial feature library for comparison to determine the target user group, this application will provide further detailed descriptions of this in subsequent specific embodiments, and will not be elaborated on here.

[0087] This application uses a Haar cascade detector for facial detection, then uses a model fitting method to extract facial features, and compares these features with data in a preset facial feature library. It can achieve fast and accurate user identity authentication, improve the security and personalization of services, and is particularly effective in reducing fraud and enhancing user experience in banking transaction environments that require strict identity verification, while improving operational efficiency and the overall reliability of the system.

[0088] In some optional implementations, the above step of "inputting the target facial features into a preset facial feature library for comparison to determine the target user group" includes the following steps:

[0089] Perform principal component analysis on the target facial features to obtain a facial descriptor;

[0090] The user group corresponding to the facial descriptor is obtained in a preset facial feature library to determine the target user group.

[0091] In this embodiment, the principal component analysis (PCA) processing can be a statistical technique for simplifying data while retaining important features in the data set. The PCA can be used to reduce the dimension of facial image data. The use of PCA processing can also help reduce the impact of environmental changes (such as lighting and posture differences) on recognition accuracy, thereby improving processing speed and recognition efficiency.

[0092] The above-mentioned facial descriptor can be a vector used to represent facial features. The above-mentioned facial descriptor can include geometric features such as the position and relative distance of facial key points (such as corners of the eyes, tip of the nose, corners of the mouth, etc.) and appearance features such as skin texture, color, local image gradient and direction.

[0093] In a possible embodiment, in order to implement the above-mentioned principal component analysis processing process, the following steps can be performed: the above-mentioned target facial features are standardized to ensure that each dimension (each pixel of the image) has zero mean and unit variance, the covariance matrix of each dimension is calculated, the correlation between the dimensions in the above-mentioned target facial features is found, the covariance matrix is ​​feature decomposed, the facial eigenvalues ​​and the corresponding facial feature vectors are extracted, and the target facial feature vector (i.e., the above-mentioned facial descriptor) is selected according to the size of the above-mentioned facial eigenvalues. For example, the target facial feature vector with the largest facial eigenvalue is selected as the above-mentioned facial descriptor, and the generated facial descriptor is compared with the user facial data in the preset facial feature library to confirm the identity of the user.

[0094] The present application performs principal component analysis (PCA) on target facial features to generate facial descriptors, and compares these descriptors with data in a preset facial feature library, thereby achieving efficient and accurate user identification and classification, improving the speed and efficiency of facial recognition, and enhancing the system's ability to identify user identities by accurately matching facial features. The use of PCA processing also helps reduce the impact of environmental changes (such as lighting and posture differences) on recognition accuracy, thereby improving the robustness and reliability of the system in a variety of application scenarios.

[0095] In some optional implementations, before step S204, the following steps are further included:

[0096] Get the recommended training data set;

[0097] Perform feature encoding on user behavior data and network log data to obtain push data features;

[0098] Estimate the push data features and obtain the model parameters of the service recommendation model;

[0099] Build a self-attention network function and embed it into the hidden layer of the preset general model;

[0100] The model parameters of the service recommendation model are imported into the general model, and the general model is trained using the recommended training data set to obtain a pre-trained service recommendation model.

[0101] In this embodiment, the above-mentioned recommendation training data set may be composed of a plurality of the above-mentioned user behavior data.

[0102] In this embodiment, the preset general model can be a model preliminarily selected for service recommendation. The model can be downloaded from the Internet by selecting a suitable artificial intelligence model. By training the general model in subsequent steps and adjusting the corresponding parameters of the general model, a relatively ideal service recommendation model can be obtained. The hidden layer is a set of one or more layers of neurons located between the input layer and the output layer. These neurons do not directly interact with the external environment (that is, they do not receive external input or output results to the outside), but they process user behavior data, transmit signals, and perform complex transformations on input data within the network.

[0103] In a possible embodiment, the user's registration information, transaction data and network log data are converted into feature representations (i.e., the above-mentioned push data features) that can be understood by a deep learning model through one-hot encoding, and the above-mentioned push data features are learned through regression analysis, decision trees, neural networks, etc., so as to estimate the model parameters of the service recommendation model, build a self-attention network with a Transformer structure, and embed the self-attention network function into the hidden layer of a preset general model. Finally, the model parameters of the above-mentioned service recommendation model are loaded into a selected general model framework (such as a convolutional neural network, a recurrent neural network, etc.), and the recommended training data set is used to train and optimize the model to obtain a pre-trained service recommendation model.

[0104] This application can deeply understand and analyze users' behavior patterns and preferences by training a service recommendation model, thereby providing more accurate and relevant service recommendations. By adding a self-attention mechanism to the service recommendation model, the service recommendation model is more efficient in processing complex user behavior data, which can improve the accuracy of the service recommendation model's recommended services and user satisfaction.

[0105] In some optional implementations, step S204 includes the following steps:

[0106] Perform feature analysis on the target user portrait through the input layer to obtain the target feature vector;

[0107] The processing layer calculates the push degree of the target feature vector to obtain a service push score, and determines the target push service based on the service push score;

[0108] Output the target push service through the output layer.

[0109] In this embodiment, the pre-trained service recommendation model includes an input layer, a processing layer, and an output layer. Specifically, the input layer of the pre-trained service recommendation model performs feature analysis on the target user portrait to obtain a target feature vector, and the target feature vector is used as the input of the processing layer. The processing layer calculates the push degree of the target feature vector to obtain a service push score. At the same time, the processing layer also determines the target push service based on the service push score, and uses the target push service as the input of the output layer, and outputs the target push service through the output layer.

[0110] In a possible embodiment, the feature analysis may include performing Z-score normalization on the transaction data in the target user portrait to obtain a transaction data feature vector, performing One-Hot Encoding classification on the user registration information to obtain a user registration information feature vector, and performing time series analysis on the network log information to obtain a network log information feature vector, and merging the transaction data feature vector, the user registration information feature vector and the network log information feature vector as the target feature vector.

[0111] Specifically, regarding the above-mentioned specific implementation process of calculating the push degree of the target feature vector through the processing layer to obtain the service push score, and determining the target push service based on the service push score, this application will provide further details in subsequent specific embodiments, and will not be elaborated on here.

[0112] Through detailed data processing and feature engineering, this application converts transaction data, user registration information and network logs into comprehensive feature vectors through respective optimization methods. This method can greatly improve the accuracy and efficiency of the service recommendation system, enable the recommendation model to more comprehensively understand user behavior and needs, and provide users with highly personalized service recommendations through precise data analysis, thereby enhancing user experience and improving user satisfaction.

[0113] In some optional implementations, the above step of "calculating the push degree of the target feature vector through the processing layer to obtain a service push score, and determining the target push service based on the service push score" includes the following steps:

[0114] Get service options;

[0115] The target feature vector and service options are calculated through collaborative filtering algorithm to obtain the service push score;

[0116] The service push scores are sorted, and the service option with the highest service push score is selected as the target push service.

[0117] In this embodiment, the above-mentioned service options may be user-selectable operation items in the above-mentioned smart ATM, and the above-mentioned user-selectable operation items may include withdrawal services, deposit services, account balance inquiry services, bank statement printing services, transfer services, retirement savings account services, tax optimization suggestions, credit card application services, credit card bill inquiry services, and credit limit management services and other banking services.

[0118] In this embodiment, the collaborative filtering algorithm may be an algorithm for predicting and recommending what the user may be interested in. The collaborative filtering may include user-based collaborative filtering and item-based collaborative filtering. Specifically, the user-based collaborative filtering may be to find other users similar to the target user by calculating the similarity between different users, and then recommend services that these similar users like. The item-based collaborative filtering may be to calculate the similarity between services and recommend similar services to users.

[0119] In a possible embodiment, when the above-mentioned item-based collaborative filtering is used as the above-mentioned collaborative filtering algorithm, the similarity between the above-mentioned service options is calculated by methods such as cosine similarity or Pearson correlation coefficient, and a service push score is generated based on the similarity between the above-mentioned service options, that is, the higher the similarity between the service options, the higher the service push score. For the target user, the above-mentioned smart ATM predicts the current user's preference for unselected services by considering the current target user group's evaluation of the above-mentioned service options and the similarity between the service options, that is, if some services were selected by similar user groups at the same time in the past, then there may be a high similarity between these services. For example, if user A has a high evaluation of service X, and service X is very similar to service Y, then the system may predict that user A will also have a high evaluation of service Y.

[0120] In a possible embodiment, when the user-based collaborative filtering is used as the collaborative filtering algorithm, the user registration information feature vector in the target feature vector is analyzed, and the similarity between the user registration information feature vectors is calculated by methods such as cosine similarity or Pearson correlation coefficient, and similar users of the target user corresponding to the target feature vector are found. The preference coefficient of the target user is predicted based on the evaluation scores of the similar users on the service options, that is, the higher the evaluation score of the service options, the greater the preference coefficient of the target user. The service push score is determined based on the preference coefficient of the target user, that is, the greater the preference coefficient of the target user, the higher the service push score. For example, if the evaluation score of the first service option by a user group similar to the target user is higher than that of other service options, the first service option will be recommended to the target user first.

[0121] In this embodiment, all service options are sorted according to the calculated service push scores, and the service with the highest score is selected to be recommended to the user.

[0122] This application makes intelligent recommendations for service options in smart ATMs by applying user-based collaborative filtering and item-based collaborative filtering algorithms, which can achieve highly personalized and accurate service matching. By analyzing user behaviors and preferences as well as the similarities between services, it provides banking services that best meet user needs, helping banks to better manage and promote their products and provide customized services for specific user groups, thereby improving overall business performance and market competitiveness.

[0123] In some optional implementations, after step S204, the following steps are further included:

[0124] Obtain user feedback data;

[0125] Based on the target push service and user feedback data, the back propagation algorithm is used for fitting to obtain the prediction error;

[0126] The prediction error is compared with a preset threshold. If the prediction error is greater than the preset threshold, the pre-trained service recommendation model is iteratively updated until the prediction error is less than or equal to the preset threshold.

[0127] In this embodiment, the above-mentioned back propagation algorithm, i.e., the error back propagation algorithm (BP algorithm), is a learning algorithm suitable for multi-layer neural networks. It is based on the gradient descent method and is used for error calculation of deep learning networks. The input and output relationship of the BP network is essentially a mapping relationship: the function performed by a BP neural network with n inputs and m outputs is a continuous mapping from n-dimensional Euclidean space to a finite field in m-dimensional Euclidean space, and this mapping is highly nonlinear. The learning process of the BP algorithm consists of a forward propagation process and a backward propagation process. In the forward propagation process, the input information passes through the input layer and the hidden layer, is processed layer by layer and transmitted to the output layer, and is transferred to the backward propagation, and the partial derivatives of the objective function to each neuron weight are obtained layer by layer, forming the gradient of the objective function to the weight vector, which is used as the basis for modifying the weight.

[0128] In a possible embodiment, the above-mentioned smart ATM can collect the user's actual behavior data (i.e., the above-mentioned user feedback data) based on the user's feedback behaviors such as clicks and purchases, and use it as part of the training model. It is fitted through the back propagation algorithm and the prediction error is compared with the preset threshold. If the prediction error is greater than the preset threshold, the pre-trained service recommendation model is iteratively updated until the prediction error is less than or equal to the preset threshold. The model is continuously iteratively trained and updated to improve the accuracy and personalization effect of the recommendation system.

[0129] This application collects user feedback data and adjusts the pre-trained service recommendation model in combination with the back propagation algorithm. The system can continuously adjust and improve its prediction algorithm to reduce prediction errors and improve the accuracy of recommendations. If the prediction error exceeds the preset threshold, the model will be iteratively updated to make the recommendation system more sensitive and adaptable to the user's actual needs and preferences, thereby enhancing user experience and satisfaction, ensuring that the relevance and personalization of service recommendations continue to improve, and making the recommendation service more in line with the user's real-time needs.

[0130] It is understandable that in the specific implementation of the present application, it involves data such as user data to be processed, network log data, target user portraits, and target push services. When the embodiments in the present application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of relevant data need to comply with relevant laws, regulations and standards of relevant countries and regions.

[0131] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned information such as the user data to be processed, network log data, target user portraits, and target push services, the above-mentioned information such as the user data to be processed, network log data, target user portraits, and target push services can also be stored in a node of a blockchain.

[0132] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0133] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0134] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0135] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0136] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0137] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of a service recommendation device, and the device embodiment is similar to Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0138] like Figure 3 As shown, the service recommendation device 300 described in this embodiment includes: an acquisition module 301, a determination module 302, a construction module 303 and a processing module 304. Among them:

[0139] The acquisition module 301 is used to acquire the user data to be processed, wherein the user data to be processed includes facial recognition data and user behavior data;

[0140] A determination module 302, for determining a target user group based on the facial recognition data, and determining network log data of the target user group based on the user behavior data;

[0141] A construction module 303 is used to construct a target user portrait based on the target user group and the network log data;

[0142] The processing module 304 imports the target user portrait into a pre-trained service recommendation model, performs push processing on the target user group through the pre-trained service recommendation model, and obtains a target push service.

[0143] In some optional implementations of this embodiment, the determining module 302 includes:

[0144] A detection submodule, used for performing face detection on the face recognition data through a Haar cascade detector to obtain a target face image;

[0145] A feature extraction submodule, used for extracting features of the target facial image based on the model fitting method to obtain target facial features;

[0146] The comparison submodule is used to obtain a user group corresponding to the target facial features in a preset facial feature library and determine the target user group.

[0147] In some optional implementations of this embodiment, the comparison submodule includes:

[0148] A processing unit, configured to perform principal component analysis on the target facial features to obtain a facial descriptor;

[0149] The comparison unit is used to obtain the user group corresponding to the facial descriptor in a preset facial feature library and determine the target user group.

[0150] In some optional implementations of this embodiment, before the processing module 304, the following is further included:

[0151] A training data set acquisition module is used to acquire a recommended training data set, wherein the recommended training data set is composed of a plurality of user behavior data;

[0152] An encoding module, used to perform feature encoding on the user behavior data and network log data to obtain push data features;

[0153] An estimation module, used to estimate the push data characteristics and obtain model parameters of the service recommendation model;

[0154] A building module, used to build a self-attention network function, and embed the self-attention network function into a hidden layer of a preset general model;

[0155] The import module is used to import the model parameters of the service recommendation model into the general model, and train the general model using the recommended training data set to obtain the pre-trained service recommendation model.

[0156] In some optional implementations of this embodiment, the processing module 304 includes:

[0157] A feature analysis submodule, used to perform feature analysis on the target user portrait through the input layer to obtain a target feature vector;

[0158] A calculation submodule, configured to calculate the push degree of the target feature vector through the processing layer to obtain the service push score, and determine the target push service based on the service push score;

[0159] The output submodule is used to output the target push service through the output layer.

[0160] In some optional implementations of this embodiment, the calculation submodule includes:

[0161] An acquisition unit, used for acquiring service options;

[0162] A calculation unit, configured to calculate the target feature vector and the service options by using a collaborative filtering algorithm to obtain the service push score, wherein one service push score corresponds to one service option;

[0163] The sorting unit is used to sort the service push scores and select the service option with the highest service push score as the target push service.

[0164] In some optional implementations of this embodiment, after the processing module 304, the following is further included:

[0165] A module for obtaining user feedback data is used to obtain user feedback data;

[0166] A fitting module, used to perform fitting based on the target push service and the user feedback data using a back propagation algorithm to obtain a prediction error;

[0167] The updating module is used to compare the prediction error with a preset threshold. If the prediction error is greater than the preset threshold, the pre-trained service recommendation model is iteratively updated until the prediction error is less than or equal to the preset threshold.

[0168] In this embodiment, by obtaining user data to be processed, the target user group is determined based on the facial recognition data, and the network log data of the target user group is determined based on the user behavior data, and a target user portrait is constructed based on the target user group and the network log data. The target user portrait is imported into a pre-trained service recommendation model, and the target user group is pushed through the pre-trained service recommendation model to obtain a target push service. By combining facial recognition and user behavior data, a detailed user portrait is constructed and combined with historical network log data, the needs and preferences of the target user group can be accurately identified and analyzed, the accuracy of the service and user satisfaction are improved, and the user experience is optimized. For user groups who are not very proficient in use, they can more conveniently obtain banking services suitable for them through smart ATMs, thereby improving the efficiency and convenience of intelligent services.

[0169] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0170] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 with components 41-43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable gate arrays (Field-Programmable Gate Array, FPGA), digital processors (Digital Signal Processor, DSP), embedded devices, etc.

[0171] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0172] The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk equipped on the computer device 4, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions of the service recommendation method, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0173] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run computer-readable instructions stored in the memory 41 or process data, such as computer-readable instructions for running the service recommendation method.

[0174] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0175] The present embodiment provides a computer device, which obtains user data to be processed, determines a target user group based on the facial recognition data, and determines the network log data of the target user group based on the user behavior data, constructs a target user portrait based on the target user group and the network log data, imports the target user portrait into a pre-trained service recommendation model, and pushes the target user group through the pre-trained service recommendation model to obtain a target push service. By combining facial recognition and user behavior data, a detailed user portrait is constructed and combined with historical network log data, the needs and preferences of the target user group can be accurately identified and analyzed, the accuracy of the service and user satisfaction are improved, and the user experience is optimized. For user groups who are not very proficient in use, it is more convenient to obtain banking services suitable for them through smart ATMs, thereby improving the efficiency and convenience of intelligent services.

[0176] The present application also provides another implementation, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the service recommendation method as described above.

[0177] The present embodiment provides a computer-readable storage medium, which obtains user data to be processed, determines the target user group based on the facial recognition data, and determines the network log data of the target user group based on the user behavior data, builds a target user portrait based on the target user group and the network log data, imports the target user portrait into a pre-trained service recommendation model, and pushes the target user group through the pre-trained service recommendation model to obtain a target push service. By combining facial recognition and user behavior data, a detailed user portrait is built and combined with historical network log data, the needs and preferences of the target user group can be accurately identified and analyzed, the accuracy of the service and user satisfaction are improved, and the user experience is optimized. For user groups who are not very proficient in use, it is more convenient to obtain banking services suitable for them through smart ATMs, thereby improving the efficiency and convenience of intelligent services.

[0178] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0179] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A service recommendation method, applied to a smart teller machine, characterized in that: The steps include: Acquire user data to be processed, wherein the user data to be processed includes facial recognition data and user behavior data; determining a target user group based on the facial recognition data, and determining web log data of the target user group based on the user behavior data; Building a target user profile based on combining the target user group with the network log data; The target user portrait is imported into a pre-trained service recommendation model, and push processing is performed on the target user group through the pre-trained service recommendation model to obtain a target push service.

2. The service recommendation method according to claim 1, characterized in that: The step of determining the target user group based on the facial recognition data specifically includes: Performing face detection on the face recognition data by using a Haar cascade detector to obtain a target face image; Extracting features of the target facial image based on the model fitting method to obtain target facial features; A user group corresponding to the target facial feature is obtained from a preset facial feature library to determine the target user group.

3. The service recommendation method according to claim 2, characterized in that: The step of obtaining a user group corresponding to the target facial feature in a preset facial feature library and determining the target user group specifically includes: Performing principal component analysis on the target facial features to obtain a facial descriptor; A user group corresponding to the facial descriptor is obtained from a preset facial feature library to determine the target user group.

4. The service recommendation method according to claim 1, characterized in that: Before the step of importing the target user portrait into the pre-trained service recommendation model, and performing push processing on the target user group through the pre-trained service recommendation model to obtain the target push service, the method specifically includes: Acquire a recommended training data set, wherein the recommended training data set is composed of a plurality of the user behavior data; Performing feature encoding on the user behavior data and network log data to obtain push data features; Estimating the push data features to obtain model parameters of a service recommendation model; Building a self-attention network function, and embedding the self-attention network function into a hidden layer of a preset general model; The model parameters of the service recommendation model are imported into the general model, and the general model is trained using the recommended training data set to obtain the pre-trained service recommendation model.

5. The service recommendation method according to claim 1 or 4, characterized in that: The pre-trained service recommendation model includes an input layer, a processing layer and an output layer. The step of performing push processing on the target user group through the pre-trained service recommendation model to obtain the target push service specifically includes: Performing feature analysis on the target user portrait through the input layer to obtain a target feature vector; Calculating the push degree of the target feature vector by the processing layer to obtain the service push score, and determining the target push service based on the service push score; The target push service is output through the output layer.

6. The service recommendation method according to claim 5, characterized in that: The step of calculating the push degree of the target feature vector by the processing layer to obtain the service push score, and determining the target push service based on the service push score specifically includes: Get service options; The target feature vector and the service options are calculated by a collaborative filtering algorithm to obtain the service push score, where one service push score corresponds to one service option; The service push scores are sorted, and the service option with the highest service push score is selected as the target push service.

7. The service recommendation method according to claim 1, characterized in that: After the step of performing push processing on the target user group through the pre-trained service recommendation model to obtain the target push service, the method further includes: Obtain user feedback data; Based on the target push service and the user feedback data, a back propagation algorithm is used to perform fitting to obtain a prediction error; The prediction error is compared with a preset threshold. If the prediction error is greater than the preset threshold, the pre-trained service recommendation model is iteratively updated until the prediction error is less than or equal to the preset threshold.

8. A service recommendation device, applied to a smart teller machine, characterized in that: include: An acquisition module, used to acquire user data to be processed, wherein the user data to be processed includes facial recognition data and user behavior data; a determination module, configured to determine a target user group based on the facial recognition data, and to determine network log data of the target user group based on the user behavior data; A construction module, used to construct a target user portrait based on the target user group and the network log data; The processing module imports the target user portrait into a pre-trained service recommendation model, performs push processing on the target user group through the pre-trained service recommendation model, and obtains a target push service.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores computer-readable instructions, and the processor implements the steps of the service recommendation method according to any one of claims 1 to 7 when executing the computer-readable instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, the steps of the service recommendation method according to any one of claims 1 to 7 are implemented.

Images