User rights matching method and device

Abstract

The present invention can be used in the field of artificial intelligence technology. The present invention provides a user rights matching method and device, including: obtaining user data of a target user and historical redemption rights data of the target user; matching rights for the target user according to the user data, historical redemption rights data, and a user rights matching model pre-established based on a neural network. The present invention establishes a relationship between rights resources, users, and associated users, and provides a rights matching method with more extensiveness and iterative optimization mechanism. Based on knowledge graph technology, the present invention can link users with users, and users with hidden rights and interests, so as to assist in pushing more accurately and make the push results traceable. On the one hand, the rights and interests library resources can be fully utilized so that all kinds of rights and interests resources can be released; more targeted rights and interests benefits can be pushed to users to better meet the real needs of customers, reduce user search time, and improve user experience.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence technology, and specifically relates to a user rights matching method and device. Background Art

[0002] In today's financial field, customer rights have become an important tool for daily marketing and customer acquisition. In order to improve user retention and contribution, traditional banks have cooperated with various merchants to provide customers with a variety of rights services, covering various life scenarios such as travel, medical care, and business. For example, hotel reservation services, car rental services, ticket reservation services, life customization services, etc. The rights and interests that can be enjoyed vary depending on the customer level, among which the rights and interests for high-net-worth customers and private banking customers are the most abundant. After receiving the rights and interests online, customers can directly enjoy the corresponding benefits offline.

[0003] Specifically, banks usually establish a merchant cooperation rights database to centralize the various rights that can be provided and display them uniformly in the channel. Users can complete the redemption after initiating a rights request. The existing rights issuance rules cannot match users and rights according to user characteristics and user levels, and cannot conduct in-depth learning based on the relationship between customers. It is impossible to issue the most suitable rights activities to customers, and the user experience and retention rate are low.

[0004] In the existing technology, the method of matching users with products by using big data, user portraits and historical behavior analysis is increasingly widely used. The existing recommendation system only focuses on the relationship between a single customer and a product in terms of analysis methods, and does not fully extend to the related customers of historical ordering customers, nor does it explore the relationship between related customers and products. In the scenario of strong correlation between bank customers, it cannot fully exert its effect and is not suitable for equity resource matching. Therefore, a method of information matching that can cover more related customers is needed. Summary of the invention

[0005] The present invention belongs to the field of artificial intelligence technology. The user rights matching method and device provided by the present invention establish a relationship among rights resources, users, and related users, and provide a rights matching method with more extensiveness and iterative optimization mechanism. Based on knowledge graph technology, the present invention can connect users with users, and users with hidden rights and interests, so as to assist in pushing more accurately and make the push results traceable. On the one hand, it can make full use of the rights and interests library resources so that all kinds of rights and interests resources can be released; on the other hand, it can push more targeted rights and interests benefits to users, better meet the real needs of customers, reduce user search time, and improve user experience.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a user rights matching method, comprising:

[0008] Obtain user data of a target user and historical redemption rights data of the target user;

[0009] The target user is matched with rights and interests according to the user data, the historical exchange rights and interests data and a user rights and interests matching model pre-established based on a neural network.

[0010] In one embodiment, the method for establishing the user rights matching model includes the following steps:

[0011] Generate training data based on the pre-generated equity resource library and the historical exchange equity data;

[0012] Generate an initial model of the user equity matching model using a convolutional neural network;

[0013] The initial model is trained using the training data to generate the user equity matching model.

[0014] In one embodiment, generating training data according to the pre-generated equity resource library and the historical exchange equity data includes:

[0015] Constructing a rights and interests knowledge graph according to the rights and interests resource library;

[0016] Build the initial customer knowledge graph based on the historical redemption rights data of target customers;

[0017] A customer rights and interests training graph is constructed based on the rights and interests knowledge graph and the initial customer knowledge graph.

[0018] In one embodiment, constructing a rights knowledge graph according to the rights resource library includes:

[0019] Splitting the equity information in the equity resource library into structured data and unstructured data;

[0020] A rights and interests knowledge graph is constructed based on the structured data and unstructured data.

[0021] In one embodiment, the user rights matching method further includes:

[0022] Obtaining the associated customer information of the target customer;

[0023] A target customer knowledge graph is constructed based on the associated customer information.

[0024] In one embodiment, the step of training the initial model using the training data to generate the user equity matching model includes:

[0025] Generate training end conditions based on the target customer knowledge graph;

[0026] The initial model is trained according to the training end condition and the customer equity training map to generate a user equity matching model.

[0027] In one embodiment, the training of the initial model according to the training end condition and the customer equity training graph to generate a user equity matching model includes:

[0028] Mapping entities and relationships in the customer equity training graph to space vectors through different mapping matrices for identification, so as to generate input data for a pre-generated convolutional neural network initial model;

[0029] Inputting the input data into the initial model to train the initial model;

[0030] When the training end condition is reached, the training ends to generate the user equity matching model.

[0031] In one embodiment, the user rights matching method further includes:

[0032] Obtaining usage data and feedback information of the target user on its matching rights and interests;

[0033] The user equity matching model is iteratively optimized according to the usage data and the feedback information.

[0034] In a second aspect, the present invention provides a user rights matching device, comprising:

[0035] A user data acquisition module, used to acquire user data of a target user and historical redemption rights data of the target user;

[0036] The rights and interests matching module is used to match the rights and interests of the target user according to the user data, the historical exchange rights and interests data and a user rights and interests matching model pre-established based on a neural network.

[0037] In one embodiment, the user rights and interests matching device further includes: a matching model building module, which is used to build the user rights and interests matching model, and the matching model building module includes:

[0038] A training data generating unit, configured to generate training data according to a pre-generated equity resource library and the historical exchange equity data;

[0039] An initial model generating unit, used to generate an initial model of the user equity matching model using a convolutional neural network;

[0040] The matching model building unit is used to train the initial model using the training data to generate the user equity matching model.

[0041] In one embodiment, the training data generating unit includes:

[0042] A rights graph construction unit, used to construct a rights knowledge graph according to the rights resource library;

[0043] An initial graph construction unit, used to construct an initial customer knowledge graph based on historical redemption rights data of target customers;

[0044] A training graph construction unit is used to construct a customer equity training graph based on the equity knowledge graph and the initial customer knowledge graph.

[0045] In one embodiment, the equity graph construction unit includes:

[0046] A rights and interests information splitting unit, used for splitting the rights and interests information in the rights and interests resource library into structured data and unstructured data;

[0047] The equity graph construction subunit is used to construct an equity knowledge graph based on the structured data and unstructured data.

[0048] In one embodiment, the user rights matching device further includes:

[0049] A related customer information acquisition module, used to acquire the related customer information of the target customer;

[0050] The target customer graph acquisition module is used to construct a target customer knowledge graph based on the associated customer information.

[0051] In one embodiment, the matching model building unit includes:

[0052] An end condition generating unit, used to generate a training end condition based on the target customer knowledge graph;

[0053] The matching model establishment subunit is used to train the initial model according to the training end condition and the customer equity training map to generate a user equity matching model.

[0054] In one embodiment, the matching model building subunit includes:

[0055] An input data generating unit, used for mapping entities and relationships in the customer equity training graph to space vectors through different mapping matrices for identification, so as to generate input data for a pre-generated convolutional neural network initial model;

[0056] A model training unit, used for inputting the input data into the initial model to train the initial model;

[0057] The target model generating unit is used to terminate the training when the training termination condition is reached to generate the user equity matching model.

[0058] In one embodiment, the user rights matching device further includes:

[0059] A feedback information acquisition module, used to acquire the target user's usage data and feedback information of its matching rights;

[0060] A model optimization module is used to iteratively optimize the user equity matching model according to the usage data and the feedback information.

[0061] In a third aspect, the present invention provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of a user rights matching method when executing the program.

[0062] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of a user rights matching method when executed by a processor.

[0063] From the above description, it can be seen that the user rights matching method and device provided by the embodiment of the present invention first obtain the user data of the target user and the historical exchange rights data of the target user; finally, the target user is matched with rights according to the user data, the historical exchange rights data and the user rights matching model pre-established based on the neural network. The deep learning and resource matching model provided by the present invention establishes a relationship between rights resources, users, and associated users, and provides a rights matching method with more extensiveness and iterative optimization mechanism. On the one hand, the rights library resources can be fully utilized so that all kinds of rights resources are released; on the other hand, more targeted rights and benefits are pushed to users, which better meets the real needs of customers, reduces user search time, and improves user experience. At the same time, based on the knowledge graph technology, the present invention can connect users with users, and users with rights hidden relationships, so as to assist in pushing more accurately and make the push results traceable. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0065] Figure 1 The user rights matching method in the embodiment of the present invention is shown in FIG. Figure 1 ;

[0066] Figure 2 The user rights matching method in the embodiment of the present invention is shown in FIG. Figure 2 ;

[0067] Figure 3 This is a flow chart of step 300 in the user rights matching method in an embodiment of the present invention;

[0068] Figure 4 This is a flow chart of step 301 in the user rights matching method in an embodiment of the present invention;

[0069] Figure 5 This is a flow chart of step 3011 in the user rights and interests matching method in an embodiment of the present invention;

[0070] Figure 6 The user rights matching method in the embodiment of the present invention is shown in FIG. Figure 3 ;

[0071] Figure 7 This is a flow chart of step 303 in the user rights matching method in an embodiment of the present invention;

[0072] Figure 8 This is a flowchart of step 3032 in the user rights matching method in an embodiment of the present invention;

[0073] Fig. 9 The user rights matching method in the embodiment of the present invention is shown in FIG. Figure 4 ;

[0074] Fig.10 It is a flowchart of a method for matching user rights and interests in a specific application example of the present invention;

[0075] Fig.11 The structure of the user rights matching device in the specific application example of the present invention is shown in FIG. Figure 1 ;

[0076] Fig.12 The structure of the user rights matching device in the specific application example of the present invention is shown in FIG. Figure 2 ;

[0077] Fig.13 It is a structural schematic diagram of the matching model building module 30 in a specific application example of the present invention;

[0078] Fig.14 is a structural diagram of a training data generating unit 301 in a specific application example of the present invention;

[0079] Fig.15 It is a structural schematic diagram of the equity graph construction unit 3011 in a specific application example of the present invention;

[0080] Fig.16 The structure of the user rights matching device in the specific application example of the present invention is shown in FIG. Figure 3 ;

[0081] Fig.17 It is a structural schematic diagram of the matching model building unit 303 in a specific application example of the present invention;

[0082] Fig.18 It is a structural schematic diagram of the matching model establishment subunit 3032 in a specific application example of the present invention;

[0083] Fig.19 The structure of the user rights matching device in the specific application example of the present invention is shown in FIG. Figure 4 ;

[0084] Fig. 20 It is a schematic diagram of the structure of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION

[0085] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution 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 described embodiments are 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 creative work are within the scope of protection of the present invention.

[0086] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0087] It should be noted that 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. For example, a process, method, system, product or apparatus comprising 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 apparatuses.

[0088] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0089] The embodiment of the present invention provides a specific implementation method of a user rights matching method, see Figure 1 , the method specifically includes the following contents:

[0090] Step 100: Obtain user data of a target user and historical redemption rights data of the target user.

[0091] The existing benefit recommendation methods only focus on the relationship between a single customer and a product in terms of analytical means, and do not fully extend to the associated customers of historical ordering customers, let alone explore the relationship between associated customers and products. Therefore, it is necessary to establish an association between user data and the user's historical redemption benefit data. User data includes: customer gender, age, occupation, customer level, and historical redemption benefit data includes: types of redeemed benefits, benefit names, etc.

[0092] Step 200: Match the rights and interests of the target user according to the user data, the historical exchange rights and interests data, and a user rights and interests matching model pre-established based on a neural network.

[0093] Specifically, user data and historical redemption rights data are input into the user rights matching model to determine the rights that best suit the target user, so as to increase user stickiness.

[0094] It is understandable that a neural network is a nonlinear, adaptive information processing system composed of a large number of interconnected processing units. It is proposed based on the research results of modern neuroscience, and attempts to process information by simulating the way brain neural networks process and memorize information. Artificial neural networks have four basic characteristics:

[0095] (1) Nonlinearity Nonlinear relationships are a universal characteristic of nature. The intelligence of the brain is a nonlinear phenomenon. Artificial neurons are in two different states: activated or inhibited. This behavior is mathematically expressed as a nonlinear relationship. A network composed of neurons with thresholds has better performance and can improve fault tolerance and storage capacity.

[0096] (2) Non-locality. A neural network is usually composed of multiple neurons that are widely connected. The overall behavior of a system depends not only on the characteristics of a single neuron, but may also be mainly determined by the interactions and connections between units. The non-locality of the brain is simulated through a large number of connections between units. Associative memory is a typical example of non-locality.

[0097] (3) Very qualitative. Artificial neural networks have the ability to adapt, self-organize, and self-learn. Not only can the information processed by the neural network be subject to various changes, but while processing information, the nonlinear dynamic system itself is also constantly changing. Iterative processes are often used to describe the evolution of dynamic systems.

[0098] (4) Non-convexity. The evolutionary direction of a system will depend on a certain state function under certain conditions. For example, the energy function, whose extreme value corresponds to the relatively stable state of the system. Non-convexity means that this function has multiple extreme values, so the system has multiple relatively stable equilibrium states, which will lead to the diversity of system evolution.

[0099] In artificial neural networks, neuron processing units can represent different objects, such as features, letters, concepts, or some meaningful abstract patterns. The types of processing units in the network are divided into three categories: input units, output units, and hidden units. Input units receive signals and data from the outside world; output units realize the output of system processing results; hidden units are units between input and output units and cannot be observed from outside the system. The connection weights between neurons reflect the connection strength between units, and the representation and processing of information are reflected in the connection relationship of network processing units. Artificial neural networks are a non-programmed, adaptive, brain-style information processing. Its essence is to obtain a parallel distributed information processing function through the transformation and dynamic behavior of the network, and imitate the information processing function of the human brain nervous system to varying degrees and levels. It is an interdisciplinary subject involving multiple fields such as neuroscience, thinking science, artificial intelligence, and computer science.

[0100] From the above description, it can be seen that the user rights matching method and device provided by the embodiment of the present invention first obtains the user data of the target user and the historical redemption rights data of the target user; finally, the target user is matched with rights based on the user data, the historical redemption rights data and the user rights matching model pre-established based on the neural network. The present invention proposes a model that constructs association relationships through knowledge graphs and uses convolutional neural networks and attention mechanisms for deep learning. It can recommend suitable rights to more related customers through the model based on the records of the customer's historical redemption rights, and continuously optimize the learning model through a feedback mechanism to improve the rights matching degree and rights redemption effect.

[0101] In one embodiment, see Figure 2 , the user rights matching method also includes:

[0102] Step 300: Establish the user rights matching model. Figure 3 , step 300 further comprises:

[0103] Step 301: Generate training data based on the pre-generated equity resource library and the historical exchange equity data;

[0104] The rights resource library is generally formed by summarizing the rights information of various merchants, and the historical redemption rights data refers to the rights data that the target user has already redeemed.

[0105] Step 302: Generate an initial model of the user equity matching model using a convolutional neural network;

[0106] It is understandable that Convolutional Neural Networks (CNN) is a type of feedforward neural network that includes convolution calculations and has a deep structure, and is one of the representative algorithms of deep learning. Convolutional neural networks have representation learning capabilities and can classify input information in a translation-invariant manner according to their hierarchical structure, so they are also called translation-invariant artificial neural networks.

[0107] Step 303: Use the training data to train the initial model to generate the user equity matching model.

[0108] It is understandable that, when step 303 is implemented, it is also necessary to set certain training end conditions, such as training time, number of training times, etc.

[0109] In one embodiment, see Figure 4 , step 301 comprises:

[0110] Step 3011: constructing a rights and interests knowledge graph according to the rights and interests resource library;

[0111] Specifically, data extraction technologies including entity extraction, relationship extraction, and attribute extraction are used to extract information. And knowledge fusion technology is used to construct a rights and interests graph with rights and interests services as nodes and related functional characteristics and other factors as edges. The graph nodes include (service providers, rights and interests types, service customer levels, etc.)

[0112] Step 3012: construct an initial customer knowledge graph based on the historical redemption rights data of the target customer;

[0113] Specifically, big data is used to filter out customer data that have historically redeemed various rights and interests based on the established rights and interests knowledge graph to build an initial customer knowledge graph. The nodes of the graph include (customer gender, age, occupation, customer level, types of rights and interests redeemed, etc.). In a special case, the target customer is a new customer, that is, there is no historical rights and interests redeemed data for the target customer, then the historical rights and interests redeemed data of the target customer's associated customers are mined to build the initial customer knowledge graph of the target user.

[0114] Step 3013: Construct a customer equity training graph based on the equity knowledge graph and the initial customer knowledge graph.

[0115] Specifically, by building a mapping relationship, the data of the equity knowledge graph and the data of the initial customer knowledge graph are integrated to obtain the customer equity training graph. For example, if a certain equity redeemed by a customer in the initial customer knowledge graph is matched with the equity in the equity knowledge graph, if they point to a common equity entity, the information can be integrated through the knowledge merging technology to form a customer equity training graph.

[0116] In one embodiment, see Figure 5 , step 3011 includes:

[0117] Step 30111: splitting the equity information in the equity resource library into structured data and unstructured data;

[0118] The equity data information is split through the equity database, including structured data such as service providers, service content, service time, service customer levels, and unstructured data such as service web page images and video types.

[0119] Step 30112: Construct a rights and interests knowledge graph based on the structured data and unstructured data.

[0120] In one embodiment, see Figure 6 , the user rights matching method also includes:

[0121] Step 400: Acquire the associated customer information of the target customer;

[0122] Step 500: Construct a target customer knowledge graph based on the associated customer information.

[0123] In step 400 and step 500, first obtain the target customer's associated customers, and build a target customer knowledge graph through customer portrait, historical behavior analysis, and non-financial service order records. The nodes of the graph include (the associated customer's gender, age, occupation, non-financial preferences, etc.)

[0124] In one embodiment, see Figure 7 , step 303 further comprises:

[0125] Step 3031: Generate training end conditions using the target customer knowledge graph;

[0126] Step 3032: Train the initial model according to the training end condition and the customer equity training map to generate a user equity matching model.

[0127] In one embodiment, see Figure 8, step 3032 further includes:

[0128] Step 30321: Map entities and relationships in the customer equity training graph to space vectors through different mapping matrices for identification, so as to generate input data for a pre-generated convolutional neural network initial model;

[0129] Step 30322: inputting the input data into the initial model to train the initial model;

[0130] Step 30323: When the training end condition is reached, the training ends to generate the user equity matching model.

[0131] In step 30321 to step 30323, the entities and relationships in the knowledge graph are mapped to space vectors through different mapping matrices for representation. For example, the triple corresponding to <Customer A, reservation, airport pick-up service> is<h,r,t> , h and t are the head entity vector and tail entity vector, which are obtained through word2vec model training, and r is the corresponding relationship vector. The mapping function M of the head entity vector is calculated by the relationship vector r and the head vector h rh ; The mapping function M of the tail entity vector is calculated using the tail entity vector t and the relationship vector r rt . Through different mapping matrices M rh and M rt After mapping, we get the head entity and tail entity vectors h respectively. ⊥ =M rh h and t ⊥ =M rh t,h ⊥ +r≈t ⊥ The loss function of the model is The corresponding vector value is obtained by minimizing the loss function.

[0132] The customer equity training graph is used as the stock customer equity data, and the entity vectors h1, t1, h2, t2, ... and relationship vectors r1, r2, r3, ... of customer equity are obtained through the vectorized representation model; the target customer knowledge graph formed by the target customer is used as the target prediction data, and the entity vector and relationship vector of the customer are obtained through the vectorized representation model. The vectors of entities and relationships thus obtained can be used as the input of the equity matching model.

[0133] In one embodiment, step 200 further includes:

[0134] Step 201: Determine a plurality of optional user rights according to the user data, the historical exchange rights data and the user rights matching model;

[0135] Step 202: Input the multiple optional user benefits into an attention mechanism pre-established based on the attention mechanism to determine the best user benefit.

[0136] In step 201 and step 202, the feature vector output by the convolutional neural network is used as the input vector of the model. The importance of different rights is measured by the attention mechanism, and different attention weights are assigned. For example, the equity feature vector matrix (s1, s2, s3…) and the customer feature vector m, using e(s i ,m) function calculates the "relevance" between each stock equity s and the current customer m status. The relevance function can be obtained by dot multiplication e(s,m)=s T m. According to the correlation function, the weight of each equity vector is calculated through softmax The matching degree between the existing equity information and the target customer is determined according to the size of the α weight value, and the equity s with the largest weight can be output as the equity that best matches the customer.

[0137] In one embodiment, see Fig. 9 , the user rights matching method also includes:

[0138] Step 600: Obtaining usage data and feedback information of the target user on his matching rights;

[0139] Step 700: Iteratively optimize the user equity matching model according to the usage data and the feedback information.

[0140] In step 600 and step 700, the customer's feedback on the matching rights provided by the system and the customer's browsing times, consultation times, application frequency, and usage frequency are recorded. The push result data can be used as new data for the initial customer knowledge graph to update knowledge and as a sample for model training to iteratively optimize the model.

[0141] To further illustrate the present solution, the present invention also provides a specific application example of the user rights matching method.

[0142] In this specific application example, a user equity matching system is also provided, which includes: a data acquisition module and a matching system model module. Furthermore, the data acquisition module includes: an equity generation unit, an equity data splitting unit, an equity knowledge graph construction unit, an initial data acquisition unit, an associated customer graph construction unit, and a historical data training model unit. The matching system model module includes: a knowledge graph vector representation module, an equity matching module, and a result feedback module, specifically:

[0143] 1. Data acquisition module:

[0144] 1. Equity generation unit: used to aggregate the equity information of each merchant and generate an equity database;

[0145] 2. Equity data splitting unit: splits equity data information through the equity database, including structured data such as service providers, service content, service time, service customer levels, and unstructured data such as service web page images and video types.

[0146] 3. Equity knowledge graph construction unit: Use data extraction technology including entity extraction, relationship extraction, and attribute extraction to extract information. And use knowledge fusion technology to construct an equity graph A with equity services as nodes and related functional characteristics and other factors as edges. The graph nodes include (service providers, equity types, service customer levels, etc.)

[0147] 4. Initial data acquisition unit: Use big data to filter out customer data that has historically redeemed various rights and interests based on the established rights and interests knowledge graph, and construct the initial customer knowledge graph B1. The nodes of the graph include (customer gender, age, occupation, customer level, types of rights and interests that have been redeemed, etc.)

[0148] 5. Associated customer graph construction unit: Obtain associated customers of the initial data, and construct the target customer knowledge graph B2 through customer portraits, historical behavior analysis, and non-financial service order records. The nodes of the graph include (the gender, age, occupation, non-financial preferences, etc. of the associated customers)

[0149] 6. Historical data training model unit: By building a mapping relationship, the data of the equity map A and the data of the initial customer map B1 are integrated to obtain the customer equity training map C. For example: if a certain equity redeemed by a customer in the map B1 is matched with the equity in the map A, if they point to a common equity entity, the information can be integrated through the knowledge merging technology to form the customer equity training map C.

[0150] 2. Matching system model module:

[0151] 1. Knowledge graph vector representation module:

[0152] 1) Vector representation model: The entities and relationships in the knowledge graph are mapped to space vectors through different mapping matrices for representation. For example, the triple corresponding to <Customer A, reservation, airport pick-up service> is<h,r,t> , h and t are the head entity vector and tail entity vector, which are obtained through word2vec model training, and r is the corresponding relationship vector. The mapping function M of the head entity vector is calculated by the relationship vector r and the head vector h rh ; The mapping function M of the tail entity vector is calculated using the tail entity vector t and the relationship vector r rt . Through different mapping matrices M rh and Mrt After mapping, we get the head entity and tail entity vectors h respectively. ⊥ =M rh h and t ⊥ =M rh t,h ⊥ +r≈t ⊥ The loss function of the model is The corresponding vector value is obtained by minimizing the loss function.

[0153] 2) Knowledge graph entity and relationship vectorization: The customer equity training graph C is used as the existing customer equity data, and the entity vectors h1, t1, h2, t2, ... and relationship vectors r1, r2, r3, ... of customer equity are obtained through the vectorization representation model; the customer knowledge graph B2 formed by the target customer is used as the target prediction data, and the entity vector and relationship vector of the customer are obtained through the vectorization representation model. The entity and relationship vectors obtained by this module can be used as the input of the equity matching model.

[0154] 2. Equity matching module:

[0155] 1) Convolutional neural network model: In order to make the convolutional neural network more robust, this model uses a multi-channel convolution layer with a depth of 3. The initial entity and relationship vectors obtained by word2vec model training are used as the first channel input of the convolution layer; the output entity and relationship vectors of the knowledge graph vector representation module are used as the second channel input of the convolution layer; the entity vector obtained by taking the mean of the tail entity vectors of the entities in the knowledge graph is used as the third channel input of the convolution layer, for example and<h1,r2,t2> The mean of vectors t1 and t2 can be taken as the tail entity vector of h1. The target customer vector data and the existing customer equity vector data are respectively input into the convolutional neural network model, and features are extracted through different convolution channels. The pooling layer uses the maximum pooling method to reduce the dimension of the matrix and finally obtain the output vector of the pooling layer. The output includes the target customer feature vector m and the existing customer equity feature vectors s1, s2, s3... of the training sample, which are used as the input of the attention mechanism model.<h1>

[0156] 2) Attention mechanism model: The feature vector output by the convolutional neural network is used as the input vector of the model. The importance of different rights is measured through the attention mechanism, and different attention weights are assigned. For example, the equity feature vector matrix (s1, s2, s3…) and the customer feature vector m, using e(s i ,m) function calculates the "relevance" between each stock equity s and the current customer m status. The relevance function can be obtained by dot multiplication e(s,m)=s T m. According to the correlation function, the weight of each equity vector is calculated through softmax The matching degree between the existing equity information and the target customer is determined according to the size of the α weight value, and the equity s with the largest weight can be output as the equity that best matches the customer.

[0157] 3. Result feedback module:

[0158] Record the customer's feedback on the matching rights provided by the system, as well as the customer's browsing times, consultation times, application frequency, and usage frequency. The push result data can be used as new data for knowledge graph B1 to update knowledge, and as a sample for model training to iteratively optimize the model.

[0159] See also Fig.10 Based on the above user rights matching system, the user rights matching method provided by the specific application example of the present invention includes:

[0160] S1: Obtain merchant rights and interests information and generate a rights and interests resource library.

[0161] S2: Split the equity information, which includes structured data and unstructured data, and construct the equity knowledge graph A based on it.

[0162] S3: Obtain customer data that has historically redeemed benefits to construct the initial customer knowledge graph B1.

[0163] S4: Obtain the target customer’s related customers and build the target customer knowledge graph B2.

[0164] S5: Merge the equity knowledge graph A and the initial customer knowledge graph B1 to form the customer equity training graph C.

[0165] S6: Map entities and relationships in the knowledge graph into space vectors for identification through different mapping matrices.

[0166] S7: The customer’s entity vector and relationship vector are obtained through the vectorized representation model as the input of the equity matching model.

[0167] S8: A multi-channel convolutional layer with a depth of 3 is used to input the target customer vector data and the existing customer equity vector data into the convolutional neural network model respectively, and feature extraction is performed through different convolution channels as the input of the attention mechanism model.

[0168] S9: The importance of different rights is measured through the attention mechanism, and different attention weight scores α are assigned. The right with the largest weight is output as s as the right that best matches the customer.

[0169] S10: Record the customer's feedback on the matching rights provided by the system. The result data can be used as new energy data of the knowledge graph for model iterative optimization.

[0170] From the above description, it can be seen that the user rights matching method provided by the embodiment of the present invention first obtains the user data of the target user and the historical redemption rights data of the target user; finally, the target user is matched with rights based on the user data, the historical redemption rights data, and the user rights matching model pre-established based on the neural network. The deep learning and resource matching model provided by the present invention establishes a relationship between rights resources, users, and associated users, and provides a rights matching method with more extensiveness and iterative optimization mechanism. On the one hand, the rights library resources can be fully utilized so that all kinds of rights resources are released; on the other hand, more targeted rights and benefits are pushed to users, which better meets the real needs of customers, reduces user search time, and improves user experience. At the same time, based on the knowledge graph technology, the present invention can connect users with users, and users with rights hidden relationships, so as to assist in pushing more accurately and make the push results traceable.

[0171] Based on the same inventive concept, the embodiments of the present application also provide a user rights matching device, which can be used to implement the method described in the above embodiments, such as the following embodiments. Since the principle of solving the problem by the user rights matching device is similar to that of the user rights matching method, the implementation of the user rights matching device can refer to the implementation of the user rights matching method, and the repeated parts will not be repeated. As used below, the term "unit" or "module" can be a combination of software and / or hardware that implements predetermined functions. Although the system described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived.

[0172] The embodiment of the present invention provides a specific implementation of a user rights matching device capable of implementing a user rights matching method, see Fig.11 , the user rights matching device specifically includes the following contents:

[0173] A user data acquisition module 10, used to acquire user data of a target user and historical redemption rights data of the target user;

[0174] The rights matching module 20 is used to match the rights of the target user according to the user data, the historical exchange rights data and the user rights matching model pre-established based on the neural network.

[0175] In one embodiment, see Fig.12 The user rights matching device also includes: a matching model building module 30, which is used to build the user rights matching model, see Fig.13 , the matching model building module 30 includes:

[0176] A training data generating unit 301 is used to generate training data according to the pre-generated equity resource library and the historical exchange equity data;

[0177] An initial model generating unit 302, configured to generate an initial model of the user equity matching model using a convolutional neural network;

[0178] The matching model building unit 303 is used to train the initial model using the training data to generate the user equity matching model.

[0179] In one embodiment, see Fig.14 , the training data generating unit 301 comprises:

[0180] The equity graph construction unit 3011 is used to construct an equity knowledge graph according to the equity resource library;

[0181] An initial graph construction unit 3012 is used to construct an initial customer knowledge graph based on historical redemption rights data of target customers;

[0182] The training graph construction unit 3013 is used to construct a customer equity training graph based on the equity knowledge graph and the initial customer knowledge graph.

[0183] In one embodiment, see Fig.15 , the equity graph construction unit 3011 includes:

[0184] The rights information splitting unit 30111 is used to split the rights information in the rights resource library into structured data and unstructured data;

[0185] The equity graph construction subunit 30112 is used to construct an equity knowledge graph based on the structured data and the unstructured data.

[0186] In one embodiment, see Fig.16 , the user rights matching device also includes:

[0187] The associated customer information acquisition module 40 is used to acquire the associated customer information of the target customer;

[0188] The target customer graph acquisition module 50 is used to construct a target customer knowledge graph based on the associated customer information.

[0189] In one embodiment, see Fig.17 , the matching model building unit 303 includes:

[0190] An end condition generating unit 3031 is used to generate a training end condition using the target customer knowledge graph;

[0191] The matching model establishing subunit 3032 is used to train the initial model according to the training end condition and the customer equity training map to generate a user equity matching model.

[0192] In one embodiment, see Fig.18 , the matching model building subunit 3032 includes:

[0193] An input data generating unit 30321 is used to map entities and relationships in the customer equity training graph to space vectors through different mapping matrices for identification, so as to generate input data for a pre-generated convolutional neural network initial model;

[0194] A model training unit 30322, used for inputting the input data into the initial model to train the initial model;

[0195] The target model generating unit 30323 is used to terminate the training when the training termination condition is met to generate the user equity matching model.

[0196] In one embodiment, see Fig.19 , the user rights matching device also includes:

[0197] A feedback information acquisition module 60, used to acquire the target user's usage data and feedback information on his matching rights;

[0198] The model optimization module 70 is used to iteratively optimize the user equity matching model according to the usage data and the feedback information.

[0199] From the above description, it can be seen that the user rights matching device provided by the embodiment of the present invention first obtains the user data of the target user and the historical redemption rights data of the target user; finally, the target user is matched with rights based on the user data, the historical redemption rights data and the user rights matching model pre-established based on the neural network. The deep learning and resource matching model provided by the present invention establishes a relationship between rights resources, users, and associated users, and provides a rights matching method with more extensiveness and iterative optimization mechanism. On the one hand, the rights library resources can be fully utilized so that all kinds of rights resources are released; on the other hand, more targeted rights and benefits are pushed to users, which better meets the real needs of customers, reduces user search time, and improves user experience. At the same time, based on the knowledge graph technology, the present invention can connect users with users, and users with rights hidden relationships, so as to assist in pushing more accurately and make the push results traceable.

[0200] Reference below Fig. 20 , which shows a schematic structural diagram of an electronic device 600 suitable for implementing an embodiment of the present application.

[0201] like Fig. 20As shown, electronic device 600 includes a central processing unit (CPU) 601, which can perform various appropriate operations and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage portion 608 into a random access memory (RAM) 603. In RAM 603, various programs and data required for the operation of system 600 are also stored. CPU 601, ROM 602, and RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to bus 604.

[0202] The following components are connected to the I / O interface 605: an input section 606 including a keyboard, a mouse, etc.; an output section 607 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, a modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the I / O interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed, so that a computer program read therefrom is installed as needed as the storage section 608.

[0203] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the above-mentioned method for determining the distance between people in a data center scenario are implemented, and the steps include:

[0204] Step 100: Receive the native load balancing model and target image version of the application to be upgraded;

[0205] Step 200: Determine a pod list corresponding to the native load balancing model;

[0206] Step 300: modify the image file in the pod file to the target image version according to the pod list.

[0207] In such an embodiment, the computer program may be downloaded and installed from a network via the communication section 609 , and / or installed from the removable medium 611 .

[0208] For the convenience of description, the above device is described in various units according to their functions. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0209] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0210] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0211] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, commodity or device including the element.

[0212] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0213] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A user rights matching method, characterized in that: include: Obtain user data of the target user and historical redemption rights and interests data of the target user; wherein the user data includes: user gender, age, occupation, user level; the historical redemption rights and interests data includes: types of rights and interests that have been redeemed and rights and interests names; Matching the rights and interests of the target user according to the user data, the historical exchange rights and interests data, and a user rights and interests matching model pre-established based on a neural network; The method for establishing the user rights matching model comprises the following steps: Generate training data based on the pre-generated equity resource library and the historical exchange equity data; wherein the equity resource library is formed by summarizing the equity information of each merchant; and generate an initial model of the user equity matching model using a convolutional neural network; Using the training data to train the initial model to generate the user equity matching model; The generating of training data according to the pre-generated equity resource library and the historical exchange equity data includes: Constructing a rights and interests knowledge graph according to the rights and interests resource library; Build the initial customer knowledge graph based on the historical redemption rights data of target customers; Constructing a customer equity training graph based on the equity knowledge graph and the initial customer knowledge graph; Constructing a rights and interests knowledge graph based on the rights and interests resource library includes: utilizing data extraction technologies including entity extraction, relationship extraction, and attribute extraction to extract information, and utilizing knowledge fusion technology to construct a rights and interests graph with rights and interests services as nodes and related functional factors as edges; the nodes of the rights and interests knowledge graph include: service providers, rights and interests types, and service customer levels.

2. The user rights matching method according to claim 1, characterized in that: The constructing of a rights and interests knowledge graph according to the rights and interests resource library comprises: Splitting the equity information in the equity resource library into structured data and unstructured data; A rights and interests knowledge graph is constructed based on the structured data and unstructured data.

3. The user rights matching method according to claim 2, characterized in that: Also includes: Obtaining the associated customer information of the target customer; A target customer knowledge graph is constructed based on the associated customer information.

4. The user rights matching method according to claim 3, characterized in that: The using the training data to train the initial model to generate the user equity matching model includes: Generate training end conditions based on the target customer knowledge graph; The initial model is trained according to the training end condition and the customer equity training map to generate a user equity matching model.

5. The user rights matching method according to claim 4, characterized in that: The training of the initial model according to the training end condition and the customer equity training graph to generate a user equity matching model includes: Mapping entities and relationships in the customer equity training graph to space vectors through different mapping matrices for identification, so as to generate input data for a pre-generated convolutional neural network initial model; Inputting the input data into the initial model to train the initial model; When the training end condition is reached, the training ends to generate the user equity matching model.

6. The user rights matching method according to claim 1, characterized in that: Matching the rights and interests of the target user according to the user data, the historical exchange rights and interests data, and a user rights and interests matching model pre-established based on a neural network, including: Determine a plurality of optional user rights according to the user data, the historical exchange rights data and the user rights matching model; The multiple optional user benefits are input into an attention mechanism pre-established based on the attention mechanism to determine the best user benefit.

7. The user rights matching method according to claim 1, characterized in that: Also includes: Obtaining usage data and feedback information of the target user on its matching rights and interests; The user equity matching model is iteratively optimized according to the usage data and the feedback information.

8. A user rights matching device, characterized in that: include: A user data acquisition module is used to acquire the user data of the target user and the historical exchange rights and interests data of the target user; wherein the user data includes: user gender, age, occupation, user level; the historical exchange rights and interests data includes: types of exchanged rights and interests names; A rights matching module, used for matching rights of the target user according to the user data, the historical exchange rights data and a user rights matching model pre-established based on a neural network; A matching model building module is used to build the user equity matching model. The matching model building module includes: A training data generating unit, used to generate training data according to the pre-generated equity resource library and the historical exchange equity data; wherein the equity resource library is formed by aggregating the equity information of each merchant; An initial model generating unit, used to generate an initial model of the user equity matching model using a convolutional neural network; A matching model building unit, used to train the initial model using the training data to generate the user equity matching model; The generating of training data according to the pre-generated equity resource library and the historical exchange equity data includes: Constructing a rights and interests knowledge graph according to the rights and interests resource library; Build the initial customer knowledge graph based on the historical redemption rights data of target customers; Constructing a customer equity training graph based on the equity knowledge graph and the initial customer knowledge graph; Constructing a rights and interests knowledge graph based on the rights and interests resource library includes: utilizing data extraction technologies including entity extraction, relationship extraction, and attribute extraction to extract information, and utilizing knowledge fusion technology to construct a rights and interests graph with rights and interests services as nodes and related functional factors as edges; the nodes of the rights and interests knowledge graph include: service providers, rights and interests types, and service customer levels.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the user rights matching method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the user rights matching method described in any one of claims 1 to 7 are implemented.

Images