Incentive scheme determination method and device based on causal inference, equipment and medium

Abstract

The embodiment of the invention provides an incentive scheme determination method and device based on causal inference, equipment and a medium, and belongs to the technical field of artificial intelligence. The method comprises the steps of obtaining customer basic information, historical incentive data and corresponding historical interactive behavior data of a target customer, and extracting multi-dimensional feature information of the target customer; performing deep clustering on the target customers according to the multi-dimensional feature information to obtain a plurality of target customer groups, and determining customer group feature vectors of the target customer groups; through a trained causal inference model, based on the customer group feature vector, determining an interaction improvement gain of each target customer group under each incentive strategy; and performing multi-target optimization according to the excitation cost of each excitation strategy and the corresponding interaction improvement gain to obtain an optimal excitation strategy of each target customer group, and determining an excitation scheme of each target customer group according to the optimal excitation strategy. According to the embodiment of the invention, the accuracy of issuing incentives to customers and the drainage efficiency of the interactive platform can be improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method and apparatus, device and medium for determining an incentive scheme based on causal inference. Background Art

[0002] Insurance companies usually develop some interactive platforms for customers, and use incentives from headquarters, agencies, and agents themselves to attract customers from the platform. Generally, under a given budget, different forms of incentives are issued to agents' customers, hoping to maximize the interaction between customers and agents through incentives, so as to promote more efficient business development by subsequent agents. The form of incentives is usually determined by factors such as time, quantity, and content.

[0003] In the past, incentives were basically determined manually by the headquarters’ operations staff, who lacked the understanding of the classification of different customers and group management. The effect of incentives was also reviewed based on the final total amount, which affected the accuracy of incentives issued to customers. At the same time, since the customer situation of each agent is different, the lack of refined incentive activities is not conducive to maximizing benefits and also affects the efficiency of attracting traffic to the interactive platform. Summary of the invention

[0004] The main purpose of the embodiments of the present application is to propose a method and apparatus, device and medium for determining an incentive scheme based on causal inference, aiming to improve the accuracy of issuing incentives to customers and the efficiency of attracting traffic to the interactive platform.

[0005] To achieve the above-mentioned purpose, a first aspect of an embodiment of the present application proposes a method for determining an incentive scheme based on causal inference, the method comprising:

[0006] Obtaining basic customer information, historical incentive data, and corresponding historical interactive behavior data of the target customer, and extracting multi-dimensional feature information of the target customer based on the basic customer information, the historical incentive data, and the historical interactive behavior data;

[0007] Performing deep clustering on the target customers according to the multi-dimensional feature information to obtain multiple target customer groups, and determining a customer group feature vector for each target customer group;

[0008] Determine the interaction enhancement gain of each target customer group under each incentive strategy through a pre-trained causal inference model based on the customer group feature vector and a plurality of preset incentive strategies;

[0009] Multi-objective optimization is performed based on the incentive cost of each incentive strategy and the corresponding interaction improvement gain to obtain the optimal incentive strategy for each target customer group, and then the incentive plan for each target customer group is determined based on the optimal incentive strategy.

[0010] In some embodiments, the step of obtaining basic customer information, historical incentive data, and corresponding historical interactive behavior data of the target customer, and extracting multi-dimensional feature information of the target customer based on the basic customer information, the historical incentive data, and the historical interactive behavior data includes:

[0011] Acquire the historical incentive data of the target customer in a preset first historical period and the historical interactive behavior data in a preset second historical period through the interactive platform;

[0012] Determine the number of historical incentives and the content of historical incentives based on the historical incentive data, and determine the historical interactive behavior indicators based on the historical interactive behavior data;

[0013] Acquire basic customer information of the target customer, and determine multiple basic customer attributes according to the basic customer information;

[0014] Determine the multi-dimensional feature information according to the basic attributes of the customer, the number of historical incentives, the historical incentive content, and the historical interactive behavior indicators;

[0015] The second historical period is the next period of the first historical period.

[0016] In some embodiments, the deep clustering of the target customers according to the multi-dimensional feature information to obtain multiple target customer groups, and determining the customer group feature vector of each target customer group includes:

[0017] Inputting the multi-dimensional feature information into a pre-trained SCCL clustering model to obtain a customer category corresponding to each target customer;

[0018] Dividing the target customers into a plurality of target customer groups according to the customer categories;

[0019] The customer group feature vector is generated according to the basic customer information of each target customer in the target customer group.

[0020] In some embodiments, the SCCL clustering model is trained by the following steps:

[0021] Acquire multiple sample customer feature data, wherein the sample customer feature data includes sample customer basic attributes, sample incentive times, sample incentive content, and sample interactive behavior indicators;

[0022] Performing data enhancement on each of the sample customer feature data to obtain corresponding enhanced customer feature data, and then constructing an enhanced data set based on the sample customer feature data and the enhanced customer feature data;

[0023] Construct the SCCL clustering model and determine the contrast loss function and clustering loss function;

[0024] Taking the sample customer feature data and the corresponding enhanced customer feature data as a positive sample pair, taking other data samples in the enhanced data set except the positive sample pair as negative samples, performing contrast learning and cluster learning on the positive sample pair and the negative samples, and determining a contrast loss value through the contrast loss function, and determining a cluster loss value through the cluster loss function;

[0025] The parameters of the SCCL clustering model are updated according to the contrast loss value and the clustering loss value to obtain the trained SCCL clustering model.

[0026] In some embodiments, the causal inference model is trained by the following steps:

[0027] Acquire multiple sample customer incentive data corresponding to each of the incentive strategies, wherein the sample customer incentive data includes a sample customer feature vector, incentive strategy encoding data, and a corresponding sample interactive behavior indicator;

[0028] Determine the incentive cost of each of the incentive strategies, construct a control group according to the incentive data of the sample customers corresponding to the incentive strategy with the smallest incentive cost, and construct corresponding multiple intervention groups according to the incentive data of the sample customers corresponding to the other incentive strategies;

[0029] According to the control group and the intervention group, reinforcement learning and causal inference are performed based on the X-Learner model to obtain the causal inference model.

[0030] In some embodiments, the determining of the interaction enhancement gain of each target customer group under each incentive strategy based on the customer group feature vector and a plurality of preset incentive strategies by using a pre-trained causal inference model includes:

[0031] The characteristic vectors of each customer group and the incentive strategy codes corresponding to each incentive strategy are input into the causal inference model in pairs to obtain the predicted values ​​of the interactive behavior indicators of each target customer group under each incentive strategy;

[0032] Determine the predicted value of the interactive behavior index of each target customer group under the incentive strategy with the minimum incentive cost as the interactive behavior index benchmark value of each target customer group;

[0033] The interaction enhancement gain of each target customer group under each incentive strategy is determined according to the difference between the predicted value of the interaction behavior indicator and the baseline value of the interaction behavior indicator.

[0034] In some embodiments, performing multi-objective optimization according to the incentive cost of each incentive strategy and the corresponding interaction enhancement gain to obtain the optimal incentive strategy for each target customer group includes:

[0035] Construct an objective function about incentive cost and interaction improvement gain, and determine the incentive cost constraint and positive gain constraint;

[0036] Constructing an initial population corresponding to each target customer group according to the incentive strategy and the corresponding interaction improvement gain, and determining a fitness function according to the objective function, the incentive cost constraint and the positive gain constraint;

[0037] Utilizing a genetic algorithm to optimize each of the initial populations according to the fitness function to obtain an optimal solution corresponding to each of the target customer groups;

[0038] The optimal incentive strategy for each target customer group is determined according to the optimal solution.

[0039] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application proposes a device for determining an incentive scheme based on causal inference, the device comprising:

[0040] A feature extraction module is used to obtain basic customer information, historical incentive data and corresponding historical interactive behavior data of a target customer, and extract multi-dimensional feature information of the target customer based on the basic customer information, the historical incentive data and the historical interactive behavior data;

[0041] A customer clustering module, used to perform deep clustering on the target customers according to the multi-dimensional feature information, obtain multiple target customer groups, and determine a customer group feature vector for each target customer group;

[0042] An interaction enhancement gain determination module, used to determine the interaction enhancement gain of each target customer group under each incentive strategy based on the customer group feature vector and a plurality of preset incentive strategies through a pre-trained causal inference model;

[0043] The optimization module is used to perform multi-objective optimization according to the incentive cost of each incentive strategy and the corresponding interaction improvement gain, obtain the optimal incentive strategy for each target customer group, and then determine the incentive plan for each target customer group according to the optimal incentive strategy.

[0044] To achieve the above-mentioned purpose, the third aspect of an embodiment of the present application proposes an electronic device, which includes a memory, a processor, a program stored on the memory and executable on the processor, and a data bus for realizing connection and communication between the processor and the memory. When the program is executed by the processor, the method for determining an incentive scheme based on causal inference as described in the first aspect above is implemented.

[0045] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes a storage medium, which is a computer-readable storage medium used for computer-readable storage, and the storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the incentive scheme determination method based on causal inference as described in the first aspect above.

[0046] The present application proposes a method and apparatus, device and medium for determining an incentive scheme based on causal inference, which obtains basic customer information, historical incentive data and corresponding historical interactive behavior data of target customers, and extracts multi-dimensional feature information of target customers based on the basic customer information, historical incentive data and historical interactive behavior data, deeply clusters the target customers based on the multi-dimensional feature information to obtain multiple target customer groups, and determines the customer group feature vector of each target customer group, determines the interaction enhancement gain of each target customer group under each incentive strategy through a pre-trained causal inference model based on the customer group feature vector and multiple preset incentive strategies, performs multi-objective optimization based on the incentive cost of each incentive strategy and the corresponding interaction enhancement gain, obtains the optimal incentive strategy for each target customer group, and then determines the incentive scheme for each target customer group based on the optimal incentive strategy. The embodiment of the present application extracts multi-dimensional feature information of customers based on their basic information, historical incentive data, and historical interactive behavior data, performs deep clustering of customers based on the multi-dimensional feature information to obtain multiple customer groups, determines the interaction enhancement gain of each customer group under different incentive strategies through a causal inference model, and obtains the optimal incentive strategy for each customer group through multi-objective optimization, so that the incentive plan for each customer group can be determined based on the optimal incentive strategy, thereby improving the accuracy of incentives issued to customers and the efficiency of attracting traffic to the interactive platform. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a flow chart of a method for determining an incentive scheme based on causal inference provided in an embodiment of the present application;

[0048] Figure 2 yes Figure 1 A flow chart of step S101 in FIG.

[0049] Figure 3 yes Figure 1A flow chart of step S102 in FIG.

[0050] Figure 4 is a flow chart of training the SCCL clustering model provided in an embodiment of the present application;

[0051] Figure 5 is a flow chart of a training causal inference model provided in an embodiment of the present application;

[0052] Figure 6 yes Figure 1 A flow chart of step S103 in FIG.

[0053] Figure 7 yes Figure 1 A flow chart of step S104 in FIG.

[0054] Figure 8 It is a structural schematic diagram of an incentive scheme determination device based on causal inference provided in an embodiment of the present application;

[0055] Fig. 9 is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of the present application;

[0056] Fig.10 It is a schematic diagram of the structure of the storage medium provided in an embodiment of the present application. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0058] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0059] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0060] First, some nouns involved in this application are analyzed:

[0061] Artificial intelligence (AI) is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. AI is a branch of computer science that attempts to understand the essence of intelligence and produce a new type of intelligent machine that can respond in a similar way to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing and expert systems. AI can simulate the information process of human consciousness and thinking. AI is also a 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.

[0062] Customer incentives: Improving certain operational / product data indicators through material rewards can help insurance agents improve their performance and achieve KPIs. In addition, since customers receive certain incentives from the product, their sunk costs are increased, which also increases customer stickiness.

[0063] Deep clustering: Cluster analysis is a key technology for mining the internal structure of data. In the era of big data, the data people face are usually large in scale, high in dimension, and complex in structure. Direct application of traditional clustering algorithms often fails. Deep learning makes large-scale deep feature extraction possible with its hierarchical nonlinear mapping capabilities. Therefore, clustering algorithms based on deep learning (deep clustering) have quickly become a research hotspot in the field of unsupervised learning.

[0064] Causal inference: Causal inference is one of the core issues in statistics and data science. When a phenomenon has already occurred, the process of drawing a causal conclusion is causal inference. It has been widely used in biomedicine, economic management, and social sciences. It can reveal the causal relationship between variables and discover the deep reasons behind the phenomenon. Causal inference is also considered a paradigm revolution in the field of artificial intelligence and is one of the research hotspots in this field in recent years.

[0065] Genetic Algorithm (GA): It was first proposed by John Holland of the United States in the 1970s. The algorithm was designed and proposed based on the laws of evolution of organisms in nature. It is a computational model of biological evolution that simulates the natural selection and genetic mechanism of Darwin's theory of biological evolution. It is a method of searching for the optimal solution by simulating the natural evolution process. The algorithm uses mathematical methods and computer simulation operations to convert the problem-solving process into processes similar to the crossover and mutation of chromosome genes in biological evolution. When solving more complex combinatorial optimization problems, it can usually obtain better optimization results faster than some conventional optimization algorithms. Genetic algorithms have been widely used in combinatorial optimization, machine learning, signal processing, adaptive control, artificial life and other fields.

[0066] Insurance companies usually develop some interactive platforms for customers, and use incentives from headquarters, agencies, and agents themselves to attract customers from the platform. Generally, under a given budget, different forms of incentives are issued to agents' customers, hoping to maximize the interaction between customers and agents through incentives, so as to promote more efficient business development by subsequent agents. The form of incentives is usually determined by factors such as time, quantity, and content.

[0067] In the past, incentives were basically determined manually by the headquarters’ operations staff, who lacked the understanding of the classification of different customers and group management. The effect of incentives was also reviewed based on the final total amount, which affected the accuracy of incentives issued to customers. At the same time, since the customer situation of each agent is different, the lack of refined incentive activities is not conducive to maximizing benefits and also affects the efficiency of attracting traffic to the interactive platform.

[0068] Based on this, the embodiments of the present application provide a method and apparatus, device and medium for determining an incentive scheme based on causal inference, aiming to improve the accuracy of incentives issued to customers and the efficiency of attracting traffic to the interactive platform. The embodiments of the present application introduce a causal inference model to optimize the incentive scheme of the interactive platform, which can improve the interests of the insurance company, the agent and the customer from a practical level.

[0069] The method, device, equipment and medium for determining an incentive scheme based on causal inference provided in the embodiments of the present application are specifically illustrated through the following embodiments. First, the method for determining an incentive scheme based on causal inference in the embodiments of the present application is described.

[0070] 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.

[0071] 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.

[0072] The method for determining an incentive scheme based on causal inference provided in the embodiment of the present application relates to the field of artificial intelligence technology. The method for determining an incentive scheme based on causal inference provided in the embodiment of the present application can be applied to a terminal, can be applied to a server side, or can be software running in a terminal or a server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements a method for determining an incentive scheme based on causal inference, etc., but is not limited to the above forms.

[0073] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0074] It should be noted that in each specific implementation of the present application, when it comes to the need to perform relevant processing based on data related to user identity or characteristics such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first, and the collection, use, and processing of these data will comply with relevant laws, regulations, and standards. In addition, when the embodiment of the present application needs to obtain the user's sensitive personal information, the user's separate permission or consent will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the user's separate permission or consent, the necessary user-related data for the normal operation of the embodiment of the present application will be obtained.

[0075] Figure 1 is an optional flow chart of a method for determining an incentive scheme based on causal inference provided in an embodiment of the present application. Figure 1 The method may include but is not limited to steps S101 to S104.

[0076] Step S101, obtaining basic customer information, historical incentive data, and corresponding historical interactive behavior data of a target customer, and extracting multi-dimensional feature information of the target customer based on the basic customer information, historical incentive data, and historical interactive behavior data;

[0077] Step S102, performing deep clustering on the target customers according to the multi-dimensional feature information to obtain multiple target customer groups, and determining the customer group feature vector of each target customer group;

[0078] Step S103, determining the interaction enhancement gain of each target customer group under each incentive strategy through a pre-trained causal inference model based on the customer group feature vector and a plurality of preset incentive strategies;

[0079] Step S104, performing multi-objective optimization according to the incentive cost of each incentive strategy and the corresponding interaction improvement gain, obtaining the optimal incentive strategy for each target customer group, and then determining the incentive plan for each target customer group according to the optimal incentive strategy.

[0080] In steps S101 to S104 shown in the embodiment of the present application, multi-dimensional feature information of the customer is extracted based on the customer's basic information, historical incentive data, and historical interactive behavior data, and the customer is deeply clustered based on the multi-dimensional feature information to obtain multiple customer groups. The interaction enhancement gain of each customer group under different incentive strategies is determined through a causal inference model, and the optimal incentive strategy for each customer group is obtained through multi-objective optimization. Therefore, the incentive plan for each customer group can be determined based on the optimal incentive strategy, thereby improving the accuracy of incentives issued to customers and the efficiency of attracting traffic to the interactive platform.

[0081] See also Figure 2 In some embodiments, step S101 may include but is not limited to steps S1011 to S1014:

[0082] Step S1011, obtaining the historical incentive data of the target customer in a preset first historical period and the historical interactive behavior data in a preset second historical period through the interactive platform;

[0083] Step S1012, determining the number of historical incentives and the content of historical incentives according to the historical incentive data, and determining the historical interactive behavior index according to the historical interactive behavior data;

[0084] Step S1013, obtaining basic customer information of the target customer, and determining a plurality of basic customer attributes according to the basic customer information;

[0085] Step S1014, determining multi-dimensional feature information according to the basic attributes of the customer, the number of historical incentives, the content of historical incentives, and the historical interactive behavior indicators;

[0086] Among them, the second historical period is the next period of the first historical period.

[0087] In some embodiments, customer characteristics are calculated offline based on customer data of an interactive platform: first, characteristic information of the customer is extracted from statistical data of multiple dimensions such as the incentives previously obtained by the customer on the interactive platform, basic information of the customer, and actual interactive behaviors generated by the customer. Such characteristic information includes age, gender, number of historical incentives, content of incentives, interactive behavior indicators, etc.

[0088] In some embodiments, basic customer information may include: basic information, asset information, financial transaction information, fund flow information, wage information, mobile banking information, etc. Basic information may include: age, gender, place of origin, customer group, etc. Customer groups may include: retirees, people in fishing circles, etc. Asset information may include: current deposits, time deposits, financial management, funds, precious metals, loans, etc. Financial transaction information may include: number of credit card transactions, credit card transaction amount, number of transfer and remittance transactions, amount of transfer and remittance transactions, number of fund insurance and financial management transactions, amount of fund insurance and financial management transactions, etc. Fund flow information may include: education, medical care, travel, etc. Payroll information may include: amount of payroll, average monthly fluctuation, etc. Mobile banking information may include: number of login days, active columns, active functions, etc.

[0089] A customer portrait can be constructed based on the customer's basic information, and the generated customer portrait label can be used as multiple basic attributes of the customer.

[0090] See also Figure 3In some embodiments, step S102 may include but is not limited to steps S1021 to S1023:

[0091] Step S1021, inputting the multi-dimensional feature information into the pre-trained SCCL clustering model to obtain the customer category corresponding to each target customer;

[0092] Step S1022, dividing the target customers into multiple target customer groups according to customer categories;

[0093] Step S1023: generating a customer group feature vector according to the basic customer information of each target customer in the target customer group.

[0094] In some embodiments, target customers are deeply clustered based on the SCCL clustering model to divide target customer groups: the multidimensional feature information obtained by offline calculation is input into a pre-trained SCCL clustering model to obtain the customer category corresponding to each target customer, and the target customers are divided into multiple target customer groups according to the customer category; in addition, for each target customer group, a customer group feature vector is generated through the statistical characteristics of the basic customer information of each target customer in the group.

[0095] It should be noted that the statistical features here can be the mean, variance and other features of numerical data (such as age, deposit), or the probability features of numerical data in different numerical ranges, or the probability features of attribute data (such as gender, place of origin) under different labels, ultimately forming an overall customer group feature vector of the target object group.

[0096] SCCL is a clustering method that can simultaneously consider the relationship between samples and attributes for bidirectional optimization iteration. It can minimize the differences between clusters while maintaining the internal similarity of samples.

[0097] See also Figure 4 In some embodiments, the steps of training the SCCL clustering model may include but are not limited to steps S201 to S205:

[0098] Step S201, obtaining multiple sample customer feature data, the sample customer feature data including sample customer basic attributes, sample incentive times, sample incentive content, and sample interactive behavior indicators;

[0099] Step S202, performing data enhancement on each sample customer feature data to obtain corresponding enhanced customer feature data, and then constructing an enhanced data set based on the sample customer feature data and the enhanced customer feature data;

[0100] Step S203, constructing an SCCL clustering model, and determining a contrast loss function and a clustering loss function;

[0101] Step S204, taking the sample customer feature data and the corresponding enhanced customer feature data as positive sample pairs, taking other data samples in the enhanced data set except the positive sample pairs as negative samples, performing contrast learning and cluster learning on the positive sample pairs and the negative samples, and determining the contrast loss value through the contrast loss function, and determining the cluster loss value through the cluster loss function;

[0102] Step S205, updating the parameters of the SCCL clustering model according to the contrast loss value and the clustering loss value to obtain a trained SCCL clustering model.

[0103] In some embodiments, the overall framework of SCCL consists of three parts, including a neural network, a contrast loss function, and a clustering loss function. The neural network is used to map the input data to the representation space, and the contrast loss function and the clustering loss function are used to calculate the contrast loss value and the clustering loss value, respectively.

[0104] In some embodiments, the training data of the SCCL clustering model is composed of original data and enhanced data. Generate a corresponding enhanced customer feature data through data enhancement, and finally obtain an enhanced data set

[0105] In some embodiments, a sample customer feature data and the corresponding enhanced customer feature data are selected from the enhanced data set as a positive sample pair, and the remaining 2M-2 data samples are selected as negative samples. After being input into the SCCL clustering model, the contrast loss value is calculated through the contrast loss function to shorten the distance between the positive sample pairs and increase the distance between the positive samples and the negative samples, and the clustering loss value is calculated through the clustering loss function.

[0106] In some embodiments, a joint loss value of the SCCL clustering model is determined according to the contrast loss value and the clustering loss value, and the SCCL clustering model is optimized based on the joint loss value until the model converges or reaches a preset number of iterations to obtain a trained SCCL clustering model.

[0107] See also Figure 5 In some embodiments, the steps of training the causal inference model may include but are not limited to steps S301 to S303:

[0108] Step S301, obtaining multiple sample customer incentive data corresponding to each incentive strategy, the sample customer incentive data including sample customer feature vectors, incentive strategy coding data and corresponding sample interactive behavior indicators;

[0109] Step S302, determining the incentive cost of each incentive strategy, constructing a control group according to the incentive data of sample customers corresponding to the incentive strategy with the smallest incentive cost, and constructing corresponding multiple intervention groups according to the incentive data of sample customers corresponding to other incentive strategies;

[0110] Step S303: Perform reinforcement learning and causal inference based on the X-Learner model according to the control group and the intervention group to obtain a causal inference model.

[0111] In some embodiments, the three-element vectors of time, quantity and incentive value are independently numbered using One-Hot encoding and converted into binary form to represent different possible values, and then the incentive strategy encoding data is constructed using a time function.

[0112] In some embodiments, after determining the incentive strategy encoding data, a causal inference model is selected to estimate the causal effect. The embodiment of the present application uses the X-Learner model to model each incentive strategy, and uses the strategy with the smallest total incentive cost as the benchmark strategy to calculate and evaluate the improvement in the interactive effect of other incentive strategies, i.e., uplift. X-Learner is a model based on reinforcement learning and causal inference. It can estimate the influence of each strategy on core indicators from the observed data and make decisions based on the estimated results.

[0113] In some embodiments, first, multiple sample customer incentive data under each incentive strategy are determined, including sample customer feature vectors, incentive strategy encoding data, and corresponding sample interactive behavior indicators; then, the sample customer incentive data corresponding to the incentive strategy with the smallest incentive cost is used as a control group, and the remaining sample customer incentive data are divided into multiple intervention groups according to the similarities and differences in incentive strategies, and X-Learner models are established for the control group and the intervention group respectively; X-Learner first recognizes the possible information gap between the intervention group and the control group, and then estimates the missing effect of the intervention on the other group by using the data of each group, and then uses these estimates to predict the results to obtain a trained causal inference model.

[0114] See also Figure 6 In some embodiments, step S103 may include but is not limited to steps S1031 to S1033:

[0115] Step S1031, combining the feature vectors of each customer group and the incentive strategy codes corresponding to each incentive strategy in pairs and inputting them into the causal inference model to obtain the predicted values ​​of the interactive behavior indicators of each target customer group under each incentive strategy;

[0116] Step S1032, determining the predicted value of the interactive behavior index of each target customer group under the incentive strategy with the minimum incentive cost as the interactive behavior index benchmark value of each target customer group;

[0117] Step S1033, determining the interaction enhancement gain of each target customer group under each incentive strategy according to the difference between the interaction behavior indicator prediction value and the interaction behavior indicator benchmark value.

[0118] In some embodiments, the customer group feature vector of the target customer group A and the incentive strategy code corresponding to the incentive strategy C are input into the causal inference model to obtain the predicted value of the interactive behavior index of the target customer group A under the incentive strategy C output by the model.

[0119] In some embodiments, the predicted value of the interactive behavior index of the target customer group A under the incentive strategy with the lowest incentive cost is used as the interactive behavior index baseline value of the target customer group A, and the interactive enhancement gain of the target customer group A under the incentive strategy C can be determined based on the difference between the predicted value of the interactive behavior index and the baseline value of the interactive behavior index.

[0120] It should be noted that the interaction improvement gain corresponding to the incentive strategy with the lowest incentive cost is 0, and the interaction improvement gains corresponding to other incentive strategies may be positive or negative.

[0121] See also Figure 7 In some embodiments, step S104 may include but is not limited to steps S1041 to S1044:

[0122] Step S1041, constructing an objective function of incentive cost and interaction improvement gain, and determining an incentive cost constraint and a positive gain constraint;

[0123] Step S1042, constructing the initial population corresponding to each target customer group according to the incentive strategy and the corresponding interaction improvement gain, and determining the fitness function according to the objective function, the incentive cost constraint and the positive gain constraint;

[0124] Step S1043, using a genetic algorithm to optimize each initial population according to the fitness function to obtain the optimal solution corresponding to each target customer group;

[0125] Step S1044, determining the optimal incentive strategy for each target customer group based on the optimal solution.

[0126] In some embodiments, the optimal strategy is solved through multi-objective optimization: the predicted core indicator uplift of the target customer group under different incentive strategies, i.e., the interaction uplift gain, is used to complete multi-objective optimization using a genetic algorithm to solve the optimal incentive strategy that maximizes the difference between the benefit brought by the interaction uplift gain and the incentive cost. The target constraints include the incentive cost constraint (which must not exceed the upper limit value) and the strategy needs to generate positive gain (interaction uplift gain is not less than 0). This process aims to find the most appropriate, most effective or well-balanced specific incentive plan under multi-dimensional needs.

[0127] In some embodiments, an objective function of incentive cost and interaction improvement gain is constructed, where the objective function can be expressed as min(s*at), where s represents interaction improvement gain, a represents the conversion rate of interaction and benefit, and t represents the incentive cost. The incentive cost constraint can be expressed as t≤t max , t max represents the upper limit of the incentive cost, and the positive gain constraint can be expressed as s≥0.

[0128] In some embodiments, the initial feasible solution of the objective function is determined according to the incentive strategy and the corresponding interaction improvement gain, the initial feasible solution is encoded, the fitness function is evaluated and selected, a certain number of primary populations are randomly generated in the feasible solution region, the individual fitness in the population is calculated, and selection, crossover, mutation and dimensionality increase are repeatedly performed until the convergence condition is reached, the operation is stopped, and the optimal solution is obtained.

[0129] In some embodiments, the optimal incentive strategy for each target customer group is determined based on the optimal solution, and for each target customer in the target customer group, personalized adjustments can be made to the optimal incentive strategy to obtain a personalized incentive plan for each target customer.

[0130] It can be recognized that the embodiment of the present application extracts multi-dimensional feature information of customers based on their basic information, historical incentive data, and historical interactive behavior data, performs deep clustering of customers based on the multi-dimensional feature information to obtain multiple customer groups, determines the interaction enhancement gain of each customer group under different incentive strategies through a causal inference model, and obtains the optimal incentive strategy for each customer group through multi-objective optimization, so that the incentive plan for each customer group can be determined based on the optimal incentive strategy, thereby improving the accuracy of incentives issued to customers and the efficiency of attracting traffic to the interactive platform.

[0131] Compared with the prior art, the embodiments of the present application also have the following advantages:

[0132] 1) Data-driven decision-making: Incorporate a large amount of customer feature data and incentive information into model calculations, and establish a strategy improvement model with higher prediction accuracy through causal inference. Compared with the previous incentive plan based on overall statistical results, it can more accurately understand how each customer type responds to different forms of incentives.

[0133] 2) Personalized marketing: Through deep clustering and grouping, a large number of agents or institutions can be gradually divided into small-scale interactive subsystems, and differentiated operations and activation measures can be designed according to user characteristics. This is conducive to the realization of refined operations and personalized services.

[0134] 3) Improve customer interaction rate: By performing offline calculation of user characteristics and deep clustering of SCCL, we can better understand different types of customers and design personalized incentive plans for them. By estimating the improvement through causal models, we can select the most effective strategy to stimulate interaction between target users and agents, thereby improving the overall interaction rate.

[0135] See also Figure 8 The embodiment of the present application further provides an incentive scheme determination device based on causal inference, which can realize the above incentive scheme determination based on causal inference. The incentive scheme determination device based on causal inference includes:

[0136] A feature extraction module is used to obtain the target customer's basic information, historical incentive data, and corresponding historical interactive behavior data, and extract multi-dimensional feature information of the target customer based on the basic customer information, historical incentive data, and historical interactive behavior data;

[0137] The customer clustering module is used to perform deep clustering of target customers based on multi-dimensional feature information, obtain multiple target customer groups, and determine the customer group feature vector of each target customer group;

[0138] An interaction enhancement gain determination module is used to determine the interaction enhancement gain of each target customer group under each incentive strategy based on a pre-trained causal inference model, customer group feature vectors and multiple preset incentive strategies;

[0139] The optimization module is used to perform multi-objective optimization based on the incentive cost of each incentive strategy and the corresponding interaction improvement gain, obtain the optimal incentive strategy for each target customer group, and then determine the incentive plan for each target customer group based on the optimal incentive strategy.

[0140] The specific implementation of the device for determining an incentive scheme based on causal inference is basically the same as the specific implementation of the method for determining an incentive scheme based on causal inference mentioned above, and will not be repeated here.

[0141] The embodiment of the present application also provides an electronic device, the electronic device comprising: a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for realizing connection and communication between the processor and the memory, and the program is executed by the processor to realize the above-mentioned method for determining an incentive scheme based on causal inference. The electronic device can be any intelligent terminal including a tablet computer, a car computer, etc.

[0142] See also Fig. 9 , Fig. 9 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0143] The processor 901 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0144] The memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 902, and the processor 901 calls and executes the method for determining an incentive scheme based on causal inference in the embodiment of this application;

[0145] Input / output interface 903, used to implement information input and output;

[0146] Communication interface 904, used to realize communication interaction between the device and other devices, which can be realized by wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);

[0147] A bus 905 that transmits information between various components of the device (e.g., the processor 901, the memory 902, the input / output interface 903, and the communication interface 904);

[0148] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other in communication within the device via a bus 905 .

[0149] See also Fig.10The embodiment of the present application also provides a storage medium, which is a computer-readable storage medium used for computer-readable storage. The storage medium stores one or more programs 1001, and the one or more programs 1001 can be executed by one or more processors to implement the above-mentioned incentive scheme determination method based on causal inference.

[0150] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0151] The causal inference-based incentive scheme determination method, apparatus, device and medium provided in the embodiments of the present application obtain basic customer information, historical incentive data and corresponding historical interactive behavior data of target customers, and extract multi-dimensional feature information of target customers based on the basic customer information, historical incentive data and historical interactive behavior data, perform deep clustering of target customers based on the multi-dimensional feature information to obtain multiple target customer groups, and determine the customer group feature vector of each target customer group, determine the interaction enhancement gain of each target customer group under each incentive strategy based on the customer group feature vector and multiple preset incentive strategies through a pre-trained causal inference model, perform multi-objective optimization based on the incentive cost of each incentive strategy and the corresponding interaction enhancement gain, obtain the optimal incentive strategy for each target customer group, and then determine the incentive scheme for each target customer group based on the optimal incentive strategy. The embodiment of the present application extracts multi-dimensional feature information of customers based on their basic information, historical incentive data, and historical interactive behavior data, performs deep clustering of customers based on the multi-dimensional feature information to obtain multiple customer groups, determines the interaction enhancement gain of each customer group under different incentive strategies through a causal inference model, and obtains the optimal incentive strategy for each customer group through multi-objective optimization, so that the incentive plan for each customer group can be determined based on the optimal incentive strategy, thereby improving the accuracy of incentives issued to customers and the efficiency of attracting traffic to the interactive platform.

[0152] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0153] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0154] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0155] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0156] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device 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 devices.

[0157] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0158] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0159] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0160] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0161] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0162] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A method for determining an incentive scheme based on causal inference, characterized in that: The method comprises: Obtaining basic customer information, historical incentive data, and corresponding historical interactive behavior data of the target customer, and extracting multi-dimensional feature information of the target customer based on the basic customer information, the historical incentive data, and the historical interactive behavior data; Performing deep clustering on the target customers according to the multi-dimensional feature information to obtain multiple target customer groups, and determining a customer group feature vector for each target customer group; Determine the interaction enhancement gain of each target customer group under each incentive strategy through a pre-trained causal inference model based on the customer group feature vector and a plurality of preset incentive strategies; Multi-objective optimization is performed based on the incentive cost of each incentive strategy and the corresponding interaction improvement gain to obtain the optimal incentive strategy for each target customer group, and then the incentive plan for each target customer group is determined based on the optimal incentive strategy.

2. The method for determining an incentive scheme based on causal inference according to claim 1, characterized in that: The acquiring of the target customer's basic information, historical incentive data, and corresponding historical interactive behavior data, and extracting the target customer's multi-dimensional feature information based on the basic customer information, the historical incentive data, and the historical interactive behavior data, includes: Acquire the historical incentive data of the target customer in a preset first historical period and the historical interactive behavior data in a preset second historical period through the interactive platform; Determine the number of historical incentives and the content of historical incentives based on the historical incentive data, and determine the historical interactive behavior indicators based on the historical interactive behavior data; Acquire basic customer information of the target customer, and determine multiple basic customer attributes according to the basic customer information; Determine the multi-dimensional feature information according to the basic attributes of the customer, the number of historical incentives, the historical incentive content, and the historical interactive behavior indicators; The second historical period is the next period of the first historical period.

3. The method for determining an incentive scheme based on causal inference according to claim 1, characterized in that: The step of performing deep clustering on the target customers according to the multi-dimensional feature information to obtain a plurality of target customer groups, and determining a customer group feature vector for each of the target customer groups, comprises: Inputting the multi-dimensional feature information into a pre-trained SCCL clustering model to obtain a customer category corresponding to each target customer; Dividing the target customers into a plurality of target customer groups according to the customer categories; The customer group feature vector is generated according to the basic customer information of each target customer in the target customer group.

4. The method for determining an incentive scheme based on causal inference according to claim 3 is characterized in that: The SCCL clustering model is trained by the following steps: Acquire multiple sample customer feature data, wherein the sample customer feature data includes sample customer basic attributes, sample incentive times, sample incentive content, and sample interactive behavior indicators; Performing data enhancement on each of the sample customer feature data to obtain corresponding enhanced customer feature data, and then constructing an enhanced data set based on the sample customer feature data and the enhanced customer feature data; Construct the SCCL clustering model and determine the contrast loss function and clustering loss function; Taking the sample customer feature data and the corresponding enhanced customer feature data as a positive sample pair, taking other data samples in the enhanced data set except the positive sample pair as negative samples, performing contrast learning and cluster learning on the positive sample pair and the negative samples, and determining a contrast loss value through the contrast loss function, and determining a cluster loss value through the cluster loss function; The parameters of the SCCL clustering model are updated according to the contrast loss value and the clustering loss value to obtain the trained SCCL clustering model.

5. The method for determining an incentive scheme based on causal inference according to claim 1, characterized in that: The causal inference model is trained by the following steps: Acquire multiple sample customer incentive data corresponding to each of the incentive strategies, wherein the sample customer incentive data includes a sample customer feature vector, incentive strategy encoding data, and a corresponding sample interactive behavior indicator; Determine the incentive cost of each of the incentive strategies, construct a control group according to the incentive data of the sample customers corresponding to the incentive strategy with the smallest incentive cost, and construct corresponding multiple intervention groups according to the incentive data of the sample customers corresponding to the other incentive strategies; According to the control group and the intervention group, reinforcement learning and causal inference are performed based on the X-Learner model to obtain the causal inference model.

6. The method for determining an incentive scheme based on causal inference according to claim 5, characterized in that: The method of determining the interaction enhancement gain of each target customer group under each incentive strategy based on the customer group feature vector and a plurality of preset incentive strategies through the pre-trained causal inference model includes: The characteristic vectors of each customer group and the incentive strategy codes corresponding to each incentive strategy are input into the causal inference model in pairs to obtain the predicted values ​​of the interactive behavior indicators of each target customer group under each incentive strategy; Determine the predicted value of the interactive behavior index of each target customer group under the incentive strategy with the minimum incentive cost as the interactive behavior index benchmark value of each target customer group; The interaction enhancement gain of each target customer group under each incentive strategy is determined according to the difference between the predicted value of the interaction behavior indicator and the baseline value of the interaction behavior indicator.

7. The method for determining an incentive scheme based on causal inference according to any one of claims 1 to 6, characterized in that: The multi-objective optimization is performed according to the incentive cost of each incentive strategy and the corresponding interaction improvement gain to obtain the optimal incentive strategy for each target customer group, including: Construct an objective function about incentive cost and interaction improvement gain, and determine the incentive cost constraint and positive gain constraint; Constructing an initial population corresponding to each target customer group according to the incentive strategy and the corresponding interaction improvement gain, and determining a fitness function according to the objective function, the incentive cost constraint and the positive gain constraint; Utilizing a genetic algorithm to optimize each of the initial populations according to the fitness function to obtain an optimal solution corresponding to each of the target customer groups; The optimal incentive strategy for each target customer group is determined according to the optimal solution.

8. A device for determining an incentive scheme based on causal inference, characterized in that: The device comprises: A feature extraction module is used to obtain basic customer information, historical incentive data and corresponding historical interactive behavior data of a target customer, and extract multi-dimensional feature information of the target customer based on the basic customer information, the historical incentive data and the historical interactive behavior data; A customer clustering module, used to perform deep clustering on the target customers according to the multi-dimensional feature information, obtain multiple target customer groups, and determine a customer group feature vector for each target customer group; An interaction enhancement gain determination module, used to determine the interaction enhancement gain of each target customer group under each incentive strategy based on the customer group feature vector and a plurality of preset incentive strategies through a pre-trained causal inference model; The optimization module is used to perform multi-objective optimization according to the incentive cost of each incentive strategy and the corresponding interaction improvement gain, obtain the optimal incentive strategy for each target customer group, and then determine the incentive plan for each target customer group according to the optimal incentive strategy.

9. An electronic device, characterized in that: The electronic device includes a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for realizing connection and communication between the processor and the memory. When the program is executed by the processor, the steps of the method for determining an incentive scheme based on causal inference as described in any one of claims 1 to 7 are realized.

10. A storage medium, the storage medium being a computer-readable storage medium, used for computer-readable storage, characterized in that: The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of the incentive scheme determination method based on causal inference as described in any one of claims 1 to 7.

Images