Customer state prediction method and device based on Bayesian network, equipment and medium

Abstract

The invention provides a customer state prediction method and device based on a Bayesian network, equipment and a medium, and relates to application of a machine learning model in the field of financial science and technology. The customer state prediction method based on the Bayesian network comprises the following steps: acquiring first service data of a target customer in a first time period; inputting the first service data into a pre-trained Bayesian network model to obtain a first probability that the target customer is in each target state at the first time node; based on the first probability of each target state, determining a service intervention operation for the target customer; and in response to execution of the service intervention operation, obtaining second service data of the target customer in a second time period, and inputting the second service data into a pre-trained Bayesian network model to obtain a second probability that the target customer is in each target state at a second time node.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of machine learning, can be applied to the field of financial technology, and more particularly to a customer state prediction method and device based on machine learning, equipment and medium. BACKGROUND

[0002] In the financial field, with the increasing opening of the financial market and the intensifying competition, the stability of the customer resources of a financial institution has become a core element affecting its profitability and market competitiveness. Customer loss not only directly leads to a reduction in the economic benefits of various business incomes, but also damages the brand reputation and long-term development potential of the financial institution. Therefore, how to accurately and efficiently identify and predict whether there is a risk of loss in the customer group, and take effective intervention operations accordingly, has become a key problem in customer relationship management for the financial institution. SUMMARY

[0003] In view of the above problems, the present application provides a customer state prediction method and device based on a Bayesian network, equipment and medium.

[0004] The first aspect of the present application provides a customer state prediction method based on a Bayesian network, the method comprising: acquiring first business data of a target customer in a first time period; inputting the first business data into a pre-trained Bayesian network model to obtain a first probability of the target customer being in each target state at a first time node; determining a business intervention operation for the target customer based on the first probability of each target state; in response to the business intervention operation being executed, acquiring second business data of the target customer in a second time period, and inputting the second business data into the pre-trained Bayesian network model to obtain a second probability of the target customer being in each target state at a second time node; wherein the first time node is after the first time period, the second time node is after the first time node, and the second time period is at least the time period between the first time node and the second time node.

[0005] According to the embodiments of the present application, the business intervention operation for the target customer is determined based on the first probability of each target state, comprising: determining a risk label of the target customer according to the first probability of each target state, the risk label representing the risk degree of the target customer losing at the first time node; matching an intervention operation corresponding to the risk label in an operation database as the business intervention operation for the target customer; wherein the operation database is configured with a plurality of different types of intervention operations, each intervention operation has different priority, and the priority of the intervention operation is used to match the corresponding risk label.

[0006] According to an embodiment of the present application, the method further comprises: inputting the first service data and the first probability of each target state into the pre-trained Bayesian network model simultaneously to obtain a third probability of the target customer being in each target state at a second time node without performing the service intervention operation; comparing the second probability and the third probability of each target state to determine a probability change value of each target state in the case of performing the service intervention operation.

[0007] Based on the probability change value of each target state, the priority of the service intervention operation is adjusted in the operation database.

[0008] According to an embodiment of the present application, the target state comprises a stable state, a risk state and a loss state; and based on the probability change value of each target state, the priority of the service intervention operation is adjusted in the operation database, which comprises: if the probability change value of the risk state and / or the loss state represents that the second probability is less than the third probability, the priority of the service intervention operation is improved based on the change size of the probability change value; and if the probability change value of the stable state represents that the second probability is greater than the third probability, the priority of the service intervention operation is improved based on the change size of the probability change value.

[0009] According to an embodiment of the present application, before determining the service intervention operation for the target customer based on the first probability of each target state, the method further comprises: for each intervention operation in the operation database, inputting each intervention operation and the first service data into the pre-trained Bayesian network model to obtain a fourth probability of the target customer being in each target state at the first time node in the case of performing each intervention operation; comparing the first probability and the fourth probability of each target state to determine a probability change value of each target state in the case of performing each intervention operation; and determining the priority of each intervention operation in the operation database based on the probability change value of each target state; wherein each prediction operation is configured to be performed at a third time node, and the third time node is a time node before the first time node.

[0010] According to an embodiment of the present application, the Bayesian network model is trained in the following manner: obtaining historical service data of the target customer, the historical service data comprising customer basic information, customer behavior data, service data, service intervention operation for each customer and customer label; performing feature processing on each type of service data to obtain feature data of each type, so as to construct a feature data set; training the Bayesian network model based on the feature data set; wherein the Bayesian network model is constructed based on the following function: a state change function representing a change relationship of the first probability of each target state of the target customer at the first time node to the second probability of each target state at the second time node, the state change function being defined at least by the feature data of each type.

[0011] According to an embodiment of the present application, after obtaining the first probability of the target customer being in each target state at the first time node, the method further comprises: adding the first business data to the historical business data to obtain updated historical business data, and re-determining each type of feature data according to the updated historical business data to update the feature data set; in the case of updating the feature data set, adjusting the state change function to update the Bayesian network model to obtain an updated Bayesian network model.

[0012] The second aspect of the present application provides a customer state prediction device based on a Bayesian network, the device comprising: a data acquisition module configured to acquire first business data of a target customer within a first time period; a first model application module configured to input the first business data into a pre-trained Bayesian network model to obtain a first probability of the target customer being in each target state at a first time node; an operation determination module configured to determine a business intervention operation for the target customer based on the first probability of each target state; and a second model application module configured to, in response to the business intervention operation being executed, acquire second business data of the target customer within a second time period and input the second business data into the pre-trained Bayesian network model to obtain a second probability of the target customer being in each target state at a second time node; wherein the first time node is after the first time period, the second time node is after the first time node, and the second time period is at least a time period between the first time node and the second time node.

[0013] The third aspect of the present application provides an electronic device comprising: one or more processors; and a memory configured to store one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the method.

[0014] The fourth aspect of the present application further provides a computer-readable storage medium having stored thereon a computer program or instructions, wherein the computer program or instructions are executed by a processor to implement the steps of the method.

[0015] The fifth aspect of the present application further provides a computer program product comprising a computer program or instructions, wherein the computer program or instructions are executed by a processor to implement the steps of the method. BRIEF DESCRIPTION OF DRAWINGS

[0016] The above and other objects, features and advantages of the present application will become more apparent from the following description of embodiments of the present application, taken in conjunction with the accompanying drawings, in which:

[0017] Figure 1 An application scenario diagram of the customer state prediction method, device, equipment and medium based on a Bayesian network according to an embodiment of the present application is schematically shown;

[0018] Figure 2 Fig. 1 schematically shows one of flowcharts of a method for predicting customer state based on Bayesian network according to an embodiment of the present application;

[0019] Figure 3 Fig. 2 schematically shows another of flowcharts of a method for predicting customer state based on Bayesian network according to an embodiment of the present application;

[0020] Figure 4 Fig. 3 schematically shows a block diagram of an apparatus for predicting customer state based on Bayesian network according to an embodiment of the present application;

[0021] Figure 5 Fig. 4 schematically shows a block diagram of an electronic device suitable for implementing the method for predicting customer state based on Bayesian network according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] Hereinafter, embodiments of the present application will be described with reference to the accompanying drawings. It is to be understood, however, that the description is merely exemplary of the present application, and is not intended to limit the scope of the present application. In the following detailed description of the embodiments of the present application, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, it will be apparent to one skilled in the art that the present application can be practiced without these specific details. In other instances, well-known structures and functions have not been described in detail in order to avoid obscuring aspects of the present application.

[0023] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the present application. As used herein, the term "including" "comprising" and the like are meant to be inclusive, but not limiting to the components, steps, operations and / or functions that are described, but that additional or other components, steps, operations, and / or functions can also be utilized.

[0024] All terms used herein, including technical and scientific terms, have the meanings commonly understood by one of ordinary skill in the art unless otherwise defined herein. It should be noted that the terms used herein are defined as having meanings that are consistent with the context of the specification in which the terms are used, and should not be interpreted in an idealized or overly formal way.

[0025] In instances where a term similar to "at least one of A, B, and C, etc." is used, in general, it should be interpreted to include one or more of A, B, or C in addition to the possibility that "A, B, and C each by themselves are included." (e.g., a system having "at least one of A, B, and C" shall mean a system herein having A alone, solely B, solely C, A and B together, A and C together, B and C together, and / or A, B, and C together, etc.).

[0026] In the technical solutions of the present application, the user information (including but not limited to user personal information, user image information, user device information such as location information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved are all information and data authorized by the user or authorized by all parties, and the collection, storage, use, processing, transmission, provision, application, and application of related data comply with relevant laws, regulations, and standards, take necessary security measures, do not violate public order and good customs, and provide corresponding operation portals for users to choose authorization or refusal.

[0027] In the scenario of making automated decisions using personal information, the method, device, and system provided by the embodiments of the present application all provide corresponding operation portals for users to choose to agree or refuse the automated decision result; if the user chooses to refuse, the expert decision process is entered. The expression "automated decision" here refers to the activity of automatically analyzing, evaluating the behavior habits, interests and hobbies, or economic, health, and credit conditions of individuals, and making decisions through computer programs. The expression "expert decision" here refers to the activity of making decisions by personnel who are engaged in a certain field of work, have specialized experience, knowledge, and skills, and have reached a certain level of professional expertise.

[0028] Figure 1 An application scenario diagram of a customer state prediction method, device, equipment, medium, and program product based on a Bayesian network according to an embodiment of the present application is schematically shown.

[0029] As Figure 1 shown, the application scenario 100 according to the embodiment can include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 is a medium for providing a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 can include various connection types, such as wired, wireless communication links, or optical fiber cables, etc.

[0030] The user can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only as examples).

[0031] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be various electronic devices with displays and support web browsing, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0032] Server 105 can be a server that provides various services, such as a backend management server that supports websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (this is just an example). The backend management server can analyze and process data such as received user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices.

[0033] It should be noted that the customer state prediction method based on Bayesian networks provided in this application embodiment can generally be executed by server 105. Correspondingly, the customer state prediction device based on Bayesian networks provided in this application embodiment can generally be located in server 105. The customer state prediction method based on Bayesian networks provided in this application embodiment can also be executed by a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105. Correspondingly, the customer state prediction device based on Bayesian networks provided in this application embodiment can also be located in a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105.

[0034] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0035] It should be noted that the customer state prediction method based on Bayesian networks in this application can be applied to the fintech field, and also to cross-business scenarios between fintech and other related fields. For example, by combining e-commerce platform consumption records, shopping preferences, and credit payment behavior, changes in customer spending power and financial needs can be analyzed to provide dynamic financial product recommendations for e-commerce platform financial users.

[0036] The following will be based on Figure 1 The described scene, through Figures 2-3 A detailed description is provided of a customer state prediction method based on Bayesian networks according to embodiments of this application.

[0037] Figure 2Fig. 1 shows a flowchart of a method for predicting customer state based on Bayesian network according to an embodiment of the present application.

[0038] As shown in Fig. 1, the method for predicting customer state based on Bayesian network according to an embodiment of the present application comprises operations S201-S204. Figure 2

[0039] In operation S201, first business data of a target customer in a first time period is obtained.

[0040] In an embodiment of the present application, the business data can be various types of information related to customer financial activities, including but not limited to: customer basic information, account information, transaction records, service records (such as complaint times, customer service contacts), product usage, customer feedback, etc.

[0041] In operation S202, the first business data is input into a pre-trained Bayesian network model to obtain a first probability of the target customer being in each target state at a first time node.

[0042] In an embodiment of the present application, the Bayesian network model is a machine learning model based on a probability graph, which represents the dependency relationship between variables (such as customer state, behavior characteristics, intervention operation) through a directed acyclic graph, and uses Bayesian theorem for probability reasoning; wherein, in the present application, the Bayesian network model can process time series data and predict the change of customer state over time through a dynamic Bayesian network.

[0043] In an embodiment of the present application, the target state represents the classification state that the customer may be in at a specific time node, including but not limited to: stable state, the customer has a good relationship with the financial institution and shows no signs of attrition; risk state, the customer shows potential attrition tendency and needs attention and intervention; attrition state, the customer has terminated the main business relationship with the financial institution.

[0044] In operation S203, a business intervention operation for the target customer is determined based on the first probability of each target state.

[0045] In an embodiment of the present application, the business intervention operation is a specific action or strategy taken by the financial institution to reduce the risk of customer attrition. For example, recommending financial products, providing rate discounts, sending care messages, arranging customer manager phone follow-up, providing exclusive financial services, etc.

[0046] ​In operation S204, in response to the service intervention operation being performed, second service data of the target customer in a second time period is obtained, and the second service data is input into the pre-trained Bayesian network model to obtain second probabilities of the target customer being in each target state at a second time node; wherein the first time node is after the first time period, the second time node is after the first time node, and the second time period is at least a time period between the first time node and the second time node.

[0047] For example, all relevant data of the target customer in the past three months (i.e., the "first time period") is extracted. These data include but are not limited to: account balance changes, transaction numbers and amounts, held financial products, login mobile banking frequencies, customer service consultation records, etc. The first service data is standardized and then input into the pre-deployed dynamic Bayesian network model. The model runs the inference algorithm and outputs the probabilities of the target customer being in the "stable", "risk", and "churn" states at the end of the next month (i.e., the "first time node"), denoted as P1(stable), P1(risk), and P1(churn), to obtain the first probabilities. According to the first probabilities, it is determined that the customer is in a high risk if P1(risk)+P1(churn)>0.7, for example. Subsequently, the operation database is queried to select one or more appropriate service intervention operations according to preset rules (such as risk level matching), such as "pushing a high-yield short-term financial product to the customer". When the intervention operation is performed, the financial institution can perform the above intervention operation through a mobile application message push, a short message, or a customer manager call, etc.

[0048] Further, new behavior data of the customer is continuously collected in the one month after the intervention is performed (i.e., the second time period, which ends at the second time node), such as whether the customer has viewed the financial product or whether the customer has transferred funds, etc. At the second time node, these new data are input into the Bayesian network model again as the second service data to obtain new prediction probabilities of the customer being in each state at the time node, including P2(stable), P2(risk), and P2(churn), to obtain the second probabilities.

[0049] By using the above method, the service data of the target customer in a specific time period is obtained, and with the aid of the pre-trained Bayesian network model, the probabilities of the customer being in each target state at different time nodes can be obtained, the customer state can be accurately predicted, the financial institution can be provided with a basis for understanding the customer situation in advance, and the financial institution can timely master the customer dynamics and prepare for possible customer churn, etc. At the same time, the service intervention operation is determined based on the predicted probabilities of each target state, and the service data is obtained again after the intervention is performed to make a prediction, forming a closed-loop customer state prediction and intervention process, so that the financial institution can timely adjust the strategy according to the customer state changes, take targeted measures to maintain the customer relationship, and effectively reduce the customer churn risk.

[0050] According to the embodiments of the present application, in operation S203, the business intervention operation for the target customer is determined based on the first probability of each target state, including: determining the risk label of the target customer according to the first probability of each target state, the risk label representing the risk degree of the target customer existing in the first time node; matching the intervention operation corresponding to the risk label in the operation database as the business intervention operation for the target customer; wherein the operation database is configured with a plurality of different types of intervention operations, and the priority of each intervention operation is different, and the priority of the intervention operation is used to match the corresponding risk label.

[0051] In the embodiments of the present application, the priority of the intervention operation represents the importance degree or the recommended order of different business intervention operations in the operation database. The intervention operation with high priority can be considered more effective or cost-effective for customers with a certain risk level.

[0052] Specifically, based on the first probability of each target state, the risk label representing the risk degree of the target customer existing in the first time node is generated by the risk label determination module. Then, the intervention operation matched with the risk label in the operation database is retrieved and used as the business intervention operation for the target customer. The operation database stores a plurality of different types of intervention operations, and each intervention operation is assigned a different priority, which is used to match the corresponding risk label, so as to realize the accurate push of the intervention strategy.

[0053] For example, a comprehensive risk score is calculated according to the first probability. For example, a weighted formula is used: risk score = 0.3*P1(risk) + 0.7*P1(churn). Then, the risk label is divided according to the score range: score < 0.3, low risk; 0.3 ≤ score < 0.6, medium risk; score ≥ 0.6, high risk. Access the operation database, which stores the recommended intervention operations corresponding to different risk labels and their priorities. For example: high risk, priority 1: exclusive financial consultant phone call; priority 2: give 100 yuan of financial coupons. Medium risk, priority 1: push customized financial products; priority 2: send a caring message. Low risk, priority 1: no operation; priority 2: recommend points exchange activities.

[0054] By using the above method, the risk label is determined according to the first probability of the target customer in each target state, and the risk degree of customer churn is quantified, so that the risk degree is clear and intuitive. Then, the intervention operation corresponding to the risk label is matched in the operation database, which can provide precise and adaptive intervention operations for customers with different risk degrees, improve the pertinence and effectiveness of intervention, and avoid resource waste.

[0055] Figure 3Fig. 2 shows a flowchart of a method for predicting customer state based on a Bayesian network according to an embodiment of the present application.

[0056] According to an embodiment of the present application, as shown in Fig. 1, the method for predicting customer state based on a Bayesian network comprises the following steps. Figure 3

[0057] In operation S301, the first business data and the first probability of each target state are input into the pre-trained Bayesian network model to obtain the third probability of the target customer being in each target state at the second time node without performing the business intervention operation.

[0058] In operation S302, the second probability and the third probability of each target state are compared to determine the probability change value of each target state in the case of performing the business intervention operation.

[0059] In operation S303, the priority of the business intervention operation is adjusted in the operation database based on the probability change value of each target state.

[0060] Specifically, the first business data and the first probability of the target customer being in each target state at the first time node are input into the pre-trained Bayesian network model to simulate that the model calculates the third probability of the target customer being in each target state at the second time node without performing the business intervention operation, thereby providing a comparison benchmark for evaluating the effect of the business intervention operation. Then, the second probability of each target state obtained after performing the business intervention operation is compared with the third probability obtained without performing the intervention operation, and the probability change value of each target state in the case of performing the business intervention operation is determined through a specific calculation method. The probability change value reflects the degree of influence of the business intervention operation on the probability of the target customer being in different states. According to the probability change value of each target state, the priority of the business intervention operation is adjusted in the operation database according to a preset adjustment rule. In this way, the intervention strategy can be dynamically optimized according to the actual effect of the business intervention operation, and the pertinence and effectiveness of the intervention can be improved.

[0061] For example, while performing the intervention operation, the first business data and the first probability are input into the model, but the variable of the business intervention operation is not included (or it is set to no intervention). The model performs a counterfactual reasoning to simulate the state probability of the customer at the second time node if no measures are taken, obtaining P3(stable), P3(risk), and P3(churn), i.e., the third probability. The probability difference value between the probability after performing the intervention and the probability without intervention, i.e., the probability change value, is calculated as follows:

[0062] Δ = P2(stable) - P3(stable); ​

[0063] Δ = P2(risk) - P3(risk);

[0064] Δ = P2(churn) - P3(churn);

[0065] Based on the above change value, the effectiveness of the intervention operation can be determined. For example, if Δ = +0.25 (25% increase in stable probability), and Δ = -0.20 (20% reduction in churn probability), the intervention operation is considered to be more effective.

[0066] By using the above method, the first business data and the first probability of each target state are input into the model to obtain the third probability of the customer being in each target state at the second time node without performing the intervention operation. The second probability after performing the intervention operation is compared to determine the probability change value, so that the actual impact effect of the business intervention operation on the customer state can be clearly evaluated, helping the financial institution to understand whether the intervention measures taken are effective and in which aspects they have an effect. At the same time, based on the probability change value, the priority of the business intervention operation is adjusted in the operation database, so that the financial institution can dynamically optimize the intervention strategy according to the actual intervention effect. If a certain intervention operation significantly increases the probability of the customer moving to a more favorable state, the priority thereof is increased so as to be preferentially used in subsequent similar situations.

[0067] According to an embodiment of the present application, the target state includes a stable state, a risk state, and a churn state; in operation S303, based on the probability change value of each target state, the priority of the business intervention operation is adjusted in the operation database, including: if the probability change value of the risk state and / or the churn state represents that the second probability is less than the third probability, the priority of the business intervention operation is increased based on the change size of the probability change value; if the probability change value of the stable state represents that the second probability is greater than the third probability, the priority of the business intervention operation is increased based on the change size of the probability change value.

[0068] In the embodiments of the present application, if the probability change value of the risk state and / or the churn state indicates that the second probability after performing the intervention operation is less than the third probability without performing the intervention operation, that is, the intervention operation reduces the probability of the target customer being in the risk state and / or the churn state, then according to the specific change size of the probability change value, the priority of the business intervention operation is increased according to the set rules. This means that the intervention operation that has a positive effect on reducing the risk of customer churn will have its priority improved so as to be preferentially used in subsequent similar situations. If the probability change value of the stable state indicates that the second probability after performing the intervention operation is greater than the third probability without performing the intervention operation, that is, the intervention operation increases the probability of the target customer being in the stable state, and based on the change size of the probability change value, the priority of the business intervention operation is increased according to the set rules.

[0069] For example, the "financial coupon giving" strategy originally having a priority of 2 will have its priority increased to 1 if it shows a significant effect of reducing the churn probability for multiple high-risk customers. If a certain intervention operation has a long-term poor effect (the change value is close to 0 or negative), the priority of the intervention operation will be reduced or marked as "to be evaluated".

[0070] By using the above method, the priority of the business intervention operation is adjusted according to the probability change value of different target states, which can more specifically and targetedly optimize the intervention strategy. When the probability of the risk state and / or the churn state is reduced due to the intervention operation, or the probability of the stable state is increased due to the intervention operation, the priority of the corresponding intervention operation is increased based on the probability change size, so that the financial institution can focus on strengthening the intervention measures that have a positive effect on improving the customer state, further improve the effect of customer relationship management, reduce the risk of customer churn, and improve the stability of customers.

[0071] According to the embodiments of the present application, before determining the business intervention operation for the target customer based on the first probability of each target state, the method further comprises: inputting each intervention operation and the first business data into a pre-trained Bayesian network model to obtain a fourth probability of the target customer being in each target state at the first time node in the case of performing each intervention operation; comparing the first probability and the fourth probability of each target state to determine the probability change value of each target state in the case of performing each intervention operation; and determining the priority of each intervention operation in the operation database based on the probability change value of each target state; wherein each prediction operation is configured to be performed at a third time node, and the third time node is a time node before the first time period.

[0072] Specifically, for each intervention operation in the operation database, the intervention operation is input into the pre-trained Bayesian network model together with the first business data. Based on the input, the Bayesian network model calculates the fourth probability that the target customer is in each target state at the first time node under the condition that each intervention operation is performed.

[0073] Further, the first probability of each target state is compared with the fourth probability obtained after performing the corresponding intervention operation, and the probability change value of each target state under the condition that each intervention operation is performed is determined through a specific calculation method. The probability change value reflects the degree of change of the probability of each target state of the target customer under the condition that each intervention operation is performed, and provides a quantitative index for evaluating the effect of the intervention operation.

[0074] Finally, the priority of each intervention operation in the operation database is determined according to the probability change value of each target state. Since each prediction operation is set to be performed at the third time node (which is before the first time period), through the priority determination mechanism, the intervention strategy can be optimized before the intervention operation is actually performed, and the intervention measure that is most beneficial to improving the customer state can be selected.

[0075] By using the above method, the probability change value of each target state determines the priority of each intervention operation, which helps the financial institution to screen and optimize the intervention strategy in advance, select the intervention operation that is most likely to improve the customer state and reduce the risk of loss, and improve the effectiveness of customer relationship management.

[0076] According to an embodiment of the present application, the Bayesian network model is trained by the following steps:

[0077] Step one, obtain the historical business data of the target customer, the historical business data including customer basic information, customer behavior data, service data, business intervention operation for each customer and customer label.

[0078] Step two, for each type of business data, feature processing is performed respectively to obtain feature data of each type, so as to construct a feature data set.

[0079] Step three, train the Bayesian network model based on the feature data set; wherein the Bayesian network model is constructed based on the following function: state change function, the state change function representing the change relationship of the first probability of each target state of the target customer from the first time node to the second probability of each target state of the target customer at the second time node, and the state change function being defined at least by each type of feature data.

[0080] In the embodiments of the present application, the historical business data is used to train the Bayesian network model, and the past accumulated customer business data set. Among them, the customer basic information includes the static attribute data of the customer, such as age, occupation, income, etc.; the customer behavior data includes the dynamic behavior record of the customer in the financial activity, such as transaction, login, etc.; the service data includes the data of the customer interacting with the financial institution service, such as complaint, consultation, etc.; and the customer label includes the real state or result label of the customer in the historical data, such as whether to lose.

[0081] In the embodiments of the present application, the feature processing is to extract the features suitable for model input by performing cleaning, conversion, encoding and other operations on the original business data; and the feature data set for model training is formed after the feature processing.

[0082] In the embodiments of the present application, the state change function refers to a function defined in the Bayesian network model, which describes how the state of the customer evolves over time and external factors

[0083] Specifically, the historical business data of the target customer is collected, which covers customer basic information, customer behavior data, service data, business intervention operations implemented for each customer, and customer labels and other aspects. For different types of business data, appropriate feature processing methods are used to convert them into feature data that can be understood and processed by the Bayesian network model, and then a feature data set is constructed. Feature processing aims to extract key features in the data so that the model can better learn the patterns and rules in the data. Finally, the Bayesian network model is trained using the constructed feature data set.

[0084] In the embodiments of the present application, the Bayesian network model is constructed based on the state change function, which describes the change relationship of the first probability of each target state of the target customer at the first time node to the second probability of each target state of the target customer at the second time node, and is defined at least by various types of feature data. Through training, the model can accurately predict the change of the target customer state based on the input feature data. See the following formula for details.

[0085]

[0086] Among them, is the state change function, are different types of feature data, characterizes the state of the target customer at t time node as s, characterizes the state of the target customer at t+1 time node as , represents the probability of the result appears under the given condition , an edge probability of the feature X, a prior probability of the target state Y, a posterior probability between each target state and each feature data, a likelihood probability between each feature data and each target state.

[0087] wherein, .

[0088] In the embodiments of the present application, in the model training stage, the correlation functions of each feature data and each target state can be calculated by Bayes theorem. Including but not limited to: the prior probability of each target state, the posterior probability between each target state and each feature data, and the posterior probability likelihood probability between each feature data and each target state, also including the edge probability of each feature data.

[0089] By using the above method, the rich historical business data of the target customer is obtained, including customer basic information, behavior data, service data, business intervention operation and customer label, and various types of business data are processed to construct a feature data set, which can fully mine useful information in the historical data, provide comprehensive and accurate data support for the training of the Bayesian network model, and make the model learn the complex relationship between the customer state change and various factors, thereby improving the accuracy and reliability of the model prediction. Based on the state change function, the Bayesian network model is constructed, which accurately describes the change relationship of the target customer from the first time node to the second time node of each target state probability through various feature data definitions, and provides a scientific and reasonable structure and logic framework for the model.

[0090] According to the embodiments of the present application, after obtaining the first probability of the target customer being in each target state at the first time node; the customer state prediction method based on machine learning further comprises: adding the first business data to the historical business data to obtain updated historical business data, and re-determining each type of feature data according to the updated historical business data to update the feature data set; in the case of updating the feature data set, adjusting the state change function to update the Bayesian network model to obtain an updated Bayesian network model.

[0091] Specifically, after obtaining the first probability of each target state, the obtained first business data is added to the original historical business data to obtain updated historical business data. Based on the updated historical business data, the feature processing method is re-applied to determine the feature data of each type, and then the feature data set is updated so that the data set can reflect the latest customer business information. In the case of updating the feature data set, the state change function is adjusted according to the new feature data. Since the Bayesian network model is constructed based on the state change function, the adjustment of the state change function will lead to the change of the model structure or parameters, thereby realizing the update of the Bayesian network model and obtaining the updated Bayesian network model.

[0092] Through the above method, the newly obtained first business data is added to the historical business data, the feature data is re-determined to update the feature data set, and the state change function is adjusted according to the data set update, and then the Bayesian network model is updated. The model can continuously learn the new rules of customer behavior and state change as new data is continuously accumulated, and the accuracy and effectiveness of the model are maintained.

[0093] In the embodiments of the present application, in addition to the traditional customer transaction data and basic information data, more dimensional data can also be included, such as after obtaining the information use permission granted by the customer, obtaining the customer's interaction data on social media related to the bank, and the industry climate data of the customer. The development trend data of the industry in which the customer is located can also be obtained. For example, for an enterprise customer engaged in manufacturing industry, if the industry climate is declining, combined with the behavior data such as the flow of funds of the bank account, the customer churn risk can be more comprehensively evaluated.

[0094] Based on the above customer state prediction method based on Bayesian network, the present application provides a customer state prediction device based on Bayesian network, which will be described in detail below. Figure 4 The device will be described in detail.

[0095] Figure 4 The structure block diagram of the customer state prediction device based on Bayesian network according to the embodiments of the present application is schematically shown.

[0096] As Figure 4 shown, the customer state prediction device based on Bayesian network 400 of the embodiments includes a data acquisition module 410, a first model application module 420, an operation determination module 430, and a second model application module 440.

[0097] According to the embodiments of the present application, the data acquisition module 410 is configured to acquire first business data of a target customer in a first time period;

[0098] The first model application module 420 is configured to input the first service data into the pre-trained Bayesian network model to obtain a first probability of the target customer being in each target state at a first time node.

[0099] The operation determination module 430 is configured to determine a service intervention operation for the target customer based on the first probability of each target state.

[0100] The second model application module 440 is configured to, in response to the service intervention operation being performed, acquire second service data of the target customer in a second time period, and input the second service data into the pre-trained Bayesian network model to obtain a second probability of the target customer being in each target state at a second time node; wherein the first time node is after the first time period, the second time node is after the first time node, and the second time period is at least a time period between the first time node and the second time node.

[0101] According to an embodiment of the present application, the operation determination module 430 is further configured to determine a risk label of the target customer according to the first probability of each target state, the risk label representing a risk degree of the target customer losing at the first time node; and match an intervention operation corresponding to the risk label in an operation database as the service intervention operation for the target customer; wherein the operation database is configured with a plurality of intervention operations of different types, each intervention operation has a different priority, and the priority of the intervention operation is used to match a corresponding risk label.

[0102] According to an embodiment of the present application, the operation determination module 430 is further configured to input the first service data and the first probability of each target state into the pre-trained Bayesian network model at the same time to obtain a third probability of the target customer being in each target state at the second time node without performing the service intervention operation; compare the second probability and the third probability of each target state to determine a probability change value of each target state in the case of performing the service intervention operation; and adjust the priority of the service intervention operation in the operation database based on the probability change value of each target state.

[0103] According to an embodiment of the present application, the target state includes a stable state, a risk state and a loss state; and the operation determination module 430 is further configured to, if the probability change value of the risk state and / or the loss state represents that the second probability is less than the third probability, improve the priority of the service intervention operation based on the change size of the probability change value; and if the probability change value of the stable state represents that the second probability is greater than the third probability, improve the priority of the service intervention operation based on the change size of the probability change value.

[0104] According to an embodiment of the present application, the operation determination module 430 is further configured to, for each intervention operation in the operation database, input each intervention operation and the first service data into the pre-trained Bayesian network model to obtain a fourth probability of the target customer being in each target state at the first time node if each intervention operation is performed; compare the first probability and the fourth probability of each target state to determine a probability change value of each target state if each intervention operation is performed; and determine a priority of each intervention operation in the operation database based on the probability change value of each target state; wherein each predicted operation is configured to be performed at a third time node, and the third time node is a time node before the first time node.

[0105] According to an embodiment of the present application, the Bayesian network-based customer state prediction apparatus further comprises a model training module.

[0106] The model training module is configured to obtain historical service data of the target customer, wherein the historical service data comprises customer basic information, customer behavior data, service data, service intervention operations for each customer, and customer labels; perform feature processing on each type of service data to obtain feature data of each type, so as to construct a feature data set; and train the Bayesian network model based on the feature data set; wherein the Bayesian network model is constructed based on a state change function, the state change function representing a change relationship between a first probability of each target state of the target customer at a first time node and a second probability of each target state at a second time node, and the state change function being defined by at least each type of feature data.

[0107] According to an embodiment of the present application, the model training module is further configured to add the first service data to the historical service data to obtain updated historical service data, and re-determine each type of feature data based on the updated historical service data to update the feature data set; and adjust the state change function to update the Bayesian network model to obtain an updated Bayesian network model in the case of updating the feature data set.

[0108] According to an embodiment of the present application, any of the plurality of modules of the data acquisition module 410, the first model application module 420, the operation determination module 430 and the second model application module 440 can be combined in one module, or any of the plurality of modules can be split into a plurality of modules. Alternatively, at least part of the functions of one or more of the modules can be combined with at least part of the functions of the other modules, and implemented in one module. According to an embodiment of the present application, at least one of the data acquisition module 410, the first model application module 420, the operation determination module 430 and the second model application module 440 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on chip, a system on board, a system on package, an application specific integrated circuit (ASIC), or any other reasonable manner of integrating or packaging a circuit, etc. in hardware or firmware, or implemented in any one of software, hardware and firmware or in a proper combination of any of the plurality of the above. Alternatively, at least one of the data acquisition module 410, the first model application module 420, the operation determination module 430 and the second model application module 440 can be at least partially implemented as a computer program module which, when executed, can perform the corresponding functions.

[0109] Figure 5 A block diagram of an electronic device suitable for implementing the Bayesian network based customer state prediction method according to an embodiment of the present application is schematically shown.

[0110] As shown in Figure 5 The electronic device 500 according to an embodiment of the present application includes a processor 501 which can perform various appropriate actions and processes according to a program stored in a read only memory (ROM) 502 or a program loaded from a storage portion 508 into a random access memory (RAM) 503. The processor 501 can include, for example, a general purpose microprocessor (e.g. a CPU), an instruction set processor and / or a related chipset and / or a special purpose microprocessor (e.g. an application specific integrated circuit (ASIC)), etc. The processor 501 can also include an on-board memory for cache use. The processor 501 can include a single processing unit or a plurality of processing units for performing different actions of the method processes according to an embodiment of the present application.

[0111] In the RAM 503, various programs and data required for the operation of the electronic device 500 are stored. The processor 501, the ROM 502, and the RAM 503 are connected to each other via the bus 504. The processor 501 performs various operations of the method flow according to the embodiments of the present application by executing the programs in the ROM 502 and / or the RAM 503. It should be noted that the programs can also be stored in one or more memories other than the ROM 502 and the RAM 503. The processor 501 can also perform various operations of the method flow according to the embodiments of the present application by executing the programs stored in the one or more memories.

[0112] According to the embodiments of the present application, the electronic device 500 can further include an input / output (I / O) interface 505, which is also connected to the bus 504. The electronic device 500 can further include one or more of the following components connected to the input / output (I / O) interface 505: an input part 506 including a keyboard, a mouse, and the like; an output part 507 including a cathode ray tube (CRT), a liquid crystal display (LCD), and the like, and a speaker, and the like; a storage part 508 including a hard disk, and the like; and a communication part 509 including a network interface card such as a LAN card, a modem, and the like. The communication part 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to the input / output (I / O) interface 505 as necessary. A removable medium 511 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, and the like is mounted on the drive 510 as necessary, so that a computer program read therefrom is installed in the storage part 508 as necessary.

[0113] The present application also provides a computer readable storage medium, which can be included in the device / apparatus / system described in the above embodiments, or can exist separately without being assembled into the device / apparatus / system. The above computer readable storage medium carries one or more programs, which, when executed, implement the method for predicting customer state based on Bayesian network according to the embodiments of the present application.

[0114] According to an embodiment of the present application, the computer readable storage medium can be a non-transitory computer readable storage medium, for example, can include but is not limited to: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In this application, a computer readable storage medium can be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. For example, according to an embodiment of the present application, a computer readable storage medium can include one or more of the above-described ROM 502 and / or RAM 503, and / or one or more memory devices other than the ROM 502 and the RAM 503.

[0115] Embodiments of the present application also include a computer program product, which includes a computer program containing program codes for executing the method shown in the flow chart. When the computer program product is run in a computer system, the program codes are used to make the computer system implement the Bayesian network-based customer state prediction method provided by the embodiments of the present application.

[0116] The above-described functions defined in the system / device / apparatus of the embodiments of the present application are performed when the computer program is executed by the processor 501. According to an embodiment of the present application, the above-described system, device, module, unit, etc. can be implemented by computer program modules.

[0117] In one embodiment, the computer program can rely on tangible storage media such as optical storage media, magnetic storage media, etc. In another embodiment, the computer program can also be transmitted, distributed, and downloaded in the form of a signal on a network medium, and be downloaded and installed through the communication part 509, and / or installed from the detachable medium 511. The program codes contained in the computer program can be transmitted by any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the foregoing.

[0118] In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 509, and / or installed from the detachable medium 511. When the computer program is executed by the processor 501, the above-described functions defined in the system of the embodiments of the present application are performed. According to an embodiment of the present application, the above-described system, device, apparatus, module, unit, etc. can be implemented by computer program modules.

[0119] According to embodiments of the present application, program code for implementing the computer programs provided by embodiments of the present application can be written in any combination of one or more programming languages, and can be implemented in a high-level procedural and / or object-oriented programming language, and / or in assembly / machine language. Programming languages include, but are not limited to, Java, C++, python, C” language, or similar programming languages. Program code can execute entirely on a user's computing device, partly on a user's device, partly on a remote computing device, or entirely on a remote computing device or server. In the latter scenario, the remote computing device can be connected to the user's computing device through any kind of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computing device, such as through the Internet using an Internet Service Provider (ISP).

[0120] The computer program instructions can also be loaded onto a computer or other programmable information processing apparatus to cause a series of operations to be performed on the computer or other programmable information processing apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable information processing apparatus implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0121] Those skilled in the art will understand that features recited in the various embodiments of the present application can be combined and / or integrated in various combinations, even if such combinations have not been explicitly recited in the present application. In particular, the features recited in the various embodiments of the present application can be combined and / or integrated in various combinations, without departing from the spirit and scope of the present application. All such combinations are within the scope of the present application.

Claims

1. A Bayesian network based method for predicting customer state, characterized by, The method comprises: obtaining first business data of a target customer in a first time period; inputting the first business data into a pre-trained Bayesian network model to obtain a first probability of the target customer being in each target state at a first time node; determining a business intervention operation for the target customer based on the first probability of each target state; in response to the business intervention operation being executed, obtaining second business data of the target customer in a second time period, and inputting the second business data into the pre-trained Bayesian network model to obtain a second probability of the target customer being in each target state at a second time node; wherein the first time node is after the first time period, the second time node is after the first time node, and the second time period is at least a time period between the first time node and the second time node.

2. The method of claim 1, wherein, determining a business intervention operation for the target customer based on the first probability of each target state comprises: determining a risk label of the target customer according to the first probability of each target state, the risk label representing a risk degree of the target customer losing at the first time node; matching an intervention operation corresponding to the risk label in an operation database as the business intervention operation for the target customer; wherein the operation database is configured with a plurality of different types of intervention operations, each intervention operation having a different priority, and the priority of the intervention operation is used to match the corresponding risk label.

3. The method of claim 2, wherein, The method further comprises: inputting the first business data and the first probability of each target state into the pre-trained Bayesian network model simultaneously to obtain a third probability of the target customer being in each target state at the second time node without executing the business intervention operation; comparing the second probability and the third probability of each target state to determine a probability change value of each target state in the case of executing the business intervention operation; adjusting the priority of the business intervention operation in the operation database based on the probability change value of each target state.

4. The method of claim 3, wherein, The target state comprises a stable state, a risk state, and a loss state; adjusting the priority of the business intervention operation in the operation database based on the probability change value of each target state comprises: if the probability change value of the risk state and / or the loss state represents that the second probability is less than the third probability, then increasing the priority of the business intervention operation based on the change size of the probability change value; if the probability change value of the stable state represents that the second probability is greater than the third probability, then increasing the priority of the business intervention operation based on the change size of the probability change value.

5. The method of claim 2, wherein, Before determining the business intervention operation for the target customer based on the first probability of each target state, the method further comprises: for each intervention operation in the operation database, inputting the each intervention operation and the first business data into the pre-trained Bayesian network model to obtain a fourth probability of the target customer being in each target state at the first time node in the case of the target customer executing the each intervention operation; determining a probability change value of each target state in a case where the each intervention operation is performed by comparing the first probability and the fourth probability of the each target state; determining a priority of each intervention operation in the operation database based on the probability change value of each target state; wherein the each predicted operation configuration is executed at a third time node, and the third time node is a time node before the first time period.

6. The method of claim 1, wherein, The Bayesian network model is trained in the following way: obtaining historical business data of a target customer, the historical business data including customer basic information, customer behavior data, service data, business intervention operation for each customer and customer label; respectively performing feature processing on each type of business data to obtain each type of feature data, so as to construct a feature data set; training the Bayesian network model based on the feature data set; wherein the Bayesian network model is constructed based on the following function: a state change function, the state change function representing a change relationship that the first probability of each target state of the target customer at a first time node changes to the second probability of each target state of the target customer at a second time node, and the state change function being defined at least by each type of feature data.

7. The method of claim 6, wherein, After obtaining the first probability of each target state of the target customer at the first time node, the method further comprises: adding the first business data to the historical business data to obtain updated historical business data, and re-determining each type of feature data according to the updated historical business data to update the feature data set; in a case where the feature data set is updated, adjusting the state change function to update the Bayesian network model to obtain an updated Bayesian network model.

8. A Bayesian network-based customer state prediction apparatus characterized by comprising: The device comprises: a data acquisition module for acquiring first business data of a target customer within a first time period; a first model application module for inputting the first business data into a pre-trained Bayesian network model to obtain a first probability of each target state of the target customer at a first time node; an operation determination module for determining a business intervention operation for the target customer based on the first probability of each target state; a second model application module for acquiring second business data of the target customer within a second time period in response to the business intervention operation being executed, and inputting the second business data into the pre-trained Bayesian network model to obtain a second probability of each target state of the target customer at a second time node; wherein the first time node is after the first time period, the second time node is after the first time node, and the second time period is at least a time period between the first time node and the second time node.

9. An electronic device comprising: one or more processors; a memory for storing one or more computer programs, characterized in that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1-7.

10. A computer readable storage medium having stored thereon a computer program or instructions, characterized in that, The computer program or instructions, when executed on a processor, implement the steps of the method according to any one of claims 1-7.

11. A computer program product comprising computer programs or instructions, characterized in that, The computer program or instructions, when executed on a processor, implement the steps of the method according to any one of claims 1-7.

Images